saimm 201507 jul

VOLUME 115 NO. 7 JULY 2015

The MSA Group is a leading provider of

exploration, geology, mining and environmental solutions to the mining industry.

Tel: +27 (0)11 880 4209

Email: [email protected]

www.msagroupservices.com

@TheMSAGroup

themsagroup

/company/the-msa-group

▲

ii JULY 2015 The Journal of The Southern African Institute of Mining and Metallurgy

OFFICE BEARERS AND COUNCIL FOR THE2014/2015 SESSION

Honorary PresidentMike TekePresident, Chamber of Mines of South Africa

Honorary Vice-PresidentsNgoako RamathlodiMinister of Mineral Resources, South AfricaRob DaviesMinister of Trade and Industry, South AfricaNaledi PandoMinister of Science and Technology, South Africa

PresidentJ.L. Porter

President ElectR.T. Jones

Vice-PresidentsC. MusingwiniS. Ndlovu

Immediate Past PresidentM. Dworzanowski

Honorary TreasurerC. Musingwini

Ordinary Members on Council

V.G. Duke T. PegramM.F. Handley S. RupprechtA.S. Macfarlane N. SearleM. Motuku A.G. SmithM. Mthenjane M.H. SolomonD.D. Munro D. TudorG. Njowa D.J. van Niekerk

Past Presidents Serving on CouncilN.A. Barcza J.C. Ngoma R.D. Beck S.J. Ramokgopa J.A. Cruise M.H. Rogers J.R. Dixon G.L. Smith F.M.G. Egerton J.N. van der Merwe G.V.R. Landman W.H. van NiekerkR.P. Mohring

Branch ChairmenBotswana L.E. DimbunguDRC S. MalebaJohannesburg I. AshmoleNamibia N.M. NamateNorthern Cape C.A. van WykPretoria N. NaudeWestern Cape C. DorflingZambia D. MumaZimbabwe S. NdiyambaZululand C.W. Mienie

Corresponding Members of CouncilAustralia: I.J. Corrans, R.J. Dippenaar, A. Croll,

C. Workman-DaviesAustria: H. WagnerBotswana: S.D. WilliamsUnited Kingdom: J.J.L. Cilliers, N.A. BarczaUSA: J-M.M. Rendu, P.C. Pistorius

The Southern African Institute of Mining and Metallurgy

PAST PRESIDENTS

*Deceased

* W. Bettel (1894–1895)* A.F. Crosse (1895–1896)* W.R. Feldtmann (1896–1897)* C. Butters (1897–1898)* J. Loevy (1898–1899)* J.R. Williams (1899–1903)* S.H. Pearce (1903–1904)* W.A. Caldecott (1904–1905)* W. Cullen (1905–1906)* E.H. Johnson (1906–1907)* J. Yates (1907–1908)* R.G. Bevington (1908–1909)* A. McA. Johnston (1909–1910)* J. Moir (1910–1911)* C.B. Saner (1911–1912)* W.R. Dowling (1912–1913)* A. Richardson (1913–1914)* G.H. Stanley (1914–1915)* J.E. Thomas (1915–1916)* J.A. Wilkinson (1916–1917)* G. Hildick-Smith (1917–1918)* H.S. Meyer (1918–1919)* J. Gray (1919–1920)* J. Chilton (1920–1921)* F. Wartenweiler (1921–1922)* G.A. Watermeyer (1922–1923)* F.W. Watson (1923–1924)* C.J. Gray (1924–1925)* H.A. White (1925–1926)* H.R. Adam (1926–1927)* Sir Robert Kotze (1927–1928)* J.A. Woodburn (1928–1929)* H. Pirow (1929–1930)* J. Henderson (1930–1931)* A. King (1931–1932)* V. Nimmo-Dewar (1932–1933)* P.N. Lategan (1933–1934)* E.C. Ranson (1934–1935)* R.A. Flugge-De-Smidt

(1935–1936)* T.K. Prentice (1936–1937)* R.S.G. Stokes (1937–1938)* P.E. Hall (1938–1939)* E.H.A. Joseph (1939–1940)* J.H. Dobson (1940–1941)* Theo Meyer (1941–1942)* John V. Muller (1942–1943)* C. Biccard Jeppe (1943–1944)* P.J. Louis Bok (1944–1945)* J.T. McIntyre (1945–1946)* M. Falcon (1946–1947)* A. Clemens (1947–1948)* F.G. Hill (1948–1949)* O.A.E. Jackson (1949–1950)* W.E. Gooday (1950–1951)* C.J. Irving (1951–1952)* D.D. Stitt (1952–1953)* M.C.G. Meyer (1953–1954)* L.A. Bushell (1954–1955)

* H. Britten (1955–1956)* Wm. Bleloch (1956–1957)* H. Simon (1957–1958)* M. Barcza (1958–1959)* R.J. Adamson (1959–1960)* W.S. Findlay (1960–1961)

D.G. Maxwell (1961–1962)* J. de V. Lambrechts (1962–1963)* J.F. Reid (1963–1964)* D.M. Jamieson (1964–1965)* H.E. Cross (1965–1966)* D. Gordon Jones (1966–1967)* P. Lambooy (1967–1968)* R.C.J. Goode (1968–1969)* J.K.E. Douglas (1969–1970)* V.C. Robinson (1970–1971)* D.D. Howat (1971–1972)

J.P. Hugo (1972–1973)* P.W.J. van Rensburg (1973–1974)* R.P. Plewman (1974–1975)

R.E. Robinson (1975–1976)* M.D.G. Salamon (1976–1977)* P.A. Von Wielligh (1977–1978)* M.G. Atmore (1978–1979)* D.A. Viljoen (1979–1980)* P.R. Jochens (1980–1981)

G.Y. Nisbet (1981–1982)A.N. Brown (1982–1983)

* R.P. King (1983–1984)J.D. Austin (1984–1985)H.E. James (1985–1986)H. Wagner (1986–1987)

* B.C. Alberts (1987–1988)C.E. Fivaz (1988–1989)O.K.H. Steffen (1989–1990)

* H.G. Mosenthal (1990–1991)R.D. Beck (1991–1992)J.P. Hoffman (1992–1993)

* H. Scott-Russell (1993–1994)J.A. Cruise (1994–1995)D.A.J. Ross-Watt (1995–1996)N.A. Barcza (1996–1997)R.P. Mohring (1997–1998)J.R. Dixon (1998–1999)M.H. Rogers (1999–2000)L.A. Cramer (2000–2001)

* A.A.B. Douglas (2001–2002)S.J. Ramokgopa (2002-2003)T.R. Stacey (2003–2004)F.M.G. Egerton (2004–2005)W.H. van Niekerk (2005–2006)R.P.H. Willis (2006–2007)R.G.B. Pickering (2007–2008)A.M. Garbers-Craig (2008–2009)J.C. Ngoma (2009–2010)G.V.R. Landman (2010–2011)J.N. van der Merwe (2011–2012)G.L. Smith (2012–2013)M. Dworzanowski (2013–2014)

Honorary Legal AdvisersVan Hulsteyns Attorneys

AuditorsMessrs R.H. Kitching

Secretaries

The Southern African Institute of Mining and MetallurgyFifth Floor, Chamber of Mines Building5 Hollard Street, Johannesburg 2001P.O. Box 61127, Marshalltown 2107Telephone (011) 834-1273/7Fax (011) 838-5923 or (011) 833-8156E-mail: [email protected]

The Journal of The Southern African Institute of Mining and Metallurgy JULY 2015 ▲iii

ContentsJournal Commentby B. Genc and P. den Hoed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPresident’s Corner by J.L. Porter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi–vii

Spontaneous combustion risk in South African coalfieldsby B. Genc and A. Cook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563Processing low-grade coal to produce high-grade productsby G.J. de Korte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569

Feasibility study of electricity generation from discard coalby B. North, A. Engelbrecht, and B. Oboirien . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573The value proposition of circulating fluidized-bed technology for the utility power sectorby R. Giglio and N.J. Castilla . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581Gasification of low-rank coal in the High-Temperature Winkler (HTW) processby D. Toporov and R. Abraham. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589

Support stability mechanism in a coal face with large angles in both strike and dip by L.Q. Ma*, Y. Zhang, D.S. Zhang, X.Q. Cao, Q.Q. Li, and Y.B. Zhang. . . . . . . . . . . . . . . . . . . 599An economic risk evaluation approach for pit slope optimizationby L.F. Contreras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607Investigation of stress in an earthmover bucket using finite element analysis: a generic model for draglinesby O. Gölbas, ı and N. Demirel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623Radio frequency propagation model and fading of wireless signal at 2.4 GHz in an underground coal mineby A. Patri and D. S. Nimaje . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 629Peak particle velocity prediction using support vector machines: a surface blasting case studyby S.R. Dindarloo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637Large-scale deformation in underground hard-rock minesby E. Karampinos, J. Hadjigeorgiou, P. Turcotte, and F. Mercier-Langevin . . . . . . . . . . . . . . . . 645Visions for challenging assets in the South African coal sector by Z. van Zyl. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653

International Advisory Board

R. Dimitrakopoulos, McGill University, CanadaD. Dreisinger, University of British Columbia, CanadaE. Esterhuizen, NIOSH Research Organization, USAH. Mitri, McGill University, CanadaM.J. Nicol, Murdoch University, AustraliaE. Topal, Curtin University, Australia


Editorial BoardR.D. BeckJ. Beukes

P. den HoedM. Dworzanowski

B. GencM.F. Handley

R.T. JonesW.C. Joughin

J.A. LuckmannC. MusingwiniJ.H. PotgieterR.E. Robinson

T.R. StaceyR.J. Stewart

Editorial ConsultantD. Tudor

Typeset and Published by

The Southern African Institute ofMining and MetallurgyP.O. Box 61127Marshalltown 2107Telephone (011) 834-1273/7Fax (011) 838-5923E-mail: [email protected]

Printed by Camera Press, Johannesburg

Advertising Representative

Barbara SpenceAvenue AdvertisingTelephone (011) 463-7940E-mail: [email protected] SecretariatThe Southern African Instituteof Mining and Metallurgy

ISSN 2225-6253 (print)

ISSN 2411-9717 (online)

THE INSTITUTE, AS A BODY, ISNOT RESPONSIBLE FOR THESTATEMENTS AND OPINIONSADVANCED IN ANY OF ITSPUBLICATIONS.Copyright© 1978 by The Southern AfricanInstitute of Mining and Metallurgy. All rightsreserved. Multiple copying of the contents ofthis publication or parts thereof withoutpermission is in breach of copyright, butpermission is hereby given for the copying oftitles and abstracts of papers and names ofauthors. Permission to copy illustrations andshort extracts from the text of individualcontributions is usually given upon writtenapplication to the Institute, provided that thesource (and where appropriate, the copyright)is acknowledged. Apart from any fair dealingfor the purposes of review or criticism underThe Copyright Act no. 98, 1978, Section 12, ofthe Republic of South Africa, a single copy ofan article may be supplied by a library for thepurposes of research or private study. No partof this publication may be reproduced, storedin a retrieval system, or transmitted in any formor by any means without the prior permissionof the publishers. Multiple copying of thecontents of the publication without permissionis always illegal.

U.S. Copyright Law applicable to users In theU.S.A.The appearance of the statement of copyrightat the bottom of the first page of an articleappearing in this journal indicates that thecopyright holder consents to the making ofcopies of the article for personal or internaluse. This consent is given on condition that thecopier pays the stated fee for each copy of apaper beyond that permitted by Section 107 or108 of the U.S. Copyright Law. The fee is to bepaid through the Copyright Clearance Center,Inc., Operations Center, P.O. Box 765,Schenectady, New York 12301, U.S.A. Thisconsent does not extend to other kinds ofcopying, such as copying for generaldistribution, for advertising or promotionalpurposes, for creating new collective works, orfor resale.

GENERAL PAPERS AND TECHNICAL NOTE

COAL CONFERENCE PAPERS

IFSA CONFERENCE PAPERS


▲

iv JULY 2015 The Journal of The Southern African Institute of Mining and Metallurgy

A blessing at the best of times, electricity has becomethe bane of life in South Africa: it adds quality tolife, but when supply is erratic, as we all know too

well, the effects cripple and evoke anger. Constraints in thesupply of electricity are damaging the economy. Somepredictions even foretell a crisis of monumental proportions.Eskom can barely meet current demand. It will fall short ofgrowing demand in the near future. There are simply too fewpower stations to meet the country’s current and growingenergy needs. Building new power stations—an obvioussolution—takes years, so there is no quick fix. Making mattersworse is the sharply rising cost of electricity. South Africanswill be digging deeper into their pockets, and industry will seeits operating costs increase. There is, in short, a debilitatingdisparity between the provision of power, on the one hand,and needs and expectations on the other. The disparity raisesdistressing alarms in many quarters.

What, you might ask, does this alarming andembarrassing predicament have to do with coal, the theme(one could say) of the papers appearing in this issue of thejournal? Coal is vital to the South African economy. Exported,it exceeds all other commodities in bulk and revenue.1 It is allbut integral to the generation of electricity in South Africa:more than 95% of our electricity is generated from coal. Yetsome critics, notably the American environmentalist andauthor Bill McKibben, have argued that—

‘There is an urgent need to stop subsidizing the fossil fuelindustry, dramatically reduce wasted energy, andsignificantly shift our power supplies from oil, coal, andnatural gas to wind, solar, geothermal, and otherrenewable energy sources.’2

South Africa has heeded the call. It, too, is committed toreducing carbon emissions: in the next 50 years coal will dropto 20% of the energy mix; the rest of demand will be met bynuclear and renewable energy.3 But half a century is a longtime, and all the while in this country electricity will continueto be generated from coal. Coal mining, processing, andcombustion, in all likelihood, will be with us for a long time.This, however, does not mean that the old practices can orshould continue as before.

Two conferences last year attracted papers that addressedquestions of change and challenge in coal mining and powergeneration. Both conferences were held in Johannesburg. Thefirst one focused on ‘21st century challenges to the southernAfrican coal sector’4. Coal mining, as for mining in any sector,faces new challenges as high-grade, readily accessible seamsare mined out, the grade of resources declines, and operationsswitch to coalfields in more remote locations. Attention andefforts are now directed at mining thin seams, beneficiatingfines, and transporting bulk material from locations that liesome distance from available routes. These and otherchallenges were highlighted and discussed in 22 paperspresented at the conference, three of which appear in thisissue of the journal.

Two of the papers presented at the coal conferencediscussed a matter at the heart of a sub-theme of the secondconference.5 IFSA 2014, a conference on industrial fluidization

and fluid-bed technologies, was held towards the end of theyear. It is held every three years; this one was the fifth in theseries. Although ‘industrial’ appears in its title, the paperspresented over the two days covered topics both fundamentaland applied. A scan of the titles of papers listed in theproceedings, however, reveals a telling bias: only six of the 28papers covered topics other than the combustion orgasification of a variety of carbonaceous fuels (coal, includingdiscard coal, biomass, and oil). The three papers selected forinclusion in this issue of the journal discuss three elements ofthe subject: namely, the role of fluid-bed technologies (1) inconverting coal of all grades (2) into electricity or syngas (3).

Standard textbooks list the advantages of fluidized beds inthe design of reactors. Circulating fluidized beds (or CFBs)operate in a regime called fast fluidization: the mixing ofparticles, which aggregate in clusters that break apart andreform, is extensive and slip velocities (between gas andsolids) are an order of magnitude greater than the terminalvelocity of the particles. Higher fluidizing velocities lead to thepneumatic transport of particles. CFBs confer advantages onthe burning of carbonaceous fuels for power generation:6 agiven design can stably burn a variety of fuels in type (coal orbiomass) and quality; the solid fuel does not need to bepulverized or necessarily dried; temperatures are uniform andheat transfer is even; limestone captures SO2 in the bed, whichdoes away with the need for flue-gas desulphurization; andbecause temperatures are lower (850°C, compared with 1500°Cin conventional pulverized-fuel boilers) ash does not melt orslag; and the formation of NOx is minimal, which does awaywith the need for selective catalytic reduction (or SCR). Theseadvantages impart flexibility, save money, and significantlylessen hazardous components in flue-gas emissions.

The committee drafting the Integrated Resource Plan (IRP2010) recognized these advantages when it recommended theprocurement of fluid-bed boilers for the burning of high-ashdiscard coals. Estimates put this resource at about 1.5 billiontons. This measure would go a long way in utilizing a resourcethat has been cast aside and, in doing so, bring some redressto the energy needs of this country. In the meantime loadshedding, higher prices for electricity, and constraints insupply remain a burden. Nevertheless, unless we act now,implementing measures that will work for a country that isrich in coal, the future will be a bleak one.

B. Genc and P. den Hoed

Journal CommentHeaping coals of fire upon our heads

1This and many other facts in the editorial piece are quoted from thepreface and papers of the proceedings of IFSA 2014.

2W.E. McKibben3Department of Energy, Integrated Resource Plan for Electricity2010–2030, Revision 2, Final report, promulgated on 25 March 2011;Integrated Resource Plan for Electricity (IRP) 2010–2030, Updatereport, 21 November 2013

4This conference styled itself as a symposium.5‘The future of coal in the low-carbon economy and the impact of natural

gas’ by Dave Collins (MAC Consulting) and ‘South Africa’s answers toclimate change: challenges and opportunities in clean-coaltechnologies’ by Professor Rosemary Falcon (University of theWitwatersrand).

6See, for example, ‘The value proposition of circulating fluidized-bedtechnology for the utility power sector’ in this issue of the Journal.

▲

vi JULY 2015 The Journal of The Southern African Institute of Mining and Metallurgy

In this month’s article I want to update readers about two important aspects of the Institute –firstly, the launch of the new Botswana Branch, and secondly the Mineral EconomicsDivision of the SAIMM.

SAIMM Botswana Branch

The process to set up a Branch in Botswana began in 2006. Alan Clegg, a Council memberand the Chairman of the Regional Branch Organizing Committee at the time, was

instrumental in establishing and setting the process in motion. The Botswana Government requiresprofessional societies such as ours to be registered with the Registrar of Societies. As with any process that has to

be done from a distance without being able to visit the respective departments and people personally, this was quite a lengthytask and took a number of years before it was completed. In the meantime, James Arthur was elected as the Branch Chairman.Although presentations were set up in four main centres, logistics were a problem due to the large geographical area which hadto be covered. The Base Metals Conference, organized through the Technical Programme Committee: Metallurgy, was a highlightduring 2008. It was held at the Mwana Lodge in Chobe.

James Arthur had to return to South Africa, and unfortunately there were no further activities planned.Recently the SAIMM saw the need for a person who would grow the regional Branches, and in October 2014 we employed

Malcolm Walker as Regional Development Manager. One of Malcolm’s first tasks was to speak to people in Botswana with aview to resuscitating the Branch. We soon realized that the Branch would need to be started afresh, and Malcolm went toBotswana to personally meet people in the industry and to see if there was an appetite for an SAIMM Branch.

In terms of By-law F, which guides the operation of Branches, twelve corporate members are required to submit a writtenrequest to the Council of the SAIMM for the establishment of a Branch. Members in Botswana were sent an e-mail to find out ifthey were interested and at least twenty responses were received. This was seen as a good sign, as there are 77 members of theSAIMM in Botswana.

After further planning and arrangements, it was agreed that the Branch would be launched in conjunction with a technicalvisit. This took place on 5 June, when a visit to the Diamond Trading Company Botswana was arranged, coupled with thelaunch of the Botswana Branch and the election of the Committee. I am pleased to say that this went well, and a Chairman andCommittee were elected as follows: Chairman – Len Dimbungu, Vice Chairman – Andries Bester, Secretary – Craig Robertson,and three Committee Members – Michael Musonda, Omphile Ntabeni, and Wiesiek Masztalerz.

Malcolm and I also met with the Chairman of the Botswana Chamber of Mines, Charles Siwawa, to discuss the registrationprocess and the role that the SAIMM can play in the local industry. I am confident that with assistance from Charles and theenthusiasm of the Branch Committee, we will be arranging a number of technical presentations and conferences in Botswana.

The Mineral Economics Division

The Mineral Economics Division of the SAIMM was established to keep a watching brief on the changing nature of mining andits interface with the political economics of the resource-rich countries in Southern Africa. A workshop was held in February2012 entitled ‘Towards a Multi-stakeholder Dialogue on Critical Issues facing the Southern African Mining Industry’ to furtherexamine and discuss these issues. A key component of the workshop was the session organized and led by Mike Solomon, theChairman of the Mineral Economics Division, on the rise of resource nationalism. This resulted in the publication of the report‘The Rise of Resource Nationalism: a Resurgence of State Control in an Era of Free Markets or the Legitimate Search for a NewEquilibrium?’, which was subsequently presented at the Mining Indaba.

The dialogues came about as a direct result of a comprehensive, academically sound study consisting of an in-depth lookinto the issues related to state participation in the mining sector from a global and historical perspective. The objective was toinform national debate through rigorous and exhaustive research that would provide a platform for evidence-based dialogue.

Following the inaugural three-day dialogues hosted in 2012, Mining Dialogues 360° (MD360) produced a summary reportof the key issues identified by the participants, which is available on their website (www.miningdialogues360.co.za). Theorganization continued to engage with various constituencies, and in the wake of Marikana hosted further dialogues, mostnotably with key members from civil society organizations and the various church bodies that are active in the miningcommunities. The outcome of the one-day dialogue was a set of five key recommendations for King Leruo Molotlegi of theRoyal Bafokeng (who provided much of the funding for the work) to present at a meeting of CEOs of the affected platinumcompanies. In 2013, the Mining Dialogues research team completed the first of a series of in-depth studies into the social andeconomic footprints of each of South Africa´s major mining sectors.

In addition, MD360 participated in think tanks hosted by other industry bodies and has forged co-operative alliances withthe ICMM, the Centre for Sustainability in Mining, the Africa Futures Forum, the WEF Global Agenda Council, and the RoyalInstitute of International Affairs (also known as Chatham House).

A full merger of Mining for Change and MD360 took place in early 2014. The consolidation of these entities into a singlestreamlined platform created a strengthened organization with networks across the South African mining and regulatorysectors, with obvious benefits such as not competing for funding, more efficient staffing, and reduced management andoverhead costs.

President’s

Corner

The Journal of The Southern African Institute of Mining and Metallurgy JULY 2015 ▲vii

Mining Dialogues 360° is held under the auspices of the SAIMM and enjoys the support of the South African Chamber ofMines, the International Council on Mining and Metals (ICMM), the Centre for Sustainability in Mining and Industry (CSMI),The Royal Institute of International Affairs (Chatham House), the universities of the Witwatersrand and Stellenbosch, theWorld Economic Forum´s Mining and Minerals Council, the International Institute for Sustainable Development (IISD), theUnited Nations Environment Programme (UNEP), the mining industry, labour bodies, civil society organizations, governmentand related institutions, and the investment community. MD360 is a not-for-profit company (NPC) that operates with thesupport of grants and financial contributions.

Key milestones and achievements – 2010 to present2010 - Research commissioned by the SAIMM and funded by the Royal Bafokeng into resource nationalism in the global

context and issues in South Africa.2011 - Produced an extensive, in-depth report titled ‘The Rise of Resource Nationalism: a Resurgence of State Control in

an Era of Free Markets or the Legitimate Search for a New Equilibrium? A Study to Inform Multi-stakeholder Dialogue on State Participation in Mining’.

- February: the report is presented at the Mining Indaba to much acclaim from the industry, resulting in calls for thedocument to form the basis of focused dialogues on the issues.

2012 - April: Mining Dialogues 360 Degrees (MD360) is formed as a not-for-profit company.- July: the inaugural 3-day meeting is held with the dialogues oversubscribed, and the participatory format was

received exceptionally well by all stakeholders. - A website was created to form an ongoing communications platform and central repository of information. The

platform has been maintained and is updated on a daily basis. http://www.miningdialogues360.co.za- The final meeting report predicts a major industry disaster. Almost exactly one month later , in August, Marikana

erupts.- October: in order to provide leadership to tackle the issues, MD360 calls a dialogue between 15 key participants

from civil society and faith-based organizations. This think tank resulted in a five-point strategy for Kgosi Leruo topresent at a meeting called with the platinum CEOs to identify and seek solutions to the crisis. The points are verywell received but again, no action is taken by industry.

- November: Lonmin commissions MD360 to do a ‘deep-dive’ research piece on the social and economic footprint ofthe company at its North West operations.

2013 - September: the report is completed.- Other platinum companies show great interest in participating in the footprint exercise in order to create a wider

view and understanding of the landscape.- The independent recommendations of the research team are documented in a paper titled ‘A Platinum Compact’

and shared with the highest levels of government and key advisors, including Pravin Gordhan, Godfrey Oliphant, Musa Mabuza, Roger Baxter, and Gwede Mantashe.

2014 - MD360 is approached by the Farlam Commission to assist with information relating to Lonmin. Owing to an NDAsigned with Lonmin, MD360 is limited in what can be shared with the commission. Judge Farlam subpoenas thereport from Lonmin and it is widely quoted in the Phase 2 findings of Dr Kally Forrest.

Current - In light of the lack of progress with the government’s Framework for Sustainability in Mining agreement, and inresponse to numerous calls for an industry forum that is properly representative of all stakeholders (not justgovernment, labour, and industry), MD360 has developed the terms of reference for a new three-year researchwork and dialogue programme.

- The programme has received the approval of the SAIMM and has the support of the ICMM, CSMI, the universitiesof the Witwatersrand and Stellenbosch, the Royal Institute of International Affairs (Chatham House), and theWorld Economic Forum’s Mining and Minerals Council.

- The MD360 Board has appointed a new, influential and high-profile Advisory Council to oversee the programme and provide guidance on the issues.

- The Platinum Compact recommendations are under serious consideration by the Emergency Task Response Teamfor Mining under the oversight of Minister Radebe in the Office of the Presidency.

- The organization is currently engaged in fundraising to support the work programme and dialogues.The above information is taken from the various documents available on MD360 and is therefore presented very factually.

My own involvement has been as the Chairman of the Advisory Council. I am committed to the work being done by MD360and I encourage the industry to support the various initiatives. It is only by interested parties’ contributing to the discussions,and where possible funding the initiatives, that we will see a positive change in our industry. You are welcome to send yourcomments and questions to me so that we can continue meaningful debate.’

J.L. PorterPresident, SAIMM

PAPERS IN THIS EDITIONThese papers have been refereed and edited according to internationally accepted standards and

are accredited for rating purposes by the South African Department of Higher Education andTraining

These papers will be available on the SAIMM websitehttp://www.saimm.co.za

Coal Conference PapersSpontaneous combustion risk in South African coalfieldsby B. Genc and A. Cook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563Laboratory tests have been undertaken for five consecutive years in order to determine both the Wits-EHAC index and the crossing point temperature of 119 coal samples.These parameters, when combined, give an indication of thespontaneous combustion propensities of the samples. The database of results, which is continually being updated, provides the basis for an improved risk evaluation methodology for spontaneous combustion.

Processing low-grade coal to produce high-grade productsby G.J. de Korte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 569The coals found in the Waterberg, Soutpansberg, and other coalfields are of relatively low quality compared to the coal mined from the traditional areas of the Witbank, Highveld, and Ermelo coalfields. Processing low-yielding coals into good-quality products while ensuring that coal mining remains economically viable will require the investigation and implementation of more cost-effective coal processing technologies.

IFSA Conference PapersFeasibility study of electricity generation from discard coalby B. North, A. Engelbrecht, and B. Oboirien . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573A detailed economic assessment of the feasibility of electricity generation from discard coal, comprising material and energy balances and the construction of a discounted cash flow (DCF) table, is presented, showing that the process is potentially attractive from an economic perspective.

The value proposition of circulating fluidized-bed technology for the utility power sectorby R. Giglio and N.J. Castilla . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581This paper presents an outlook for future coal supply, quality, and price, as well as a review of the technical and economic benefits of circulating fluidized-bed technology when firing low-quality fuels for utility power generation.

Gasification of low-rank coal in the High-Temperature Winkler (HTW) processby D. Toporov and R. Abraham . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589The High-Temperature Winkler (HTW) gasification process is designed to utilize low-rank feedstock such as coals with a high ash content, lignite, or biomass. The process is characterized by a bubbling fluidized bed, where coal devolatilization and partial oxidation and gasification of coal char and volatiles take place and by a freeboard, where partial combustion and gasification of coal char take place.

PAPERS IN THIS EDITIONThese papers have been refereed and edited according to internationally accepted standards and

are accredited for rating purposes by the South African Department of Higher Education andTraining

These papers will be available on the SAIMM websitehttp://www.saimm.co.za

General Papers and Technical NoteSupport stability mechanism in a coal face with large angles in both strike and dip by L.Q. Ma, Y. Zhang, D.S. Zhang, X.Q. Cao, Q.Q. Li, and Y.B. Zhang . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599A mechanical model has been developed that considers the impact of the coal seam dip angle on the support stability in the strike direction. The research findings were successfully applied to a fully mechanized coal face with large angles both in strike and dip at the Xinji Coal Mine in China.

An economic risk evaluation approach for pit slope optimizationby L.F. Contreras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607An alternative methodology of pit slope design is proposed, where the economic impacts of potential slope failures are calculated and used as the elements on which to apply the acceptability criteria for design. The paper discusses the concepts used for interpreting the probability of slope failure and describes an approach for estimating the economic impacts of slope failure by construction of a ‘risk map’.

Investigation of stress in an earthmover bucket using finite element analysis: a generic model for draglinesby O. Gölbas, ı and N. Demirel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623This study aims to develop a generic finite-element model of the stress on an operating dragline bucket. Simulation work and sensitivity analyses provide the indicators of failure and stress values.

Radio frequency propagation model and fading of wireless signal at 2.4 GHz in an underground coal mineby A. Patri and D. S. Nimaje . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 629A novel method is proposed for determining the parameters of a suitable radio propagation model, and is illustrated with the results of a practical experiment carried out in an underground coal mine in Southern India.

Peak particle velocity prediction using support vector machines: a surface blasting case studyby S.R. Dindarloo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637In this study, the support vector machine algorithm was employed for prediction of the peak particle velocity (PPV) induced by surface mine blasting. The major advantages of the method are the very high accuracy of predictions and fast computation times.

Large-scale deformation in underground hard-rock minesby E. Karampinos, J. Hadjigeorgiou, P. Turcotte, and F. Mercier-Langevin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645Field observations and convergence measurements at two underground mines in Canada were used to provide guidelines of the anticipated squeezing ground levels at these operations. The choice of a favourable angle ofinterception between the drift and the foliation can result in a more manageable squeezing level and increase the performance of an appropriate support system for squeezing ground conditions.

Visions for challenging assets in the South African coal sector by Z. van Zyl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653This paper, which is based on research on trends in underground coal mining as well as 16 years’ practical experience in electronic monitoring of mining machinery and productivity optimization, illustrates the benefits of moving from reactive event-based management systems towards a more adaptable and flexible process-based system

Introduction The risk of spontaneous combustion is wellknown in the South African coal miningindustry. All major coal producers in SouthAfrica, such as Anglo Thermal Coal, BHPBilliton Energy Coal South Africa (Becsa),Exxaro, and Xstrata Coal or their predecessorshave experienced spontaneous combustionincidents in their history. As stated by Phillipset al. (2011) there is always the risk ofspontaneous combustion in undergroundmining (e.g. the Goedehoop fire in 2008), butspontaneous combustion can take place onboth underground and surface coal mines. Thecurrent problem is in surface mines and nearlyalways in mines extracting previously workedseams i.e. where old bord and pillar workingsare exposed. In the coming years it is very

possible that the rate of spontaneouscombustion will increase from its present lowlevels, due to factors such as higher ventilationpressures, an increased rate of mining, moreworking of previously mined seams, etc.

It is also fairly certain that coal mining willface tougher environmental legislation limitingemissions in the near future. To ascertain theareas where spontaneous combustion risks arehigh, it is necessary to improve currentlaboratory procedures for testing andevaluating coal samples, combine the resultwith site and field data, and if necessary revisethe laboratory rating system to better reflectSouth African conditions.

The current laboratory tests are conductedin order to determine both the Wits-EHACindex and the crossing-point temperature,which are combined to obtain the propensitiesof the coal samples to undergo spontaneouscombustion. This has resulted in a database ofresults to review and evaluate South Africancoal seams. Using this database, the high-riskareas in terms of spontaneous combustion canbe identified. The tests, involving 119samples, cover five consecutive years, between2008 and 2012. The samples were from a widevariety of different coal seams and producingcoalfields. All samples have been subjected toa series of laboratory tests, and the resultsanalysed. A comprehensive database of theseresults is available, and is being continuallyupdated as new test results are added.

The spontaneous combustion testAt the School of Mining Engineering at theUniversity of the Witwatersrand (Wits), an

Spontaneous combustion risk in SouthAfrican coalfieldsby B. Genc* and A. Cook†

SynopsisThe risk of spontaneous combustion is well known in the South Africancoal mining industry. In the coming years it is very possible that theincidence of spontaneous combustion will increase from current levels, dueto factors such as an increased rate of mining, re-working of previouslymined seams, more stooping and total extraction for underground mines,and higher stripping ratios for surface mines, leading to more spoils. It isalso fairly certain that coal mining will face tougher environmentalemissions legislation in the near future. To determine the areas where therisks of spontaneous combustion are high, it is necessary to improve onour current laboratory procedures for testing and evaluating coal samples,combining the results with site and field data, and if necessary revising thelaboratory rating system to refine our understanding of South Africanconditions.

Currently, laboratory tests are conducted in order to determine boththe Wits-EHAC index and the crossing-point temperature which, whencombined, give an indication of the spontaneous combustion propensitiesof the coal samples. This procedure has enabled the establishment of adatabase of results to review and evaluate South African coal seams.Using this database, the high-risk areas in terms of spontaneouscombustion are identified. Tests have been undertaken for five consecutiveyears, between 2008 and 2012. In total, 119 coal samples from differentcoal seams and production coalfields have been analysed and classifiedthrough a series of laboratory tests. A comprehensive database of theseresults is available, and is continually being updated as new test resultsare added. This database will continue to expand, and to provide the basisfor an improved risk evaluation methodology for spontaneous combustion.

Keywordscoal spontaneous combustion, risk assessment, Wits-EHAC liability index,crossing-point temperature.

* University of the Witwatersrand, Johannesburg.† Latona Consulting Pty. Ltd., Johannesburg.© The Southern African Institute of Mining and

Metallurgy, 2015. ISSN 2225-6253. This paperwas first presented at the 21st Century challengesto the southern African coal sector, 4–5 March2014, Emperors Palace, Hotel Casino ConventionResort, Johannesburg.

563The Journal of The Southern African Institute of Mining and Metallurgy VOLUME 115 JULY 2015 ▲

http://dx.doi.org/10.17159/2411-9717/2015/v115n7a1

Spontaneous combustion risk in South African coalfields

apparatus was developed nearly 30 years ago to measure thepropensity of coal to undergo spontaneous combustion. Thisresearch was funded by the Government Mining Engineer’sExplosion Hazard Advisory Committee (EHAC). Theapparatus is used to test coals under predefined conditions,and a combustibility index (Wits-EHAC) is obtained.Although the propensity of coal to combust spontaneouslycan be determined using various laboratory techniques,ignition temperature tests are commonly used to studyspontaneous combustion, as they yield rapid results. Ignitiontemperature tests use two methods:

➤ Crossing-point temperature (XPT)➤ Differential thermal analysis (DTA)

to determine both the Wits-EHAC index and the XPT. TheWits-EHAC index is defined as:

Wits-EHAC index = (Stage II slope/XPT) × 500 (Gouws,1987)

When the temperature differential between a coal sampleand an inert sample is plotted against the inert temperature,the portion of the graph where the coal is heating morerapidly than the inert sample, i.e. where an exothermicreaction is taking place, is referred to as Stage II.

According to Gouws (1987), the characteristics of thecurves plotted using the obtained results (i.e. ignitiontemperature tests) are used to determine the propensity ofcoal for self-heating, and this is the basis for the Wits-EHACliability index. It is important to understand that when anindex value of coal is greater than five, there is a highpropensity for spontaneous combustion, and when an indexvalue is less than three, there is a low propensity forspontaneous combustion. An index value of between threeand five indicates that the coal sample has a relativelymedium risk of undergoing spontaneously combustion. Asindicated in Table I, a higher index value represents a higherrisk of a coal self-heating (Gouws, 1987).

The testing apparatus used for the Wits-EHAC indexconsists of an oil bath, six coal and inert material cellassemblies, an oil circulator, a heater, a flow meter used forair flow monitoring, an air supply compressor, and acomputer. The temperatures are recorded every 20 seconds by

the microcomputer during an average of 3–4 hours’ testingtime.

Detailed information regarding the testing apparatusused, as well as the testing procedure, is well documented byGenc et al. (2013).

ResultsThe tests were done over five consecutive years, between2008 and 2012, with the spontaneous combustion liabilityindex being obtained for all 119 samples. Table II shows thesummary of the results. During this period there were nolow-risk samples. Figure 1 represents graphically the totalnumber of tests in terms of medium and high propensity.

Table III shows the results for spontaneous combustiontests in 2008 when 15 tests were conducted. The test resultsinclude the XPT in degrees Celsius (°C) and the Wits–EHACindex. The names of the mines have been abbreviated. All ofthe coal samples tested produced results that ranged frommedium to high propensity to spontaneous combustion, withan almost 50/50 split between medium (8) and high (7)propensity. The minimum calculated Wits–EHAC index was

▲

564 JULY 2015 VOLUME 115 The Journal of The Southern African Institute of Mining and Metallurgy

Table II

Spontaneous combustion test results between2008 and 2012

2008 2009 2010 2011 2012 Total

High 7 13 20 6 6 52Medium 8 5 7 22 25 67Total 15 18 27 28 31 119

Table I

Spontaneous combustion liability index

Index Spontaneous combustion liability

0-3 Low3-5 Medium> 5 High

Table III

Spontaneous combustion test results (2008)

Mine Wits-EHAC Crossing-point Spontaneous index temperature combustion

(°C) liability

GB 4.87 127.4 MediumGB 5.03 125.1 HighUm 5.22 119.3 HighAt 4.31 126.7 MediumMS 5.41 126.7 HighMS 5.51 132.3 HighMS 5.8 129.7 HighKl 5.55 126.7 HighMa 5.15 124.9 HighGe 3.65 153.5 MediumTw 4.09 133.6 MediumUi 4.86 128 MediumMb 4.96 131.6 MediumSS 4.55 130.9 MediumSL 4.9 110.4 Medium

Figure 1 – Spontaneous combustion liability test results (2008–2012)

3.65, and the maximum 5.8. Figure 2 shows the 2008 testresults for spontaneous combustion liability. The Wits-EHACindex margins are indicated using colours. The yellow partshading indicates the coal samples that have a value of morethan 3 but less than 5, which are thought to possess amedium risk to spontaneous combustion. The red shadingindicates samples that have a value of more than 5 and are,therefore, thought to have a high risk of spontaneouscombustion.

Table IV shows the 2009 tests results for spontaneouscombustion, when 18 tests were conducted. All of the testedcoal samples showed medium to high propensity to sponta-neously combust, and most of the collieries have highpropensity (13 out of 18). The minimum calculatedWits–EHAC index was 4.14, and the maximum 5.72. Figure 3shows the 2009 test results for spontaneous combustionliability.

Table V shows the 2010 tests results for spontaneouscombustion, when 27 tests were conducted. The 2010 resultsshow a very similar trend to the 2009 results, as most of thecollieries tested had results in the high propensity range (20out of 27). The minimum calculated Wits–EHAC index was4.64 and the maximum 5.64. Figure 4 shows the 2010 testresults for spontaneous combustion liability.

Table VI shows the 2011 tests results for spontaneouscombustion, when 28 tests were conducted. Although therange was similar to that previously observed (i.e. medium tohigh), most of the collieries tested had results in the mediumpropensity range (22 out of 28). The reason for the differencein the test results from one year to the next is because everycoal seam has different physical and chemical properties, andthese impact on its propensity for spontaneous combustion.The minimum calculated Wits-EHAC index was 3.1, which is

still just above the low range identified by Gouws (1987),while the maximum index was 5.91. Figure 5 shows the2011 test results for spontaneous combustion liability.

Finally, Table VII shows the 2012 tests results ofspontaneous combustion, when 31 tests were conducted. The2012 results showed a very similar trend to 2011, as most ofthe collieries tested had a medium propensity (25 out of 31).The minimum calculated Wits-EHAC index was 3.71 and themaximum 5.75. Figure 6 shows the 2012 test results forspontaneous combustion liability.



Figure 2 – Spontaneous combustion liability test results, 2008


Table IV



(°C) liability

ND 4.24 127.6 MediumSl 5.72 120.2 HighBo 4.79 115.3 MediumMo 5.09 123.3 HighM1 5.01 108.7 HighPo 5.27 136.4 HighOp 5.41 119.3 HighSp 5.31 111.9 HighSp 4.88 120.8 MediumSp 5.46 119.9 HighBw 5.35 111.4 HighPa 4.14 133.4 MediumDR 5.31 121.9 HighSs 4.53 130.9 MediumM2 5.62 119.2 HighM3 5.02 116.4 HighM4 5.18 121 HighM5 5.33 118.4 High


▲


Table V



(°C) liability

Gr4 5.64 98.6 High

Gr5 5.44 102.6 High

Xs5 5.31 105.2 High

Ma 5.14 109.3 High

Ta 5.47 110.8 High

SW 5.26 117.3 High

Sl 5.58 104.8 High

D16 4.91 118.8 Medium

D15 5.51 102.9 High

Si 5.32 113.7 High

M1 4.64 117.8 Medium

NC 5.02 119.3 High

nK 5.38 108 High

Op 5.49 108.2 High

Op 5.23 120.9 High

Op 5.1 110.8 High

Op 5.52 113.6 High

Op 5.27 117 High

Wy 5.24 114.7 High

DE 5.6 121.2 High

VaA 5.58 95.2 High

KE 4.86 120.2 Medium

KW 4.71 113.9 Medium

Xs2 4.91 116.7 Medium

SW 4.92 121.9 Medium

VaG 4.86 108.5 Medium

Mo 5.33 124.7 High

Table VI



(°C) liability

COA 3.14 161.9 MediumCOA 3.46 150.5 MediumCOA 3.58 154.3 MediumTshM 3.63 145.8 MediumTshV 3.65 144.8 MediumTshG 3.76 137.4 MediumKD 4.21 133.2 MediumM3 4.21 134.6 MediumKG 4.3 130.3 MediumSl 4.41 129.4 MediumSp 4.56 121.2 mediumSW 4.62 126.9 MediumGGV4 4.66 128.1 MediumKr 4.67 126.2 MediumTa 4.73 124.3 MediumTu 4.73 129.3 MediumGr4 4.74 118.7 MediumSp 4.79 117.3 mediumKr 4.81 114.2 MediumKr 4.81 114.2 MediumCOA 3.86 131.3 MediumGr5 4.9 112.9 MediumDr 5.01 118.7 HighGGV2 5.12 123.2 HighM2 5.18 125.9 HighM1 5.33 123.4 HighM1 5.36 126.8 HighM1 5.91 127.8 High



Analysis To address spontaneous combustion problems in SouthAfrican collieries, 119 tests were conducted over fiveconsecutive years and the results were rated according to theWits-EHAC index. The analysis shows that none of thesamples fell in the low-risk category. Sixty-seven out of 119tested collieries possess a medium risk of spontaneouscombustion (almost 56.3 per cent), while the remainingcollieries (43.7 per cent) possess a high risk of spontaneouscombustion (about a 13 per cent difference). Figure 7 showsthe propensity for spontaneous combustion percentages forall 119 tests.

The results indicate that spontaneous combustionpropensity is dependent on the properties of each coal seam.In 2010, when the largest high-risk rating percentage wasrecorded, test results showed that more than 74 per cent ofthe mines were in this high-risk category, compared withabout 21 per cent in 2011 and 19 per cent in 2012. In 2009,similar to 2010, 72 per cent of the mines had high riskratings. In 2008, the high and medium rating percentageswere very close; 47 and 53 per cent, respectively.

The results show that, given the right environmentalconditions, most of the collieries located in the Witbank andHighveld coalfields have a high risk of spontaneouscombustion. The high-risk areas also include the north-eastern part of Ogies. The southern parts of Witbankcoalfield, in the Ermelo area, are also rated high in terms ofrisk. Although 52 out of 119 collieries have a high inherentrisk of spontaneous combustion, most of the selectedcollieries possess medium risk ratings. Medium risk ratingscan be seen around the northern parts of Ermelo, as well asin KwaZulu-Natal Province where anthracite coal is mined.

It is interesting that there were no low-range resultsrecorded during the five-year testing period from 2008 to2012. This finding indicates that there is a need to re-visitthe current definition of the spontaneous combustion liabilityindex within the ranges of low, medium, or high; but thisrequires a further study as to how Gouws (1987) definedthese ranges and, if it is necessary to change the currentdefinitions, at what levels should the new criteria should beset.


The Journal of The Southern African Institute of Mining and Metallurgy VOLUME 115 JULY 2015 567 ▲


Table VII



(°C) liability

AA 4.23 139.6 MediumAA 4.29 141 MediumAA 4.15 142.9 MediumG8 4.53 131.4 MediumGr4 4.73 132.5 MediumKr 5.5 128.2 HighKh5 4.84 133.7 MediumAr8 4.78 132.1 MediumAr8 4.41 127.2 MediumAr8 4.06 138.3 MediumAr8 5.75 126 HighAr10 5.7 127.9 HighKiD 5.05 109 HighKiD 4.25 137.8 MediumKiD 4.8 125.1 MediumKiD 4.3 139.6 MediumVeD 4.43 117.9 MediumVeC 4.49 128.2 MediumVeT 4.62 119.3 MediumVeR 4.19 134.8 MediumMoW 4.75 119.9 MediumVu 5.26 124.8 HighMo 3.93 138.3 MediumTCM 5.11 128.1 HighDCM 4.56 125.1 MediumFZN 4.78 127.8 MediumFZS 4.7 124.9 MediumWKC 3.71 153.8 MediumKE 4.04 143.3 MediumKW 4.83 120.4 MediumTu 4.15 142.9 Medium

Figure 7 – Spontaneous combustion liability test results in percentage,2008–2012


ConclusionThe inherent propensity for spontaneous combustion to occurat the selected South African collieries was analysed andclassified through a series of laboratory tests. It is evidentthat, despite the low frequency of underground incidents,South African collieries do have the risk of spontaneouscombustion. In all, 119 tests were conducted between 2008and 2012. The test results indicated that the majority ofSouth African collieries have medium risk ratings (56.3 percent), and the propensity of spontaneous combustion of thecollieries ranges from medium to high.

There results and the subsequent analysis highlight asignificant concern – that there are no low-range results , andthis emphasizes the importance of monitoring the early signsof spontaneous combustion in the collieries. However, thereis also the need to re-visit the definitions of low, medium,and high risk for the spontaneous combustion liability index,and this will require further research. There is already aconsiderable body of evidence that the seams of theWaterberg coalfield are particularly prone to spontaneouscombustion, and there will be a definite need to incorporatethose results into any new research.

Based on the tests results, it was found that there arehigh spontaneous combustion risks in the Witbank andHighveld coalfields. Relatively low spontaneous combustionrisk was found in the KwaZulu-Natal coalfield, whereanthracite coal is mined, as well as the northern parts ofErmelo coalfield.

References

GENC, B. and COOK, A. 2013. Determination of spontaneous combustion risk inthe South African coalfields. 23rd International Mining Congress and Fair,Antalya, Turkey, 16–19 April 2013.

GOUWS, M.J. 1987. Crossing point characteristics and differential thermalanalysis of South African coals. MSc dissertation, University of theWitwatersrand, Johannesburg.

PHILLIPS, H., CHABEDI, K., and ULUDAG, S. 2011. Best Practice Guidelines forSouth African Collieries.http://www.coaltech.co.za/Annual_Colloquium/Colloquim%202011/Spontaneous%20Combustion%20Prevention%20and%20Control%20by%20Huw%20Phillips,%20Kelello%20Chabedi%20&%20Sezer%20Uludag.pdf[Accessed 5 January 2014]. ◆

▲


IntroductionSouth Africa’s main sources of coal supply forthe last century have been the Witbank,Highveld, and Ermelo coalfields. Thesecoalfields have been extensively mined toproduce the coal required to satisfy the needsfor power generation and other industrialrequirements, and also to establish acompetitive share for South African coal in theinternational export market. The remainingreserves of coal in these coalfields willeventually become depleted – it is predictedthat this will happen by about 2040. SouthAfrica does, however, still have extensivereserves of coal in other areas, namely theWaterberg and Soutpansberg coalfields, whilecoalfields located in the Springbok Flats, theFree State, and Molteno area remain largelyunexploited. The coal from some of these areasis, however, of relatively low quality comparedto the coal from the Witbank area.

Since the coal found in the Waterberg,Soutpansberg, and other coalfields differs fromthe coal traditionally mined, new techniqueswill be required in the future to mine, process,and utilize the coal. It is expected that thequality of the coal as-mined will becomeincreasingly poorer while the coal market willbecome increasingly more demanding in termsof the quality of the product. This willtherefore present significant challenges to thecoal industry.

History of coal processing in SouthAfricaIn the early days of coal mining in SouthAfrica, coal was selectively mined to satisfythe requirements of local industries – mainlythe gold and diamond mines and the transportindustry. The requirement was for coarse coaland it was customary to screen the coal finerthan about 6 mm, which was termed ‘duff’,from the coal as- mined. The coarse coal wassupplied to the end-users while the duff, whichcomprised a significant portion of the run-of-mine coal, was discarded. Selective miningresulted in the sub-optimum utilization of coalreserves, and mine owners realized thatwhole-seam mining would be a moresustainable option. Mining the complete coalseam, however, resulted in lower quality coaland hence some form of upgrading of the coalwas needed. Hand-picking (Figure 1) was themethod first employed to achieve thisobjective, but improved coal processingtechniques eventually followed. The first coalpreparation plant in South Africa was a jigplant constructed in the Witbank area in 1909(Coulter, 1957). The next major advance incoal processing was the commissioning of aChance washer in the Vereeniging area inabout 1935 (Coulter, 1957). Other coalprocessing plants in the Witbank area and inthe former Natal province followed.

In response to a growing demand for coaland the ever-increasing pressure to supplygood-quality coal, jig washers were installed ata number of coal mines. Following theintroduction of dense medium processingusing magnetite as the medium during the1950s, the jigs were gradually replaced by themore efficient dense medium process. The

Processing low-grade coal to producehigh-grade productsby G.J. de Korte*

SynopsisSouth Africa’s best-quality coal, located in the central Highveld basin, isbecoming depleted and alternative sources of coal, such as the Waterbergcoalfield, will have to be developed to supply the country with coal in thefuture. The quality of the coal being mined in the central basin is graduallybecoming poorer. This necessitates that more of the coal be processed toimprove the quality to meet customer requirements. The challenge to thecoal processing industry is to process low-yielding coals to produce good-quality products and at the same time ensure that coal mining remainseconomically viable. This requires that more cost-effective coal processingtechnologies be investigated and implemented.

Keywordslow-grade raw coal, new developments, low-cost coal processingtechnologies, dry processing of raw coal.

* CSIR, Pretoria.© The Southern African Institute of Mining and

Metallurgy, 2015. ISSN 2225-6253. This paperwas first presented at the 21st Century challengesto the southern African coal sector, 4–5 March2014, Emperors Palace, Hotel Casino ConventionResort, Johannesburg.



Processing low-grade coal to produce high-grade products

major change in the South African coal processing industrycame about as a result of a contract concluded with aconsortium of Japanese steel mills during the early 1970s forthe supply of 2.5 Mt of low-ash coal per annum to be used asa blend coking coal. The duration of the contract was 10years, and the coal was to be produced from the Witbank No.2 seam. This contract led to the establishment of thededicated coal railway line from the Highveld to Richards Bayand the Richards Bay Coal Terminal (RBCT). Very efficientcoal processing techniques were required to process thedifficult-to-wash No. 2 Seam coal in order to produce the low-ash coal required for the Japanese steel industry. Researchconducted by the Fuel Research Institute of South Africa(FRI) during the late 1950s and 1960s was successfullyimplemented to satisfy this requirement. The initial focus ofthe FRI research was not the Japanese market, but was aimedat extracting coking coal from the Waterberg coalfield for useby the local iron and steel industry. This know-how came inhandy for the Japanese contract and was also successfullyimplemented when Grootegeluk Mine came into production inthe late 1970s. South Africa’s coal processing industrytherefore became equipped to effectively process difficult rawcoals to produce high-quality products.

Current coal processing practiceSouth Africa today has approximately 60 coal preparationplants, most of which are located in the Witbank area. Manyof these plants produce export thermal coal, which isexported via RBCT – currently a total of some 70 Mt/a. Theexport coal typically has a heat value of 6000 kcal/kg, which

requires the raw coal to be processed at a low relativedensity. Most of the mines employ two-stage processingplants, with the first stage processing the raw coal to yield anexport product and the second stage re-processing the rejectsfrom the first stage at a higher relative density to produce athermal coal for Eskom. There are also a growing number ofsmall plants that only produce coal for Eskom. Most of theexport plants, as well as the Eskom-only plants, use densemedium drums and/or cyclones for processing coarse coaland spirals to process fine coal.

There are a number of smaller plants in the Witbank area,and also a few in KwaZulu-Natal, that produce sized productsfor the inland market. These plants tend to be equipped witha single Wemco drum to process coarse coal, dense mediumcyclones to process the small coal, and spirals to process thefine coal. The drum product is usually screened to producelarge and small nuts, while the cyclone product is screenedinto peas and duff. The spiral product is usually added to theduff.

New developments in coal processingAs mentioned previously, the quality of raw coal being minedcontinues to decrease, and several mines extract coal pillarsleft from previous bord and pillar mining operations. Theproduct yields obtained from the raw coal are lower than inthe past and processing of the coal is becoming more of anecessity. Since the coal is becoming more difficult to processand product yields are low, there is increasing pressure onthe profitability of mines and as a result, low-cost processingtechniques are being evaluated and implemented. Some of the

▲


Figure 1 – Hand-picking at Springbok Colliery

technologies recently implemented in South Africa include the3-product dense medium cyclone and dry beneficiation.

The 3-product dense medium cyclone was developed inRussia during the 1970s, but only found widespreadapplication in China in the past 10 or so years. The unit isessentially a Larcodems-like cylindrical cyclone with a conven-tional conical cyclone attached to the sinks outlet. The raw coaland medium is pumped to the primary cylindrical cyclone,where a low-density separation is effected to yield an export-quality coal. The sinks and part of the medium then enters theconical cyclone, where a high-density separation is effected onthe coal to yield an Eskom product and a final reject. The unittherefore allows two separations to be carried out with a singlemedium circuit, which results in significantly lower capital andoperating costs. The 3-product cyclone in operation at UmlalaziMine is shown in Figure 2.

The capital cost of new plants is an important consid-eration and has to be kept as low as possible while stillmaintaining efficient processing of difficult raw coals. This isachieved by simplifying plant configuration through the useof large, high-capacity processing units such as large-diameter dense medium cyclones and large-capacity screens.This reduces the number of equipment items in a plant andstill enables effective separation of coal.

South Africa is a water-scarce country and the coalindustry is under pressure to reduce the amount of waterconsumed for coal processing. In this regard, a number ofcoal processing plants have installed filter presses to closetheir water circuits. By filtering the slurry produced duringcoal processing rather than disposing of it in slurry ponds,water consumption is reduced by a factor of about three. Anadded advantage is that the product obtained from the filtermay be saleable. The filter press in operation at HakhanoMine can be seen in Figure 3.

Dry processing of coal requires no water and dryprocessing techniques are therefore very attractiveconsidering our climatic conditions. Two dry processingtechnologies have been evaluated and implemented in SouthAfrica, namely the FGX dry coal separator and X-ray sorting.The FGX unit is suited to processing of -80 mm raw coalwhile the X-ray sorter is well suited to de-stoning or pre-

beneficiation of coarse coal. Further advantages of dryprocessing are that the capital and operating costs are muchlower than dense medium processing; the product coal staysdry, which effectively increases the heat value of the coal;and no slurry is produced, which lowers the environmentalimpact of coal processing. Unfortunately, the separationefficiency of the available dry processing technologies isinferior to that of dense medium separation, and thesetechnologies are not generally applicable to all raw coals. TheFGX plant at Middelkraal Colliery is shown in Figure 4.

Some of the mines that exclusively process coal forEskom need not process the complete size range of raw coaland employ partial washing. In partial washing, the finersizes of coal are dry-screened from the plant feed and reportdirectly to the product conveyor. The coarser coal is processedand the resulting product blended with the fine raw coal toconstitute the final Eskom product. The size at which the coalis dry-screened depends on the specific quality of the rawcoal and can vary between 4 mm and 40 mm. Dry screeningat small aperture sizes is not easy, but the Bivitec and LiwellFlip-Flo screens have proven capable of this duty.

Future needsIt is expected that coal processing will become more difficult infuture as the quality of raw coal mined continues to decline.Coal processing plants will have to contend with lower yieldsand more difficult-to-process coal. At the same time, strictproduct quality specifications will have to be maintained.

The separation efficiency of the processes employed willbecome even more important and it will be necessary tobalance separation efficiency against capital and operatingcosts. The low cost of dry processes make them veryattractive, especially for small mining companies, but the lowseparation efficiency of these processes may make themuneconomical in the long run. An efficient dry process istherefore required. Dry dense medium separation offers goodefficiency but is still unproven in practice. A pilot-scale drydense medium plant is in operation in China and the SouthAfrican coal industry, through the Coaltech researchprogramme, plans to evaluate this technology in the nearfuture.



Figure 3 – Filter press at Hakhano MineFigure 2 – 3-product cyclone at Umlalazi Mine


Due to low-yielding raw coal and resulting low productyields, it will be necessary for coal processing equipment tobe able to cope with high amounts of reject coal. High spigot-capacity dense medium cyclones or suitable substitutes willbe required.

Improved fine coal beneficiation and dewateringtechniques will be needed as the amount of fine coal in run-of-mine coals is expected to increase further – especially inthose mining operations where remnant pillars are being re-mined. An additional factor to contend with in the case ofpillar re-mining operations is the influence of weathering andspontaneous heating of coal. It is further anticipated thatmore fine coal will have to be utilized in power generation,and methods to improve the transport characteristics of finecoal will therefore have to be investigated.

The effective recovery and re-use of water in coalprocessing plants will become even more important. It is alsoanticipated that the cost of water, especially in the Waterbergarea, will increase significantly in future.

Improved methods for the disposal and/or use of discardsand slurry will be required to ensure that coal mines complywith ever-increasing environmental concerns.

ConclusionMining conditions in the traditional mining areas will becomemore demanding in future and mining operations from newcoalfields will have to commence. This will require coalprocessing engineers to find new and improved methods toprocess low-grade raw coals to yield high-grade productswithin ever-increasing economic and environmentalchallenges.

ReferencesQINGRU, C. and YUFEN, Y. 2002. Current status in the development of dry benefi-

ciation technology of coal with air-dense medium fluidized bed in China.XIV International Coal Preparation Congress and Exhibition, Sandton,South Africa, 11–15 March 2002.. South African Institute of Mining andMetallurgy, Johannesburg.

COULTER, T. 1957. The history and development of coal washing in South

Africa. Fuel Research Institute Symposium, Pretoria, June 1957.

FRASER, T. and YANCEY, H.F. 1926. Artificial storm of air-sand floats coal on its

upper surface, leaving refuse to sink. Coal Age, March. pp 325–327.

HALL, I. 2013. South African Coal Roadmap. Presentation to FFF Council

Meeting, SRK Offices, Johannesburg, 27 February 2013.

HONAKER, R.Q., LUTTRELL, G.H., BRATTON, R., SARACOGLU, M., THOMPSON, E., and

RICHARDSON, V. 2007a. Dry coal cleaning using the FGX separator. SA CoalPreparation Conference and Exhibition. Sandton, 10–14 September 2007.

HONAKER, R.Q., LUTTRELL, G.H., BRATTON, R., SARACOGLU, M., THOMPSON, E., and

RICHARDSON, V. 2007b. Dry coal cleaning using the FGX separator. CoalPreparation Conference and Exhibition. Lexington, KY, 30 April 30–3 May

2007.

LITH, A. 2003. Scheiding van kool en schalie met het fluïdebed. Study report,

Technical University of Delft, Netherlands.

SHUYAN ZHAO, S. and YU, J. 2012. Novel efficient and simplified coal preparation

process. International Coal Preparation 2012. Lexington, KY. 30 April

30–3 May 2012. Paper 9.

TAKO, P.R. DE JONG, VAN HOUWELINGEN, J.A., and KUILMAN, W. 2004. Automatic

sorting and control in solid fuel processing: opportunities in European

perspective. Geologica Belgica, vol. 7, no. 3-4. pp. 325–333.

YELL, A. 2007. Problems associated with dry screening of coal. Presentation to

the South African Coal Processing Society, 31 January 2007.

ZHAO, S., ZHANG, C., XU, X., YAO, W. CHEN, J., YUAN, Z., and ZHANG, H. 2010.

Super-large gravity-fed three-product heavy medium cyclone. Proceedingsof the XVI International Coal Preparation Congress, Lexington, KY, May

2010. pp 296–305. ◆

▲


Figure 4 – FGX plant at Middelkraal Colliery

IntroductionSouth Africa has large resources of coal.Prévost (2010) reported that South Africa hascoal reserves of 33 000 Mt. Annual productionis 250 Mt, over 70% of which is utilized in thedomestic market, mostly for electricity andsynthetic fuels production. Annual exportstotal 61 Mt, which generates a large foreignincome stream for South Africa. However, theexport market demands coal of a high quality.For many producers to meet this quality (andindeed, to meet quality requirements fordomestic use), often the coal must be benefi-ciated to reduce the ash content and increasethe calorific value (CV). This results in thegeneration of waste coal. This waste coal canbe categorized into three main streams –discards, duff, and slurries.

Discards are the high-ash fraction coal.These often also contain relatively high levelsof sulphur. Discards have been reported tohave a CV in the range of 11 to 15 MJ/kg(Pinheiro, Pretorius, and Boshoff, 1999; Du

Preez, 2001) The amount of discard coalcurrently stockpiled on the surface is estimatedat 1500 Mt, and the amount of discard coalgenerated in 2009 was reported to be approxi-mately 67 Mt (Prévost, 2010). This discardedcoal represents both a loss of potentiallyusable energy and an environmental threatdue to occasional spontaneous combustion ofthe heaps. Stockpiles of discard coal can alsobecome a source of acid rock drainage (ARD).

High-ash discard coal cannot be utilized inpulverized fuel (PF) boilers, but can besuccessfully utilized in circulating fluidizedbed combustion (CFBC) boilers. AlthoughCFBC technology has lagged behind PFtechnology in terms of steam conditions, withthe commissioning of supercritical CFBCboilers this ‘disadvantage’ for CFBC has largelybeen overcome (Utt and Giglio, 2011).Additionally, capital costs for the twotechnologies have converged, making CFBCcost-competitive with PF (Utt and Giglio, 2011;Aziz and Dittus, 2011; Haripersad, 2010).

The South African Department of Energyhas released a call for 2500 MW of coal-firedbase load electricity to assist in addressing thecurrent electricity supply deficit in the country(South African Department of Energy, n.d.).

It is against this background that anassessment was made of the economic merit ofgenerating electricity from discard coal in aCFBC power station.

Potential value of discard coalAs a form of screening exercise, the value ofdiscard coal, in terms of the amount ofelectricity that could be generated from it, wasassessed. This was carried out on both theexisting stockpiles of discards and the currentarisings.

Feasibility study of electricitygeneration from discard coalby B. North*, A. Engelbrecht*, and B. Oboirien*

SynopsisThere is large electricity generation potential in discard coal, both instockpiles and current arisings. Power stations with a combined capacityof up to 18 GW electrical (GWe) could be fuelled by discard coal. Moderncirculating fluidized bed combustion (CFBC) boilers, with capital costscomparable to equivalent pulverized fuel (PF) boilers, are capable ofutilizing discard coal at a high efficiency while reducing sulphur dioxide(SO2) emissions though the use of limestone sorbent for ‘in-situ’ capture.

A detailed economic assessment of the feasibility of electricitygeneration from discard coal, comprising material and energy balancesand the construction of a discounted cash flow (DCF) table, shows that itis also potentially attractive from an economic perspective. A base caseanalysis shows positive net present values (NPVs) and an internal rate ofreturn (IRR) of 21.4%. Sensitivity analyses on critical parameters showthat the economic viability is heavily dependent on parameters such ascoal cost and the value of electricity. The project becomes unattractiveabove a coal price of approximately R300 per ton and at an electricityvalue below approximately 59c per kilowatt-hour (kWh).

Site- and project-specific information such as the delivered cost ofcoal, location and efficacy of sorbents, and effective value of the electricityproduced can be used as input to the economic analysis to evaluate sitingoptions and sorbent source options for such a power station.

Keywordsfluidized bed, discard coal, electricity generation, techno-economics,sulphur capture.

* CSIR Materials Science and Manufacturing(Energy Materials), Pretoria, South Africa.

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. This paperwas first presented at IFSA 2014, IndustrialFluidization South Africa, Glenburn Lodge, Cradleof Humankind, 19–20 November 2014



Feasibility study of electricity generation from discard coal

Electricity generation from stockpiled discard coalTable I shows the amount of electricity that could begenerated from existing discard stockpiles. It was assumedthat the discards would be utilized over a 40-year period (theassumed lifetime of a power station).

This shows that a significant amount of electricity (6.5GWe of installed capacity, or approximately 16% of SouthAfrica’s current generating capacity) could potentially begenerated from existing discard coal stockpiles throughoutthe 40-year lifetime of the power station.

Electricity generation from discard coal arisingsBased on estimates of the amount of discard coal generatedon an annual basis, a similar exercise to the above wasundertaken to estimate how much electricity could potentiallybe generated from this source. The results are shown in Table II. A power station with a capacity in excess of 11 GWecould be fuelled by the discard coal arisings.

The total amount of electricity that could be generatedfrom existing stockpiles and arisings, in terms of installedcapacity and generation per year (in gigawatt-hours peryear), is shown in Table III.

It is clear that a significant amount of electricity couldpotentially be generated from existing stockpiles and currentarisings of discard coal. But, would this be an economicallyviable undertaking? An economic analysis was undertaken todetermine the economic indicators of an FBC power station.

Economic analysis of a discard coal-fired FBC powerstationA case study of a 450 MWe station was considered. This is inline with the size of FBC power stations envisaged in theSouth African Integrated Resource Plan (South African

Department of Energy, 2010a) and plants being consideredby industry (Hall, Eslait, and Den Hoed, 2011), and is withinthe proven capacity of efficient, supercritical FBC plants (Uttand Giglio, 2011).

The analysis was undertaken in two components, both ofwhich utilized Excel® spreadsheets. The first is essentially amaterial and energy balance, in which fuel and sorbentrequirements are calculated using input data such as plantsize, plant efficiency, fuel CV, calcium to sulphur (Ca/S) ratiosetc. Additionally, in this component, operating costs, fuel andsorbent transport costs, and revenue (from the sale ofelectricity) are calculated.

The figures calculated in the first component are thenused to construct the second component, a discounted cashflow (DCF) analysis. This is used to run sensitivity analysesand to calculate economic indicators such as the net presentvalue (NPV) and the internal rate of return (IRR). The IRR isthe discount rate at which a zero NPV is seen, and isessentially a measure, as its name would suggest, of thereturn that could be made on the investment. Most companieshave a ‘hurdle rate’, and will not consider projects returningan IRR that falls below this. The IRRs (and NPVs) of variousprojects are also often compared to select the optimalinvestment out of many possible investments.

Definitions of, and example calculations of, DCF, IRR, andNPV can be found in any standard economics or financebook, e.g. Correia et al. (1989).

A list of input parameters, with a discussion andreferences (if available) follows. These values are used as abase case, and different scenarios are evaluated andsensitivity analyses presented.

Assumptions and input to economic analysis

Plant size: 450 MWe

This is in line with the size of FBC power stations envisagedin the South African Integrated Resource Plan (South AfricanDepartment of Energy, 2010a) and plants being consideredby industry (Hall, Eslait, and Den Hoed, 2011), and is withinthe proven capacity of efficient, supercritical FBC plants (Uttand Giglio, 2011).

Plant efficiency: 40%

Utt and Giglio (2011) assumed 40% efficiency for asupercritical CFB. In a prior publication, Utt, Hotta, andGoidich, (2009) reported an efficiency of 41.6% for theŁagisza power station. Jantti (2011) later reported that anefficiency of 43.3% was being achieved at Łagisza; however,it appears that this may have been calculated on the lower

▲


Table I

Electricity generation potential from existingdiscard coal stockpiles

Input data Value Source

Discard coal stockpiled 1500 Mt Prevost, 2010Utilization period 40 years AssumptionAverage CV of discards 13 MJ/kg Du Preez, 2001Efficiency (coal to electricity) 40% Estimate (SC) OutputRate of use of discards 37.5 Mt/yPower plant capacity 6.5 GWe

Table II

Electricity generation potential from discard coalarisings

Input data Value Source

Discard arisings 67 Mt/a Prevost, 2010Lifetime of plant 40 years AssumptionAverage CV of discards 13 MJ/kg Du Preez, 2001Efficiency (coal to electricity) 40 % Estimate (SC) OutputPower plant capacity 11.6 GWe

Table III

Electricity generation potential from both discardcoal stockpiles and arisings

Source GWe GWh/a

Stockpiles 6.5 54 167Arisings 11.6 96 778 Total 18.1 150 944

heating value (LHV) (or net calorific value, NCV) rather thanthe higher heating value (HHV) (or gross calorific value,GCV). It was decided therefore, keeping in mind the low-quality discard coal that would be utilized, to assume therelatively conservative figure of 40%.

Capacity factor: 85%This is the electricity that is actually produced in a year as apercentage of the electricity that could be produced. It takesinto account load-following and planned and unscheduledmaintenance. The US Electrical Power Research Institute(EPRI) assumed 85% in a study undertaken as input to theSouth African IRP, and this was adopted in the currentanalysis (South African Department of Energy, 2010b).

Fuel calorific value: 13 MJ/kgThis is drawn from studies undertaken to assess theinventory of duff and discard coal (Pinheiro, Pretorius, andBoshoff, 1999; Du Preez, 2001).

Fuel ash content: 45%Discard coal in both dumps and current arisings has a widerange of ash contents (Pinheiro, Pretorius, and Boshoff,1999; Du Preez, 2001). A figure of 45% was used, This is ingood agreement with figures quoted by Hall, Eslait, and DenHoed (2011), and is similar to the ash content of theGreenside discards tested by the CSIR in the NationalFluidized Bed Combustion Boiler (NFBC) project(Eleftheriades and North, 1987). (Note, this was a bubblingFBC.) The economic calculations are not, however, verysensitive to the coal ash content, as for the purposes of thisanalysis the coal requirements are calculated from thecalorific value of the coal rather than the ash content.

Sulphur content: 2.77%Again, there is a wide range of sulphur contents in botharising discards and in dumps. A value of 2.77% was used,this being the sulphur content of the Greenside discardstested in the NFBC (Eleftheriades and North, 1987). Thisfigure is also in agreement with sulphur contents reported byHall, Eslait, and Den Hoed, (2011). Aziz and Dittus (2011)reported a significantly lower sulphur content of 1.5% in theirstudy of a CFB power station utilizing discard coal from theDelmas coal mine. The economic study is sensitive to thesulphur content of the coal because this dictates the amountof sorbent required to reduce the sulphur oxide emissions.

Not considered here, but of merit to consider in a realapplication, is the possibility of beneficiating the discards,particularly those recovered from dumps, to reduce thesulphur content and therefore sorbent requirements(discussed below). Hall, Eslait, and Den Hoed, (2011)considered this option, whereas Aziz and Dittus (2011) didnot.

Required Ca/S ratio: 2.9, 5.3This is the molar ratio of calcium in the sorbent to sulphur inthe coal, with a stoichiometric (1:1) ratio theoretically (butnot in practice) being able to remove all the sulphur. Asshown by the research in the NFBC, the calcium content of asorbent is not necessarily a good indication of the efficacy ofthe sorbent, and therefore of the amount required(Eleftheriades and North, 1987). The physical nature of the

sorbent plays a large role. A figure of 2.9 was derived fromdata quoted by Aziz and Dittus (2011) for limestone. Utt,Hotta, and Goidich (2009) indicate that 94% of the sulphurcould be removed from a fuel containing 0.6% to 1.4%sulphur at a Ca/S ratio of 2.0 to 2.4. It was decided to use thefigure of 2.9 quoted by Aziz and Dittus (2011), as a conser-vative approach.

For dolomite (the rationale for use of which is explainedbelow), a Ca/S ratio of 5.3 was used. This is based on therelative performance of Lyttelton dolomite versus Union limeshown in the research on the NFBC. This is an estimate, butit is intended to show the effect of sorbent type and source onfinancial viability.

The Ca/S ratio is an important parameter, as it dictatesthe amount of sorbent that will be required, which is asignificant operating cost for the plant. It would be of greatvalue if the economic assessment developed here could belinked to a sorbent efficacy model, so that the required Ca/Sratio for a given sorbent can be input, rather than estimated.

Calcium carbonate content of sorbent: 30–96%While the selected Ca/S ratio drives the calculation of howmuch calcium is required, the calcium content of the sorbentthen dictates how much sorbent is required. This hasimplications for both the base cost of the limestone and thetransport cost. South African limestones typically have acalcium carbonate content in the range of 85% to 95%(Agnello, 2005). The limestone chosen for this analysis issupplied by Idwala Lime from the limestone quarry inDanielskuil, approximately 700 km from the Witbank area.Idwala Lime currently supplies the limestone for the CSIR-designed FBC high-sulphur pitch incinerator operating atSasol in Sasolburg (North et al., 1999). This limestone has ahigh calcium carbonate content, at 96%. This equates to acalcium content of 38.4%, as the molecular weight of calciumcarbonate is 100, whereas that of calcium is 40.

An advantage of in-situ sulphur capture in FBC over fluegas desulphurization (FGD) in pulverized fuel (PF)-firedboilers is that FBC can utilize relatively poor sorbents,including dolomite. Haripersad (2010), drawing heavily onAgnello (2005), concluded that the ability of FBC to utilizethese lower grade sorbents was a driver towards the adoptionof FBC technology. There would be competition with the goldmining industry and the cement industry for the high-gradelimestone required for FGD on PF plants, whereas there islittle competition for low-grade limestone and dolomite.Further, he concluded that PF with FGD would becomeresource-constrained in terms of both sorbent and water by2025. A scenario of using dolomite was therefore alsoconsidered in this current assessment.

Fixed operational costs: R202 million per yearThis was calculated from the figures quoted by EPRI for fixedcosts of an FBC power station (with limestone addition) as afactor of the installed capacity (South African Department ofEnergy, 2010b). (R404 per kW per year, escalated by theconsumer price index).

Variable operational costs: R258 million per yearThis was calculated from the figure quoted by EPRI (SouthAfrican Department of Energy, 2010b) for variable operatingcosts for an FBC power station as a factor of power sent out




in the year (R69.1 per MWh) (escalated by CPI). The costs forFBC without limestone addition were used, as in this currentanalysis the limestone costs are split out in order to assesstheir contribution to the costs, and to enable sensitivityanalyses to be carried out on the delivered cost of limestone.

Water cost: R434 000 per yearThis was derived from the water consumption indicated byEPRI (South African Department of Energy, 2010b) (33.3l/MWh) and an assumed cost of water of R3.50 per megalitre(escalated by CPI). Water costs appear to be a relatively smallcomponent of the total annual operating costs.

Fuel cost: R129 per tonThis value was essentially ‘reverse engineered’ from thecurrent electricity price and the indication by Koornneef,Junginger, and Faaij (2006) that the fuel component of thecost of electricity for ’waste coal’ is 15%. Again, in reality,there could be a great range to this value. From experience,when a waste begins to be used the owner of that wastestarts to ascribe increasing value to it. If the power stationdeveloper is also the owner of the mine, this effect will largelybe negated.

Utt and Giglio (2011) used a value of $100 per ton for a25 MJ/kg coal, and EPRI (South African Department ofEnergy, 2010b) used approximately R288 per ton for a 19.2MJ/kg coal. The cost of the fuel needs to be determined/negotiated and contracted in order to conduct an accurateeconomic viability assessment. For the purposes of thisstudy, where the specific intent is to show the potentialadvantage of using waste coal, we believe the approach ofusing the fuel cost component indicated by Koornneef,Junginger, and Faaij (2006) is valid. A wide range of fuelcosts is considered in the sensitivity analyses.

Fuel transport cost: R0.93 per kilometer per tonIt proved difficult to obtain transport costs from the transportindustry itself. An indication of road transport costs wasobtained from Blenkinsop (2012). Although not in thetransport industry, Blenkinsop is assessing the viability ofutility-scale FBC projects in southern Africa, and is thereforeregarded as a reliable source of information. He indicated arange of between R0.90 and R1.30 per kilometre per ton(including escalation by CPI). The lower limit was taken, thisbeing the transport cost indicated by Idwala Lime (below).

Fuel transport distance: zeroAs the intent is to operate a mine-mouth power station, thiswill be zero for this current assessment. It has, however,been included in the calculations in order that sensitivity tothis figure can be assessed should a potential application belocated away from the mine. Alternatively, there could bemultiple fuel feeds from multiple mines.

Sorbent cost: R449 per tonThis cost was obtained from Idwala Lime. The price has beenescalated by CPI.

Sorbent transport cost: R0.93 per ton per kilometreIdwala Lime indicated that the transport cost of their productfrom Danielskuil to Witbank is R650 per ton (after escalationby CPI). With the distance being approximately 700 km, thisequates to approximately 93c per kilometre per ton.

Sorbent transport distance: 700 kmA distance of 700 km was used for the analysis, this beingthe distance from the Idwala Lime mine in Danielskuil toWitbank. The sorbent transport distance is, however, variedin order to gauge the sensitivity of the project viability to thisparameter.

Electricity value: R0.5982 per kWhThe tariff at which Eskom is allowed to sell electricity iscurrently a hotly debated subject in South Africa. Proposedtariffs are set out in a Multi-Year Pricing Determination(MYPD) document. The National Energy Regulator of SouthAfrica (NERSA) reviews this, and makes a decision on whatit believes is a reasonable tariff increase, based on consider-ations of the cost of producing electricity and the impact thatincrease power tariffs could have on the economy of SouthAfrica.

For each of the years 2010/2011, 2011/2012, and2012/2013, an increase of 25.9% was approved by NERSA(Eskom, 2012). However, following a ‘... combined effort byGovernment and Eskom to lessen the impact of higher tariffincreases on consumers ...’, the increase for 2012/2013 wasreduced to 16% (Eskom, 2012). The revenue reported in2013/2014 was 62.82 cents per kWh, which includes a 3 centper kWh environmental levy (Eskom, 2014). A figure of59.82 cents per kWh was therefore used. It is not clear,however, how much of this value could be realized by anindependent power producer (IPP). If the electricity is to beused elsewhere (but possibly within the same company orgroup), there will be costs associated with transporting theelectricity through the Eskom grid. An analysis was thereforerun to estimate the lower limit for the electricity value thatstill results in a viable project. The value of the product,electricity, does of course have a major impact on viability. Inthe case of generation of electricity for self-use, the electricitywill not result in true revenue, but will be an avoided cost.

An alternative approach was also taken, i.e. to calculatethe cost that electricity would need to be sold at in order torealize an acceptable IRR (the hurdle rate of 20%).

Plant capital cost: R20 490 per kWeThis is a very important parameter, and unfortunatelyestimates of this varied. Utt and Giglio (2011) indicate aspecific plant cost of $2000 to $2100 per kWe installedcapacity for supercritical FBC. Tidball et al. (2010) showed arange of between approximately $1700 and $2600 per kWe(reported in 2007). This was a subcontract report written fora National Renewable Energy Laboratory (NREL) contract.EPRI (South African Department of Energy, 2010b) indicate aspecific plant cost of R16 540 per kWe. This is quoted inSouth African rands rather than US dollars because theanalysis was conducted as input to the South AfricanIntegrated Resource Plan. It was decided to use this value(corrected for four years of inflation at the average SouthAfrican inflation rate of 5.5%, giving R20 490 per kWe)because this (a) was specifically carried out for a SouthAfrican scenario and (b) specifically considered FBC powerstations.

Depreciation period: 5 yearsThis is included in the discounted cash flow as a ‘wear-and-tear’ tax allowance that is allowed on capital expenditure. The

▲


allowable depreciation was assumed to be straight-line over 5years. This approach is explained in Correia et al. (1989).

Plant lifetime: 30 yearsAlthough a power station may be kept operating for 40 yearsor more, the assumption made by EPRI (South AfricanDepartment of Energy, 2010b) of 30 years’ plant life was alsoused in this analysis.

Discount rate: 9%This is a key parameter in an economic analysis.Unfortunately, again, there is a range of values suggested.The discount rate is essentially a return that an investorwould have to receive on the investment to warrant it.Generally, the value used here is the weighted average cost ofcapital (WACC) (or weighted marginal cost of capital, WMCC).This is (simplistically) calculated from the relative weightsand contributions of equity, debt, and shares that are used tofinance the project (Correia et al. , 1989). The accuratecalculation of the WACC is in itself a science, and can involvethe application of a capital asset pricing model (CAPM) (Nell,2011). Power (2004) asserts that ‘The cost of capital is aprice, a price for a “share” of risk sold by a company.’ Assuch, factors such as where a company’s head office is listedcan significantly affect it.

For the purposes of this analysis it was decided to useavailable figures for the WACC for the only current electricityutility in South Africa, Eskom. However, even with thisnarrowed focus, a range was obtained. BUSA states that theWACC proposed by Eskom (10.3%) was possibly high, and avalue of 8% may be more realistic (BUSA, 2009). Mokoenastates that Eskom’s WACC is 8.16% (Mokoena, 2010).Mining Weekly quoted Dick Kruger, SA Chamber of Minestechno-economic assistant adviser, as saying that ‘... the10.3% applied by the utility ... should be as much as threepercentage points lower ...’ (Mining Weekly, 2012). It wasdecided to adopt a figure towards the middle of this range,namely 9%.

Tax rate: 28%This is the standard tax levied on companies by the SouthAfrican Revenue Service (2012).

Inflation: 5.5%Inflation is a variable figure. Historically South Africa hasseen periods of high inflation, whereas more recentlyinflation has been lower and more stable. Bruggemans(2011) shows a current inflation rate (2012) of 5.6%, andforecasts 5.5% and 5.9% respectively for 2013 and 2014.

The figure of 5.5% forecast for 2013 was assumed forthis study. It was further assumed that this would holdsteady over the analysis period. An inflation rate for eachfuture year could be incorporated into the DCF, but thiswould complicate the analysis, with uncertain added value. Inany event, the more important consideration is how muchmore or less than the CPI inflation rate other parameters willbe, such as fuel price, transport price etc.

Coal, water, and transport cost inflation: 7.5% (2% aboveCPI)An assumption was made that energy-related costs wouldrise at a rate above inflation. Coal is an energy product, waterhas a high electricity component to its price (due to pumping

requirements), and transport obviously requires fuel and/orelectricity.

Limestone, fixed operational and variable operationalcosts: 5.5% (equal to CPI)These commodity or equipment-type costs are assumed toinflate in line with the CPI.

Electricity price inflation (5-year 16%, CPI + 5%)The general belief that electricity price increases wouldcontinue to be well above inflation has proven to be valid,with the release of Eskom’s Multi-year Price Determination 3(MYPD3) document. Engineering News reports that increasesof 16% have been requested in MYPD3, which was releasedon 22 October 2012 (Engineering News, 2012).

As with previous MYPD submissions this will still need tobe reviewed by NERSA, but for the purposes of this analysisan increase of 16% per year was assumed for the first 5years, with increases of CPI plus 5% thereafter.

Discussion on material and energy balance and DCFThe material and energy balance, including calculation ofcosts, of the base case is shown in Table IV. In order to testthe material and energy balance, input data was derived fromthe information presented by Aziz and Dittus (2011) and thesame output in terms of fuel and sorbent requirements etc.was obtained. It was therefore concluded that the materialand energy balances were sound.

The DCF table produced from this data (plus additionalinput such as inflation estimates) is presented in Table V.

From these ‘input parameters’, a DCF table wasconstructed. The cash flow was calculated per year for 30years. A summarized form of the DCF for the base case isgiven in Table V.

A summary of the financial indicators (NPV at 10, 20,and 30 years, and the IRR) is given in Table VI.

With an IRR of 21.4%, this appears to be a potentiallyworthwhile investment opportunity, warranting furtherinvestigation (and refinement of figures). As discussedabove, investors would adopt a hurdle rate of about 20%.

Minimum value of electricity for financial viability (toachieve 20% IRR)The DCF was used to calculate the value of electricity (incents per kWh) that would deliver the adopted hurdle rate of20% (with all other parameters as per the base case). Thiswas calculated at 55.42 cents per kWh.



Table IV

Fuel and sorbent requirements and costs, andrevenue

Parameter Value

Fuel required 2.3 Mt/aSorbent required 0.6 Mt/aFuel cost 300.0 Rm/aFuel transport 0.0 Rm/aSorbent cost 272.0 Rm/aSorbent transport 393.0 Rm/aTotal sorbent cost 665.0 Rm/aElectricity value 59.5 c/kWhElectricity revenue 2000.0 Rm/a


This price is very sensitive to the chosen hurdle rate. Forexample, should an investor adopt a hurdle rate of 22%, anelectricity price of 61.54 cents per kWh would be required. Aneven more conservative investor, adopting a hurdle rate of24%, would require 67.88 cents per kWh.

Sensitivity analysisFinancial indicators were calculated using the DCF. Thesewere calculated for the base case and also used to runsensitivity analyses on the following parameters:

➤ Plant capital cost➤ Cost of coal➤ Transport distance of sorbent➤ Electricity price (at project start).

Plant capital costTo assess the sensitivity of the project to plant capital cost,this was varied from $1600 to $2800 per kWe installedcapacity. The results are shown in Figure 1.

The IRR is sensitive to the specific plant capital cost, andfalls from 26.1% to 19.74% as the specific plant cost risesfrom $1600 to $2800 per kWe. At a hurdle rate of 20%, theproject would be considered marginal at a capital cost inexcess of $2600 per kWe.

Cost of coalThe cost of coal was calculated using information fromKoornneef, Junginger, and Faaij (2006) indicating that the

fuel cost component of the cost of electricity for ’waste coal’ is15%. However, estimates varied, as indicated previously,with Utt and Giglio taking $100 per ton as a value (Utt andGiglio, 2011). In this current analysis, the coal will bepurchased in South African rands. The cost of the coal wasvaried from zero to R900 per ton. Figure 2 shows the trend ofNPV and IRR with coal price.

▲


Table V

Discounted cash flow for the base case (all values in millions of rands)

Year 0 1 2 3 4 5 10 15 20 25 30

Costs

Capital9221

Coal 300 323 347 373 401 575 826 1186 1702 2444Coal Xport 0 0 0 0 0 0 0 0 0 0Limestone 272 287 303 320 337 441 576 753 984 1287LS Xport 393 422 454 488 524 753 1080 1551 2227 3197Water 0 0 1 1 1 1 1 2 2 4Fixed opex 202 213 225 238 251 328 428 560 731 956Var. opex 258 272 287 303 319 417 545 713 931 1217Total costs 1425 1518 1616 1721 1833 2515 3457 4764 6579 9104RevenueElectricity 2000 2320 2692 3122 3622 5967 9830 16195 26680 43954Pre-tax profit 575 803 1076 1401 1789 3452 6373 11431 20101 34850Tax 161 225 301 392 501 967 1784 3201 5628 9758Post-tax profit –9221 414 578 774 1009 1288 2486 4588 8230 14473 25092Depreciation 516 516 516 516 516DCF –9221 853 921 997 1081 1173 1050 1260 1469 1678 1891NPV –9221 –8367 –7446 –6449 –5369 –4196 630 6510 13435 21406 30435

Table VI

Financial indicators for the base case

Indicator Value Units

NPV (10 years) 630.0 RmNPV (20 years) 13 435.0 RmNPV (30 years) 30 435.0 RmIRR (30 years) 21.4 %

Figure 2 – Effect of coal cost

Figure 1 – Effect of specific plant capital cost

The cost of coal has a large effect on the viability of theproject. From zero cost up to R200 per ton, the project stillshows an IRR above the hurdle rate of 20%. At R300 per ton,the IRR is 19.63%, marginally below the hurdle rate. The 10-year NPV also becomes negative. At R700 per ton, the 20-year NPV also becomes negative. The indication is that thisproject (with the assumptions on other costs and revenue)will not be viable at a coal price in excess of approximatelyR300 per ton.

Transport distance of sorbentIn order to evaluate the sourcing of sorbent, the effect oftransport distance (and therefore cost) was assessed. For thebase case, a distance of 700 km was taken. For thissensitivity analysis a range of zero to 1000 km was used.Figure 3 shows the trend of NPV and IRR with sorbenttransport distance. This analysis could also be used to assessthe options of sourcing a low-grade sorbent near to the powerstation or a high-grade sorbent further away. For this to be ofvalue, however, a full understanding of the efficacy of thesorbents would be needed.

Although the IRR at a transport distance of 1000 km, at20.3%, is still above the hurdle rate, an investor shouldinvestigate sorbent sourcing options. The limestone anddolomite deposits in South Africa are well known, but theefficacy of these sorbents in CFBCs has not been fullydetermined.

Effect of electricity price (at start of project)An electricity value of 59.82 cents per kWh was used for thebase case analysis as described above. There is, however,significant doubt as to the accuracy of that figure, as itdepends on factors such as charges to ‘wheel’ the electricitythrough the existing grid, which would lower the effectiverevenue earned. There are also indications that it could behigher. Tore Horvei (2012), who was involved in feasibilitystudies of this kind in southern Africa, indicated that thevalue of electricity could be 85 cents per kWh. In order togauge the sensitivity of the project to the electricity price itwas varied from 30 cents to 90 cents per kWh. Figure 4shows the trends of NPV and IRR with electricity price.

The electricity price has a marked effect on the viability ofthe project. At 30 cents per kWh to 50 cents per kWh theproject shows a negative NPV after 10 years. The IRR hurdle

rate of 20% is achieved only at approximately 59 cents perkWh. At the higher electricity prices, a high IRR is seen, inexcess of 30%. The conclusion that can be drawn from this isthat a potential IPP needs to understand clearly how muchrevenue will be effectively gained through the sale ofelectricity, as project viability is very sensitive to thisparameter.

Conclusions and recommendationsThere is significant electricity generation potential in discardcoal. A combined total of approximately 18 GWe installedcapacity could be fed with discard coal stockpiles andarisings.

CFBC technology has developed to the point where it is ona par with PF technology in terms of both efficiency and cost,and the ability of CFBC to utilize discard coal has beenproven.

An economic analysis indicates that generating electricityfrom discard coal via CFBC is potentially favourable. The basecase shows an IRR of 21.4%, which is above the hurdle rateadopted in this study of 20%. However, there are manyfactors to consider that affect the return on investment. Themajor elements affecting the IRR are the cost of the coal andthe value of the electricity. For a given project, the analysis(in particular ash content, sulphur content, and CV) andamount of discard coal and the logistics around getting it tothe power station must be fully understood, so that theeffective cost of the ‘free’ fuel is known. The true value of theelectricity, or the avoided cost if the electricity is generatedfor self-use, must be ascertained.

If possible, updated figures on the size and analysis ofboth discard stockpiles and arisings should be generated.This is because the dumps are being reprocessed, andmodern coal beneficiation technologies are resulting inreduced carbon content of the arisings.

Unbeneficiated run-of-mine coal could also be consideredas a feed to a CFBC power station.

The cost and efficacy of sorbent also affects the viabilityof the project. South African sorbent resources are wellknown, but sorbent efficacy in CFBCs is not. An efficacydatabase, perhaps linked to a GIS database, would enable anaccurate determination of the cost of sorbent to be made.



Figure 3 – Effect of sorbent transport distance Figure 4 – Effect of electricity price (at project start)


ReferencesAGNELLO, V. 2005. Dolomite and Limestone in South Africa: Supply and

Demand 2005. Report no. R49/2005. Department of Minerals and Energy,South Africa.

AZIZ, T. and DITTUS, M. 2011. Kuyasa mine-mouth coal-fired power project:Evaluation of circulating fluidized-bed technology. Proceedings ofIndustrial Fluidization South Africa, 2011. pp. 11–29.

BLENKINSOP, M. 2012. 11 October 2012. Personal communication.

BRUGGEMANS, C. 2011. First National Bank Five Year Economic Forward Look. https://www.fnb.co.za/economics/econhtml/forecast/fc_5yearview_new.htm.[Accessed 17 October 2012].

BUSINESS UNITY SOUTH AFRICA. 2009. Preliminary response to the Eskom RevenueApplication for the Multi Year Price Determination for the period 2010/11to 2012.13 (MYPD 2).http://www.busa.org.za/docs/PRELIMINARY%20SUBMISSION%20ESKOM%20APP LICATIONfinal.pdf [Accessed 17 October 2012].

CORREIA, C., FLYNN, D., ULIANA, E., and WORMALD, M. 1989. FinancialManagement. 2nd edn. Juta, Johannesburg, South Africa.

DU PREEZ, I. 2001. National Inventory of Discard and Duff Coal. Badger Mining.Confidential report prepared for the SA Department of Minerals andEnergy.

ELEFTHERIADES, C.M. and NORTH, B.C. 1987. Special plant features and theireffect on combustion of waste coals in a fluidized bed combustor.Proceedings of the 9th International Conference on Fluidized BedCombustion, Boston, 3–7 May 1987, Mustonen, J.P. (ed.). ASME NewYork. pp. 353–359.

ENGINEERING NEWS. 2012. Eskom seeks yearly increases of 16% to 2018.http://m.engineeringnews.co.za/article/eskom-seeks-yearly-increases-of-16-to-2018-2012-10-22 [Accessed 23 October 2012].

ESKOM. 2012. Tariffs and Charges Booklet 2012/2013.http://www.eskom.co.za/content/ESKOM%20TC%20BOOKLET%202012-13%20(FINAL)~2.pdf [Accessed 16 October 2012].

ESKOM. 2014. Tariffs and Charges Booklet 2014/2015.

HALL, I., ESLAIT, J., and DEN HOED, P. 2011. Khanyisa IPP – a 450 MWe FBCproject: Practical challenges. Proceedings of Industrial Fluidization SouthAfrica 2011. pp. 47–55.

HARIPERSAD, N. 2010. Clean Coal Technologies for Eskom. MSc thesis. Da VinciInstitute of Technology Management, Johannesburg.

HORVEI, T. 2012. 22 October 2012. Personal communication.

JANTTI, T, 2011. Lagisza 450 MWe supercritical CFB – operating experienceduring first two years after start of commercial operation. Proceedings ofCoal-Gen Europe 2011, Prague, Czech Republic, 15–17 February 2011.

KOORNNEEF, J., JUNGINGER, M., and FAAIJ, A. 2006. Development of fluidized bedcombustion – An overview of trends, performance and cost. Progress inEnergy and Combustion Science, vol. 22, no. 1. pp. 19–55.

MINING WEEKLY. 2012. Big electricity hikes will be “materially damaging” to SAmines. http://www.miningweekly.com/article/big-electricity-hikes- will-be-materially-damaging-to-sa-mines-2010-01-22 [Accessed 10 October2012].

MOKOENA, S. 2010. Guideline on municipal electricity price increase for 2011/12.http://www.busa.org.za/docs/PRELIMINARY%20SUBMISSION%20ESKOM%20APP LICATIONfinal.pdf [Accessed 16 October 2012].

NEL, S. 2011. The application of the capital asset pricing model (CAPM): aSouth African Perspective. African Journal of Business Management, vol.5, no. 13. pp. 5336–5347.

NORTH, B.C., ELEFTHERIADES, C.E., ENGELBRECHT, A.D., and RUTHERFORD-JONES, J.1999. Destruction of a high sulphur pitch in an industrial scale fluidizedbed combustor. Proceedings of 15th International Conference on FluidizedBed Combustion, Savannah, Georgia, 16–19 May 1999.

PINHEIRO, H.J., PRETORIUS, C.C., and BOSHOFF, H.P. 1999. Analysis of discard coalsamples of producing South African collieries. Confidential unpublishedreport for the South African Department of Minerals and Energy.

PRÉVOST, X.M. 2010. Personal communication. 14 October.POWER, M. 2004. How has South Africa Inc sought to reduce its high cost of

capital? OECD Development Centre Seminar: ‘Cheaper Money for SouthernAfrica – Unlocking Growth’. Paris, 7 October 2004.

SOUTH AFRICAN DEPARTMENT OF ENERGY. Not dated. Coal Baseload call.https://www.ipp-coal.co.za/Home/About [Accessed 7 May 2015].

SOUTH AFRICAN DEPARTMENT OF ENERGY. 2010a. Integrated Resource Plan.http://www.energy.gov.za/IRP/2010/IRP2010.pdf [Accessed 12 October2011].

SOUTH AFRICAN DEPARTMENT OF ENERGY. 2010b. Power Generation TechnologyData for Integrated Resource Plan of South Africa.http://www.energy.gov.za/ – Programmes and Projects - IntegratedResource Plan – EPRI report on supply side cost) [Accessed 12 October2012].

SOUTH AFRICAN REVENUE SERVICE. 2012. SARS pocket tax guide, budget 2012.http://www.treasury.gov.za/documents/national%20budget/2012/sars/Budget%202 012%20Pocket%20Guide.pdf [Accessed 17 October 2012].

TIDBALL, R., BLUESTEIN, J., RODRIGUEZ, N., and KNOKE, S. 2010. Cost andperformance assumptions for modelling electricity generatingtechnologies. NREL subcontract report NREL/SR-6A20-48595.http://www.nrel.gov/docs/fy11osti/48595.pdf [Accessed 11 October2012].

UTT, J. and GIGLIO, R. 2011. Technology comparison of CFB versus pulverized-fuel firing for utility power generation. Proceedings of IFSA 2011:Industrial Fluidization South Africa, Johannesburg, 16–17 November2011. pp. 91–99.

UTT, J., HOTTA, A., and GOIDICH, S. 2009. Utility CFB goes “supercritical” – FosterWheeler’s Lagisza 460 MWe operating experience and 600-800 MWedesigns. Proceedings of Coal-Gen 2009, Charlotte, North Carolina, 18–21August 2009. ◆

▲


Advanced CFB boilers for utility powergenerationWhen the Łagisza power plant (Figure 1),located in the Katowice area of southernPoland, began commercial operation in June2009, it marked a new era in the evolution ofcirculating fluidized-bed (CFB) technology.The plant is now celebrating its sixth year ofsuccessful commercial operation.

Besides being the most advanced operatingCFB steam generator in the world, the CFB atthe Łagisza plant has unique first-of-a-kinddesign features, such as vertical-tubesupercritical steam technology and low-temperature flue-gas heat recovery system thatallows the plant to achieve a very high netplant efficiency of 43.3% (based on the fuel’s

lower heating value). A notable feature of theŁagisza CFB is that it meets all atmosphericemission permit levels without post-combustion de-NOx or de-SOx equipment suchas selective catalytic reduction (SCR) or flue-gas desulphurization (FGD).

Like the Łagisza plant owners (PKE),Korean Southern Power Company (KOSPO)also saw value in CFB technology when itchose the technology for its 2200 MWe GreenPower Plant project in Samcheok, Korea(Figure 2). The Samcheok plant, which is nowunder construction, will utilize four larger550 MWe CFB boilers featuring ultra-supercritical steam conditions (257 barg,603/603°C). These CFB boilers will be themost advanced units in the world when theplant comes on line as expected in 2016.

Both PKE and KOSPO first consideredconventional pulverized coal (PC) technologyfor their projects, but after studying theadditional technical and economic benefits thata CFB brings, they ultimately chose CFBtechnology. The CFB boilers offer manybenefits, but two in particular played a big rolein their decision. They were:

➤ The CFB’s ability to reliably burn bothlow-rank and high-quality coals besidesbiomass and waste coal slurries (Łagiszaonly) dramatically improved thepotential for huge fuel cost savings andhigh fuel procurement security

➤ The CFB’s ability to meet atmosphericemission goals without FGD or SCRtechnology saved on capital, operatingcosts and water.

The value proposition of circulatingfluidized-bed technology for the utilitypower sectorby R. Giglio* and N.J. Castilla*

SynopsisCirculating fluidized-bed (CFB) combustion technology has been aroundfor over 40 years, but over the last 6 years it has been commerciallydemonstrated at the 500 MWe scale at the Łagisza plant located in Będzin,Poland. The CFB at the Łagisza plant has unique first-of-a-kind designfeatures, such as vertical-tube supercritical steam technology and a low-temperature flue-gas heat extraction that allows the plant to achieve highplant efficiencies of over 43% (net lower heating value). Another unusualfeature for a coal power plant is that this plant meets all its atmosphericemission permit levels without any post-combustion de-NOx or de-SOxequipment such as selective catalytic reduction (SCR) or flue-gasdesulphurization (FGD).

CFB clean coal power technology is entering the utility power sectorjust in time to help deal with declining quality in internationally tradedcoals and to promote the large-scale use of economic, low-quality domesticfuels. Owing to the very attractive price discounts, growing supplies oflow-quality Indonesian coals are outpacing the supply of high-qualityAustralian, Russian, and US coals. In Germany and Turkey, the use ofdomestic lignites for power production provides a secure and economicenergy solution while creating domestic jobs.

Conventional pulverized coal (PC) boilers will have trouble acceptingthese off-specification coals because of their narrow fuel specifications;they typically call for heating values above 5500 kcal/kg. This limitation isnot an issue for CFB technology because of its ability to burn the worstand best of coals with heating values ranging from 3900 to 8000 kcal/kg.

This paper provides an outlook for future coal supply, quality, andprice, as well as a review of the technical and economic benefits of CFBtechnology firing low-quality fuels for utility power generation.

Keywordsvalue proposition, CFB, flexibility, lignite, Łagisza, circulating fluidized-bed, power generation.

* Amec Foster Wheeler Global Power Group,Hampton, NJA, USA

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. This paperwas first presented at IFSA 2014, IndustrialFluidization South Africa, Glenburn Lodge, Cradleof Humankind, 19–20 November 2014



The value proposition of circulating fluidized-bed technology for the utility power sector

CFB’s benefits are rooted in its unique combustionprocessThe CFB’s advantages of reliability, low maintenance, a widefuel range, and smaller and less costly boilers are rooted inits unique flameless, low-temperature combustion process.As Figure 3 summarizes, unlike conventional PC or oil/gasboilers, the fuel’s ash does not melt or soften in a CFB, whichallows the CFB to avoid many of the fouling and corrosionproblems encountered in conventional boilers with an openflame.

Supercritical boiler design considerationsFor once-through supercritical boiler designs (Figure 4), thelow, even combustion temperature and heat flux throughoutthe CFB’s furnace minimizes the risk of uneven tube-to-tubetemperature variations, which permit the furnace walls to beconstructed with cost-effective and easy-to-maintain smoothvertical tubes. For additional protection, Amec FosterWheeler’s once-through CFB boilers utilize a patented low-steam mass flux design providing a natural self-coolingcharacteristic that uses buoyancy forces to increase thewater/steam flow in a tube proportionate to the amount ofheat it receives. This further minimizes tube-to-tubetemperature variations and ensures low mechanical stressesacross the furnace, thereby extending furnace life.

To cope with the uneven temperatures and heatabsorption in the furnace, most conventional PC and oil orgas once-through boilers incline and wrap the furnace walltubes around the lower section of the furnace to even outtube-to-tube heat absorption and temperatures. Althoughthis solves the heat imbalance problem, the spiral design hasseveral disadvantages compared with Amec Foster Wheeler’sCFB vertical-tube design. The spiral design requires aheavier, more complicated boiler and boiler support system

▲


Figure 3 – Comparison of conventional versus CFB boiler technology

Figure 1 – Łagisza CFB power plant located in Będzin, Poland

Figure 2 – 2200 MWe Green Power CFB plant located in Samcheok,Korea

which makes furnace tube repairs more difficult.Furthermore, the loss in steam pressure is high because thesteam path is long and the ledge formed at the interfacebetween the spiral and the vertical tube header is a naturallocation for build-up of slag.

Furnace size versus fuel qualityAs ash does not soften or melt in a CFB, the size of thefurnace does not increase as much as conventional boilerswhen firing lower quality fuels. As can be seen in Figure 5, inorder to control fouling, slagging, and corrosion, the furnaceheight of a PC boiler doubles and its footprint increases byover 60% when firing a low-quality fuel such as high-sodiumlignite, whereas the CFB boiler height increases by only 8%and its footprint increases by only 20%. This results in a CFBboiler that is smaller and costs less that a PC boiler.

Furthermore, unlike a PC boiler, a CFB boiler does notneed soot blowers to control the build-up of deposits and slagin the furnace as the ash does not soften and the circulatingsolids themselves remove deposits and minimize their build-up on the furnace wall, panels and coils.

Superheater and reheater design considerationsAnother very important feature of a CFB boiler involves thefinal superheat and reheat steam coils. These coils operate atthe highest metal temperatures in the boiler, which makesthem vulnerable to corrosion and fouling. This vulnerabilityincreases significantly for supercritical boilers with highsteam temperatures.

As shown in Figure 6, in a conventional PC or oil/gasboiler these coils are suspended from the furnace ceiling andare directly exposed to the slagging ash and corrosive gases



Figure 4 – Comparison of spiral versus vertical-tube once-through furnace design

Figure 5 – Impact on furnace size as fuel quality degrades: PC versus CFB

(sodium and potassium chlorides) in the hot furnace flue-gas. To cope with this undesirable situation, boiler designersuse expensive alloys and recommend a high level of cleaningand maintenance for these coils.

This design weakness is avoided in Amec FosterWheeler’s CFB boilers by submerging these coils in hot solidsfluidized by clean air in heat exchangers called INTREX®,which protects them from the corrosive flue-gas (see Figure 6). The bubbling solids efficiently conduct their heat tothe steam contained in the coils and as the solids never meltor soften, fouling and corrosion of these coils are minimal.Furthermore, because the high heat transfer rates of thesolids (by conduction), the coil size is many times smallerthan those in conventional boilers.

Fuel delivery systemA final important design issue involves the fuel deliverysystem to the boiler. A PC boiler requires the fuel to be finelyground and pneumatically transported and distributed tomany burners. For low-quality, high-ash fuels such as browncoals and lignite, the power consumption of fuel pulverizers

increases dramatically and the fuel delivery system requiresmore maintenance as its reliability declines. A CFB boilerdoes not require pulverizers as its fuel is only coarselycrushed and fed to the CFB boiler directly from the fuel silosvia a simple gravity feed system.

Overall plant reliabilityBased on these process and design differences, CFB powerplants have demonstrated plant availabilities well aboveconventional PC boilers, as shown by a recent studycomparing PC plant availability to Amec Foster Wheeler CFBboilers (see Figure 7). Availability is defined as a percentageof 8760 hours, the total number of hours that a plant can beoperationally available. The total includes both planned andunplanned downtime.

Power plants with CFB boilers had about a 5% (absolute)higher availability than PC plants, and this higher availabilityis maintained even for brown coals and lignites. For a 1000 MWe supercritical coal power plant, this 5% differencein plant availability can translate into a $160 million increasein power plant net income on a 10-year net present value(NPV) (see Figure 8).

Environmental performance and equipmentrequirements: PC versus CFBFrom an environmental aspect, the low-temperature CFBcombustion process (850°C for CFB versus 1500°C forPC/oil/gas) produces less NOx and allows limestone to be feddirectly into the furnace to capture SOx as the fuel burns. Inmost cases SCR or a FGD is not needed, which dramaticallyreduces the plant installed and operating cost and waterconsumption while improving plant reliability and efficiency.For a 1000 MWe power plant, the savings alone on the costsof SCR and FGD would be in the range of $250 million–$300million.


▲


Figure 6 – A comparison of boiler design features: PC boilers versus CFB boiler

Figure 7 – Availabilities* of PC and CFB power plants. *Availabilitymeans total time plant is available to run accounting for both plannedand unplanned downtime. The Amec Foster Wheeler CFB plantavailability derives from client-supplied data reported over the period2000–2008 for CFB plants located mainly in Europe. The PC valuesderive from client-supplied data over the period 2002–2011 for PC unitsthat are mainly located in Europe ¹

Avai

labi

lty (%

)

1 VGB PowerTech 2012. Availability of Thermal Power Plants2002-2011 – Report VGB-TW103Ve

A permanent change to the global coal marketSince 2005, Indonesian coal exports have grown faster thanall other countries combined; nearly quadrupling to 400 Mt in2013 (see Figure 9). Projections into the future predictIndonesian exports reaching nearly 500 Mt by 2030, abouttwice that of Australia, the world’s second largest exporter ofcoal.

Today about 50% of the coal exported from Indonesia islow-quality, high-moisture sub-bituminous in quality withgross-as-received (GAR) higher heating values ranging from3900 to 4200 kcal/kg, well below the 6000 kcal/kgbenchmark used in the international coal market for the last50 years.

Over the last 3 years, the quality of Indonesia’s exportcoal has been declining, and this trend is expected tocontinue well into the future. Today about 60% ofIndonesia’s coal mines hold low-rank sub-bituminous coals.The other 40% hold bituminous coals estimated to haveheating values less than 5200 kcal/kg. The heating value ofIndonesia’s export coals has been steadily declining and isforecast to continue, a trend that reflects the impact of miningthis lower quality coal.

The primary driver for the ballooning share of Indonesiancoal in the international coal market is simple economics. Thecurrent and forecast price discount between Indonesia’s sub-bituminous 4200 kcal/kg Ecocoal and a 6000 kcal/kgAustralian thermal coal, both on a net-as-received basis(NAR), shows a steady pattern: a 48% or $55 per metric tonaverage discount for the lower quality Indonesia coal over theperiod from 2012 to 2020 (see Figure 11). The difference in

heating value, which amounts to 30% on a comparativeenergy basis, translates into a very attractive net 18%discount for the Ecocoal, a benefit that goes right to thebottom line of a power plant’s balance sheet.

Since fuel cost makes up about 85–90% of the totaloperating cost of a large power plant, it would be foolish toignore the economic benefits of using low-quality fuels. Wecan see this in several domestic markets, where low-qualitycoals and lignites play a major role in power production. Forexample, 77% of Germany’s solid fuel power is producedfrom lignite; only 23% is produced from hard coal. In theUSA, 54% of the solid fuel power comes from low-qualitysub-bituminous coals. Use of low-rank coals and lignites forpower production is growing in Turkey, India, China,Indonesia, Australia, South Africa, and Mozambique, a trenddriven by the very low cost of these fuels relative to premiumcoals.

Until recently, low-quality coals and lignites have beenconfined to domestic markets and have not been part of theinternational coal market. This is because their economicbenefit is quickly eroded by their transportation costs, owingto the lower energy contents of the coals. But today we seemore low-quality coals and even lignites coming into theglobal coal market, a move that is driven by steep pricediscounts in a tight market for premium coals. From 2001 to2010 for example, Korean imports of Indonesian coals(mostly sub-bituminous) increased sevenfold by 38 Mt, whileimports from Australia coal grew by only 13 Mt.



Figure 10 – Average gross heating value of Indonesian export coal. Source: marketing, sales and logistics analyst, Banpu PCL

Figure 11 – Difference in prices between Indonesian Ecocoal andAustralian thermal coal delivered (CIF) to the coast of South Korea.Prices shown are nominal. Source: Amec Foster Wheeler forecast

Net

inco

me

($/m

)

Figure 8 – Impact of plant utilization factor on annual plant net incomefor a 1000 MWe supercritical steam power plant operating at autilization factor of 90% and receiving a $100 per MWe electricity tariffbased on buying coal at $100 per ton

Expo

rt vo

lum

e

Figure 9 – Global coal exports. Source: historical data and Amec FosterWheeler projections


This trend is not expected to change any time soon.Instead, it looks to be a permanent shift towards a moreflexible coal market, where buyers and sellers trade price forcoal quality, similar to markets in many other commoditiesand finished goods.

The impact of the changing coal market on coalboiler technologyThis price versus quality shift in the global coal market willlikely be viewed as good news by some observers and badnews by others, with the responses depending on their powerplant technology position. PC power plants with tight coalspecifications (here one thinks of supercritical designs) willhave a limited ability to use the discounted coals. Theseplants will have to choose either to stay within the tighteningpremium coal market or to venture into the broader coalmarket and trade lower plant outputs, reduced reliability, andhigher maintenance costs for discounts in the cost of fuel.

On the other hand, the shift will come as good news forpower generators utilizing CFB technology. Owing to theCFB’s fuel flexibility, plant owners can access the full rangeof discount coals (even for ultra-supercritical designs),buying fuels for maximum economic benefit while avoidingthe high-priced premium coals. Furthermore, the impact ofdeclining coal quality on plant output, reliability, andmaintenance is minimized with a CFB, and the risk of futurecarbon regulation is lessened because of the CFB’s ability toutilize biomass and other carbon-neutral fuels.

For new power plants, this trend clearly increases thevalue of fuel-flexible coal plants such as those utilizing CFBtechnology and will likely push towards (if not accelerate) theadoption of CFB technology in large coal-fired utility plants.The timing seems right, as CFB technology has demonstratedits capabilities in serving the utility power sector. This is notto say that new PC boiler power plants cannot be designed toburn low-rank fuels. They can. The point for consideration isthat once a PC is designed to use a specific low-rank fuel, the

plant has difficulty burning other fuels without adverselyaffecting plant performance, reliability, and maintenance.

The economic benefits of CFB technology at theutility scaleTo quantify the benefits of CFB technology on a large utility-plant scale, Amec Foster Wheeler conducted a studycomparing both the technical and economic performances oftwo supercritical 1100 MWe (gross) power plants. One of theplants used conventional PC technology and the other CFBtechnology. The study involved the development of fullpower-plant financial models, heat and material balances, aswell as conceptual plant designs for plant layout, sizing, andcost estimation purposes. For the purpose of comparison, anumber of performance metrics were evaluated. Theyincluded plant capital and operating costs, plant height andfootprint, reliability, atmospheric emissions, solid and liquidinputs, and waste streams.

The PC plant was configured with a single 1100 MWeultra-supercritical boiler that provided its steam to a single1100 MWe steam turbine generator. The plant fired anAustralian bituminous thermal coal with an NAR heatingvalue of 5500 kcal/kg and a sulphur content of 0.35%. Thecoal was priced at $95 per metric ton. SCR was installed inthe boiler to control stack NOx emissions to 50 ppmv (6% O2dry) and wet limestone FGD was installed behind the boiler tocontrol stack SOx to 50 ppmv (6% O2 dry).

The CFB plant was configured with two 550 MWe ultra-supercritical boilers that provided steam to a single 1100 MWe steam turbine generator. The CFB plant fired anIndonesian sub-bituminous thermal coal (Ecocoal) with anNAR heating value of 4200 kcal/kg and sulphur content of0.27%. The coal was priced at $55 per metric ton. SCR wasinstalled in the boiler to control NOx emissions to 50 ppmv(6% O2 dry), but no separate FGD was installed behind theCFB boiler, for the boiler itself used limestone to control stackSOx to 50 ppmv (6% O2 dry).

▲


Table I

A comparison of capital costs of 1100 MWe supercritical PC and CFB power plantsNote: Absolute design and supply boiler cost depends on scope. Source: Amec Foster Wheeler study

We compare the capital costs of boiler and pollutioncontrol equipment on a design and supply basis, excludingerection (see Table I). Even though the cost of two CFBboilers burning a low-rank coal is about 11% higher than thecost of a single large PC boiler burning a high-quality coal,meeting emission targets without installing an FGD for theCFB boilers resulted in a net $93 million savings in capital forthe CFB plant configuration.

As for operating costs (see Table II), using the discountedIndonesian coal, the CFB plant saves $66 million annually infuel costs. Adding in other operating costs such as limestone,ash disposal, gypsum sales, and maintenance increasessavings to $69 million. These savings are worth $424 millionin NPV over a 10-year period.

A full financial proforma model for both the PC and CFBplant configurations was developed to calculate the levelizedelectricity production cost for each plant configuration. Inaddition to total capital and operating costs, the proformaanalysis takes into account plant utilization, financingconditions and terms.

Figure 12 compares the proforma analyses and thecomponents that make up electricity production costs. Thesmaller capital and fuel cost components for the CFB plantresults in a net savings of $10 per megawatt-hour ofelectricity produced. This translates into $82 million annuallybased on 90% plant utilization: the savings are worth $503million NPV over a 10-year period (see Table III).



Table II

A comparison of costs of 1100 MWe supercritical PC and CFB power plants. Source: Amec Foster Wheeler study

Table III

A comparison of annual and NPV electricity production costs for 1100 MWe supercritical PC and CFB power

plants. Source: Amec Foster Wheeler study


Finally, Table IV compares other plant parameters andperformance metrics, highlighting that both the CFB and PCplants meet the same stack emission limits, but as the CFBplant does not have a separate wet FGD for SOx control, itsaves about 2 × 106 m3 of water annually.

Conclusions and observationsSix years of successful operation of the large supercriticalonce-through CFB boiler at the Łagisza power plant in Polandhas demonstrated CFB technology for utility powergeneration. KOSPO reinforces this conclusion by selectingAmec Foster Wheeler CFB technology for its 2200 MWe GreenPower Project in Samcheok, Korea.

Because combustion is flameless and occurs at lowtemperatures, CFB technology offers many benefits for utilitypower generation. Its fuel flexibility, reliability, and ability tomeet strict environmental standards with minimal post-

combustion pollution control equipment are highly valuedbenefits for utilities. Additionally, the CFB’s load-followingflexibility (CFB has the same load ramp rates as a PC, butbetter turndown) is another important value for gridscontaining a high level of intermittent renewable power. As anexample, the Łagisza unit cycles daily between 40 and 100%MCR to meet the requirements of the Polish national grid.

The CFB benefits become more compelling whenconsidering low-quality fuels. The technology is able toprovide smaller, less costly boilers as fuel quality declines,while achieving plant availabilities well beyond conventionalPC boiler technology.

The global coal market is moving away from traditionallyrigid, single-specification coal towards a more flexible price-for-coal-quality market. The convergence of the coal marketshift with the CFB’s entry into utility power application isexpected to speed up the adoption of CFB technology in thelarge utility power sector.

Owing to the large economic benefit, the use of domesticbrown coal and lignite for utility power generation is growingin Germany, Turkey, and Indonesia, all of which haveabundant supplies of economical low-quality coal andlignites. It is expected that CFBs will be utilized more in thesemarkets.

A technical and economic study conducted by AmecFoster Wheeler showed that a large utility CFB power planthas a compelling economic advantage over a traditional PCpower plant, mainly because the CFB plant does not requirepost-combustion FGD equipment and can utilize a low-qualityIndonesian coal. The numbers indicate that a 1100 MWe CFBpower plant would cost $93 million less to build and wouldproduce a net saving in the cost of producing electricity ofabout $82 million annually, worth $503 million on a 10-yearNPV basis. In today’s price-sensitive global utility marketthese numbers deserve serious consideration. ◆

▲


Table IV

A comparison of emissions, plant efficiency, fuel, limestone, ash, and FGD water flow in 1100 MWe supercriticalPC and CFB power plants. Source: Amec Foster Wheeler study

Figure 12 – A comparison of levelized electricity production costs for1100 MWe supercritical PC and CFB power plants. Source: Amec FosterWheeler study

IntroductionDuring the last three decades, primary energyconsumption has increased worldwide byabout 70% (Figure 1), reaching 11 Gt oilequivalent (Gtoe) at the end of 2009. Therewas a rapid increase in oil and natural gasconsumption, sharing 35% and 25% respec-tively of the total consumption.

Global coal demand growth under the NewPolicies Scenario (International EnergyAgency, 2010) will be around 20% between2008 and 2035, with 100% of this increaseoccurring in non-OECD countries. Global coaldemand is expected to peak around 2025 andbegin to decline slowly, returning to 2003levels by 2035 due to the restrictions imposedby climate policy measures.

Coal possesses the largest potential of allnon-renewable fuels and provides 56% of thereserves and 89% of the resources worldwide(Andruleit et al., 2013) (Figure 2). Coal, beingthe most abundant, available, and affordablefuel, has the potential to become the mostreliable and easily accessible energy source

and thus to provide a crucial contribution toworld energy security.

The major challenges facing coal areconcerned with its environmental impacts bothin production and in use. Various pollutantcontrol systems have been developed over thepast few decades and are continually evolving.These new technologies, which facilitate theuse of coal in a more environmentally friendlyway by drastically reducing pollutantemissions, are commonly known as clean coaltechnologies (CCTs). Within this concept twodifferent approaches can be considered,namely (i) reducing emissions by reducing theformation of pollutants during the coalconversion process, and (ii) developingsystems with higher thermal efficiency, so thatless coal is consumed per unit powergenerated, together with improved techniquesfor gas cleaning and for residues use ordisposal.

Nowadays, state-of-the-art combustionsystems can reach plant net efficiencies of43–45% (LHV) (Klauke, 2006; ABB Ltd, n.d)utilizing high-rank coals. With someexceptions (Germany), the use of low-rankcoals is still problematic due to the low plantefficiency and high pollution potential.

Coal gasification, being a CCT, provides anenvironmentally friendly and efficient solutionnot only for power production, but also for theproduction of a variety of chemicals such asmethanol, ammonia, and hydrogen, as well assynthetic fuels such as synthetic natural gas(SNG), gasoline, and Fischer-Tropsch liquids.

Gasification of low-rank coals is even moreattractive due to the low prices of coal, its localavailability, and high prices or even non-availability of other resources such as naturalgas and oil.

Gasification of low-rank coal in theHigh-Temperature Winkler (HTW)processby D. Toporov* and R. Abraham*

SynopsisGasification is a process of thermal conversion of solid carbonaceousmaterials into a gaseous fuel called syngas. Coal gasification is an efficienttechnology for a range of systems for producing low-emission electricityand other high-value products such as chemicals, synthetic fuels, etc.

The paper presents the High-Temperature Winkler (HTW) gasificationprocess, which is designed to utilize low-rank feedstock such as coals withhigh ash content, lignite, biomass etc. The process is characterized by abubbling fluidized bed, where coal devolatilization and partial oxidationand gasification of coal char and volatiles take place, and by a freeboardwhere partial combustion and gasification of coal char take place.

The recent development of the high-pressure HTW process is reviewed.Gasification of low-rank, high-ash coals with respect to gasificationtemperatures, conversion rates, and syngas quality is also discussed. Themain HTW design steps required for an industrial-scale design arepresented. Special attention is given to the process modelling, includingglobal thermodynamic calculation as well as detailed CFD-basedsimulation of a reacting fluidized bed. Three-dimensional numerical resultsof the HTW process are also provided and discussed.

Keywordscoal gasification, high-temperature Winkler process, HTW, reactingfluidized bed simulation, high-ash coal, low-rank coal, biomass, peat,municipal solid waste.

* Gas Technology Division, ThyssenKrupp IndustrialSolutions AG, Dortmund, Germany.

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. This paperwas first presented at IFSA 2014, IndustrialFluidization South Africa, Glenburn Lodge, Cradleof Humankind, 19–20 November 2014.



Gasification of low-rank coal in the High-Temperature Winkler (HTW) process

The High-Temperature Winkler (HTW) gasificationprocess was specially developed for utilization of low-rankfeedstock such as lignite, biomass, sub-bituminous coalswith high ash content etc. The technology development steps,the process description, as well as the design steps arediscussed in detail in the following sections.

Description of the High-Temperature Winkler(HTWTM) gasification process

HistoricalThe HTWTM fluidized-bed gasification process is based on theWinkler generator, which was developed in the 1920s inGermany by Industrie Gewerkschaft (IG). From 1920 to 1930IG investigated the possibility of using low-rank local coals,such as brown coal, instead of expensive coke, for synthesisgas production and subsequent production of ammonia andmethanol. Dr. Winkler in 1921 conceived the idea of using a

‘boiling’ bed, i.e. using particles of fuel small enough to bealmost gas-borne and hence comparatively mobile. Undersuch conditions the fuel bed behaves very much like a liquid;the gas passing through the fuel gives an appearance as if thebed were boiling, the bed finds its own level, as does a liquid,and circulation of particles within the bed is such as to givesubstantially equal temperatures throughout the bed. This iswhat we nowadays call a fluidized bed.

The first Winkler generator was put into operation atLeuna, Germany in 1926, making power gas and having acapacity of 40 000 Nm3/h. In 1930 the production ofnitrogen-free water gas began, which was obtained bycontinuous blast of pure oxygen with steam (Figure 3).

Commercial-scale Winkler gasifiers were operated atatmospheric pressure in over 40 applications around theworld. Since 2000 more than 40 new atmospheric units havebeen built in China alone. Thus, the Winkler gasificationprocess became a widely used technology, characterized bythe following advantages:

➤ Low oxygen consumption due to moderate temperatures➤ Optional use of air or pure oxygen as an oxidant➤ Simple coal preparation➤ Good partial load behaviour over a wide range of

operating conditions➤ Simple start-up and shut-down procedure➤ High operational reliability➤ No by-products in the raw gas, such as tars, phenols, and

liquid hydrocarbons, etc.In the 1970s, ThyssenKrupp Industrial Solutions (former

Uhde) together with Rheinische Braunkohlenwerke AG (nowRWE AG) commenced with the development of a pressurizedversion of the Winkler gasifier – the High-TemperatureWinkler (HTWTM) gasification process. The developmentprocess went through several steps that involved buildingand operating pilot, demonstration, and commercial plantsoperating at increased pressure, as shown in Figure 4.

▲


Figure 1 – Development of primary energy consumption worldwide(cumulative) and projections of IEA until 2035 (International EnergyAgency, 2010)

Figure 2 – Global share of all energy resources in terms of consumption as well as the production, reserves, and resources of non-renewable energyresources as at the end of 2012 (Andruleit et al., 2013)

This development led to several major enhancements tothe advantages of the atmospheric Winkler gasifier.

➤ By increasing the pressure to 10 bar, the reaction rateswere increased and thus the specific performance perunit cross-sectional area of the gasifier was increased,while the compressive energy required for thesubsequent chemical synthesis was reduced

➤ By increasing the temperature, the methane content inthe raw gas was reduced and the carbon conversionrate, and thus the gas yields, increased

➤ By recirculating the dust fines entrained from thefluidized bed it was possible to increase the carbonconversion rate

➤ Inclusion of proven and robust systems such as drydust filtration and waste heat recovery

➤ Ability to handle a great variety of feedstock (coal,peat, biomass, municipal solid waste (MSW) etc.) andhigh flexibility regarding particle size of the feedstocks

➤ High cold gas efficiency➤ Stable and smooth gasifier performance with great

inherent safety due to the large carbon inventory.HTW gasification plants, like the Oulu plant (Finland)

gasifiying peat for ammonia, the Niihama plant (Japan)gasifying MSW for power, and the Berrenrath plant(Germany) gasifying German brown coal for methanolproduction, have been operated on a commercial basis, whichhas resulted in the technology attaining industrial maturity.The Berrenrath plant was in operation for more than 12 yearsand is an excellent reference for the HTWTM gasificationtechnology (shown in Figure 5).



Figure 3 – Earlier Winkler generators (i) with traveling grate, 1930s (left) and (ii) grateless modification, 1940s (right) (Von Alberti and Rammler, 1962)

Figure 4 – Stages in the development of HTW gasification


Some typical operating figures, obtained for industrial-scale HTWTM gasification of different feedstock are listed inTable I.

The HTWTM processThe HTWTM process involves (Figure 6) a gasification unitconsisting of a feeding system, the gasifier itself, a bottomash removal system located below the gasifier, and a gas exitin the head of the gasifier with a cyclone. In the subsequentsteps the raw syngas is cooled and de-dusted and thenfurther treated in accordance to the needs of the downstreamprocesses. Screw conveyors or gravity pipes (according to thefeedstock) supply the feedstock to the HTWTM gasifier. Due tothe gasifier pressure, feeding and bottom ash removal have tobe performed by lock-hopper systems.

The gasification is controlled using the gasification agentssteam and oxygen (or air), which are injected into the gasifiervia separate nozzles. The nozzles are arranged in severallevels which are located in both the fluidized bed (FB) zoneand the freeboard zone (also called the post-gasificationzone). A high material and energy transfer rate is achieved inthe FB and this ensures a uniform temperature distributionthroughout the fluidized bed. In order to avoid the formationof particle agglomerations the temperature is maintainedbelow the ash softening point.

Additionally, the gasification agents are injected into thepost-gasification zone in order to improve the syngas qualityand the conversion rate by increasing the temperature.

In summary, the industrial-scale pressurized HTWTM

process is characterized by two temperature zones, namelythe fluidized bed with an operating window between800–1000°C and a post-gasification zone with temperaturelevels between 900 and 1200°C.

The cyclone separates approximately 95% of theentrained solids from the syngas and returns them to the FBof the gasifier, thus increasing the overall carbon conversion

▲


Figure 5 – HTW demonstration plant, Berrenrath, Germany. Productionrates: 300 t /d methanol, total opperating hours: 67 000 h (Renzenbrinket al., n.d)

Table I

HTWTM gasification process data for industrial-scale plant

Parameter German lignite Finnish peat High-ash hard coal* Dimensions

Feed 23.2 21 76.2 t/h (d.a.f.)C 68 58 70.8 wt.% d.a.f.H 4.9 6.0 6.0 wt.% d.a.f.O 25.7 33 20.7 wt-% d.a.f.N 0.7 1.9 1.7 wt.% d.a.f.S 0.6 0.3 0.8 wt.% d.a.f. ID.T. reduced >1 150 1 270 °CAsh content 4.0 7.0 48 wt.% dryParticle size range** 0-6 0-4 0-3 mmMoisture content 12 15 3 wt.%of the feed, as gasified***Thermal input 140 140 600 MWOperating pressure 10 10 30 barFluid bed temperature 810 720 870 °CFree board temperature 900 1030 1,100 °CSyngas quality CO 45 35 48 Vol. % (N2 and H2O free)H2 34 33 28 Vol. % (N2 and H2O free)CO2 17 27 21 Vol. % (N2 and H2O free)CH4 4 5 3 Vol. % (N2 and H2O free)Carbon conversion efficiency 95.5 90 93 Carbon in dry gas / carbon in feed, %Synthesis gas (CO+H2) yields 1 500 1 000 1 440 Nm3/t of feed, d.a.f.Specific oxygen consumption 0.39 0.36 0.55 O2 Nm3/kg of feed, d.a.f.Cold gas efficiency 85 75 75 %

(100 x Heating value of product gas, MWHHV / Heating value in feedstock, MWHHV)

* Estimated for coals of Indian origin ** Typical grain size for HTW process ranges between 0 and 10 mm. Coal fines can be used *** Before the gasification process the feed has to be dried to about the inherent moisture of the feed in order to improve the flow behaviour and due toeconomics (usually the content of lignite is 10 to 20% by weight).

rate. Downstream of the gasifier, the raw syngas is cooled inthe raw gas cooler and the heat is used to produce saturatedsteam that can be exported to external steam consumers.After the raw gas cooler the remaining fine ash particles areremoved from the syngas in the ceramic filter. The fly ash isfurther cooled and then discharged from the pressurizedsystem using a lock-hopper system. Subsequently, thesyngas is sent to the scrubbing system, where it is quenchedwith water to remove the chlorides. The syngas is saturated,thus making further chemical treatment like the CO-shifteasier.

Mathematical modelling of the HTWTM gasificationprocessThe gasification of solid fuel (coal, peat, biomass, MSW etc.)is a complex process governed by a number of physical andchemical phenomena. The principal steps by which thereaction progresses are the thermal decomposition of the rawfuel and the subsequent burnout of the char and the volatilematter. The following main reaction steps (Equations[1]–[10]) typically summarize the process of coal gasifi-cation:

[1]

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

When heated, coal decomposes into char and volatilematerial (Equation [1]), the former reacting slowly in thefluidized bed and in the post-gasification zone (Equations[7]–[10]), while the volatile material, consisting mainly ofwater, CO, CO2, CH4, tars, H2, and some other lighthydrocarbons conditionally named as CNHMOL, is assumed torapidly form CO and H2 (Equations [2], [3], and [4]) as themost simple reaction mechanism. Gasification temperaturesare normally so high that no hydrocarbons other thanmethane can be present in any appreciable quantity(Equations [2], [6], and [10]).

Numerical simulation of a reacting fluidized bed reactor isnot a trivial task. Prediction tools based on three differentmodel approaches for simulation of gasification of solid fuelin a fluidized bed are used at ThyssenKrupp IndustrialSolutions AG. These are based on the following methods:

➤ Black-box methods (BBM): a zero-dimensional modelresolving the overall mass and heat balances over theentire gasification reactor

➤ Fluidization methods (FM): a one-dimensional



Figure 6 – Schematic of the HTWTM process


steady-state model, which avoids the details of complexgas-solid dynamics but still maintains the fluiddynamic effects by assuming (using externally derivedempirical correlations) a multiphase pattern in the bed.Here the particle-particle interactions are not accounted

➤ Computational fluid dynamics methods (CFDM): athree-dimensional unsteady model, which considers thefluid dynamics, gas-particle, and the particle-particleinteractions entirely.

HTWTM process simulation as a design tool

The BBM approachNormally the BBM approach is based on the equilibriummodel, assuming that equilibrium is attained in the outletstreams. Equilibrium simulations yield almost no oxygen,solid carbon (above 800°C), or tar. In practice, however, thecontents of hydrocarbons and char in the gas are far fromzero, thus showing a strong kinetic limitation. Therefore, forcalculation of the HTWTH process the so-called ‘pseudo-equilibrium’ approach, implemented in HTW-specialized in-house software, is used. This method ‘supports’ theequilibrium method with empirical correlations accumulatedduring long-term operation of the HTW gasifier. A simplifiedschematic diagram of this approach is shown in Figure 7.

The pseudo-equilibrium approach allows solid carbon(present in the bottom ash product and in the dust), H2S,COS, HCN, NH3, and some heavy hydrocarbons like C6H6 andC10H8 to be contained in the outlet gas and the correspondingquantities of carbon, sulphur, nitrogen, and hydrogen arediscounted from the feedstock. This approach is supported bya large empirical database containing operational data fromdifferent HTW gasifiers.

Thus the remaining feedstock elements and the gasifi-cation agents react to attain equilibrium. The outlet gas isthen obtained by summing the gas components given by theequilibrium and those taken off initially. The underlyingreason for this approach is that the decomposition of tar andthe char conversion by gasification, as well as the sulphurand nitrogen chemistries, are mainly kinetically limited.

For this reason the reaction mechanism and the reactingspecies participating in this mechanism are pre-defined basedon experimental data obtained at the HTW pilot ordemonstration plant.

Additionally, the equilibrium of the reactions is evaluatedat a lower temperature, a ‘quasi-equilibrium temperature’,than the actual process temperature. In this way the discrep-ancies of the equilibrium predictions are attributed totemperature gradients from the bubbling fluidized bed to thepost-gasification zones by which the HTW gasifier is charac-terized. Therefore, the temperature is modified to obtain areasonable correlation with the existing HTWTM empiricaldata.

Furthermore, the split between bottom ash and dust aswell as their elemental compositions is made on the basis ofempirical correlations taken from real operating conditions.

In case there is no empirical data available for a specificfeedstock quality, the following pre-design steps are required:

➤ Laboratory determination of key feedstock parameters,such as ultimate and proximate analysis; coal ashanalysis; ash softening temperature in a reducingatmosphere; coal char reactivity; physical properties(such as bulk density and true density); bulkfluidization behaviour; calorific values etc.

➤ Determination of the key operational parameters. Realgasification tests are performed at the state-of-the-artHTWTM pilot plant (0.5 MW thermal input) shown inFigure 8. These tests are required in order to obtain realdata about the gasification temperature in both thefluidized bed and in the post-gasification zones, thecomposition of syngas, bottom ash, and dust (includingtrace elements, tars etc.), carbon conversion, agglom-eration limits, fluidization behaviour etc.

After obtaining the key feedstock and operationalparameters, the HTW-specialized in-house software using thequasi-equilibrium approach as described above can be usedfor obtaining information about:

➤ Syngas composition, production rates, and HHVs➤ Bottom ash and dust composition, production rates,

and HHVs➤ Cold gas efficiency, carbon conversion➤ Utilities (air or oxygen, steam, carbon dioxide, water,

etc.)for a given industrial-scale HTW geometry, operatingpressure, and temperature.

In practice, this relatively simple approach is very helpfulfor quick estimation of the performance of an industrial-scaleHTW gasifier for a given feedstock, load, gasification agents,pressure, and temperature.

The FM approachThe FM-based Pressurised Fluidised Bed Gasification (PFBG)program was developed at Siegen University, Germany(Hamel , 2001; Dersch and Fett, 1997) in a researchprogramme with ThyssenKrupp Industrial Solutions (TKIS)AG.

The program divides the computational domain intoseveral zones arranged in series (cells) along the main gas-particle flow path. Each cell is subdivided into a solid-freebubble phase and an emulsion phase. The emulsion phase isassumed to contain some gas and all the solid; the gas andparticles are perfectly mixed in this phase, whereas the gas

▲


Figure 7 – A simplified schematic diagram of the HTWTM quasi-equilibrium model

phase is considered to be in plug-flow. The flow pattern ineach cell is defined using semi-empirical correlations todetermine the distribution of gas species and reactingparticles, the rising velocity, minimum fluidization velocity,terminal velocity, bubble size, bubble void fraction, bubblevelocity, bubble–to-emulsion mass transfer, starting particlevelocity and other relevant variables. A system of one-dimensional conservation equations for each species is solvedby neglecting diffusion. The cell temperature is calculatedfrom the energy balance for each cell, whereas the massbalance equations are formulated separately for bubble andemulsion phases in each cell as shown in Figure 9.

The program uses kinetic data for drying,devolatilization, char conversion, and homogeneousreactions. The parameters for these reactions can be changedaccording to the specific feedstock. Normally the kinetic datais taken from the literature or (better) from experimentsperformed at the pilot plant or in similar operating conditionsto those in the gasifier.

The program is validated using the composition of the gasat the gasifier outlet. The freeboard temperature profile wasvalidated against measurements in laboratory-scale and full-scale gasifiers, as can be seen in Figure 10. The validationsshow reasonably good agreement for the main species at theoutlet. The deviations between the measured and thecalculated values for the CO, H2, H2O, and CO2 mole fractionspoint to an insufficiency of the implemented reaction modelsand can be attributed to the kinetic data available, inparticular for the water-shift reaction.

According to the literature (Gomez-Barea and Lekner,2010), the model from Siegen is among the most advancedFM models developed to date.

In practice, this much more complex approach comparedto the BBM approach is helpful for estimation of the influenceof the kinetic data on the performance of an industrial-scaleHTW gasifier for a given feedstock, load, gasification agent,pressure, and temperature.



Figure 8 – State-of-the-art HTWTM pilot plant (0.5 MW thermal input),located at the Technical University of Darmstadt, Germany

Figure 9 – Simplified schematic diagram of the FM model from the University of Siegen (Hamel, 2001; Dersch and Fett, 1997)


The CFD approachThree-dimensional simulation of coal combustion/gasificationsystems is based on a complex mathematical model, whichincludes modelling of the fluid flow, turbulence, chemicalreactions, and heat transfer in an Eulerian framework, andmodelling of the coal particles transport, heterogeneousreactions, and the associated momentum, heat, and masstransfer with the surrounding reacting fluid in a Lagrangianframework. This approach is widely used for pulverized fuelcombustion and gasification systems. However, it cannot beapplied for flows with a high volume fraction of solid matter,which are typical for fluidized bed systems.

Modelling of particle-dense flows such as in fluidizedbeds requires the introduction of new model approaches thatconsider the particle-particle interactions. Therefore incomparison with other applications such as entrained flowcombustion and gasification, three-dimensional simulation ofa reacting fluidized bed is still in the very early stage ofdevelopment and application.

TKIS AG is currently using two commercial CFD softwarepackages and also develops open-source code for 3Dsimulations of the HTW gasification process. The models arebased on a simplified discrete element method (DEM)approach assuming particles that have similar physicalproperties to form a cluster of a so-called ‘numerical’ particle.Thus the total number of the simulated particles can bereduced and simulations can be performed in an acceptabletime schedule. Such simulations are unsteady by their natureand therefore unsteady Reynolds-averaged Navier-Stokes(URANS) or large eddy simulation (LES) methods are usedfor turbulence modelling. The chemistry is modelled usingkinetic data for both homogeneous and heterogeneousreactions.

TKIS AG is actively using the CFD technique for design,assessment, and optimization of the operating conditions ofnew HTW gasifiers. Some preliminary CFD results, obtainedby the authors, for the HTW gasifier in Berrenrath can beseen in Figure 11.

As can be seen from the distribution of the particlevolume fraction, shown in Figure 11 (left), the coal afterentering the gasifier undergoes fast devolatilization. Smallparticles are entrained to the freeboard where they have aresidence time of approximately 10-15 seconds and can reactwith gasification agents before being separated by thecyclone and returned to the fluidized bed. The larger particles,together with the bed material (typically coal ash), form afluidized bed with a height about the same as the height ofthe conical section of the gasifier. The residence time of theparticles inside the fluidized bed is long enough to achievehigh carbon conversion rates.

The gas temperature distribution in the middle plane ofthe gasifier (Figure 11, right) clearly shows two temperaturezones: (i) lower temperature and uniform temperature distri-bution inside the fluidized bed, and (ii) higher temperature inthe post-gasification zone above the fluidized bed. The highertemperature is achieved with a controlled supply of oxygenjust above the fluid bed. Thus several effects are achieved,namely (i) the volatiles (hydrocarbons such as tars), whichare released during the devolatilisation inside the fluid bed,are oxidized and/or cracked, and (ii) faster endothermicgasification reactions take place in the post-gasification zone.

In practice, the CFD predictions of the HTW process canbe used for estimation of the influence of kinetic data on theperformance of an industrial-scale HTW gasifier for a givenfeedstock, load, gasification agent, pressure, andtemperature. Furthermore, the 3D information obtained from

▲


Figure 10 – Example of HTW Berrenrath simulation and validation of the FM model from the University of Siegen

such predictions can be successfully used for processoptimization, improved design of the cyclone, finding theoptimum number of nozzles and their position, localtemperature control inside the fluidized bed and also in thepost-gasification zone, etc.

ConclusionsThe High-Temperature Winkler (HTWTM) gasification processis characterized by a reacting bubbling fluidized bed operatedat elevated pressure and temperatures, thus achieving highefficiency and high flexibility in terms of feedstock quality,reaction conditions, throughput, and syngas quality. The useof nozzles for supplying the gasification agents (oxygen,steam, and CO2) provides the opportunity to achieve uniformfluidization conditions and high flexibility in temperature andstoichiometric conditions along the gasifier height.

HTWTM is a mature gasification technology for utilizationof low-rank solid feedstocks such as high-ash sub-bituminous coals, lignite, peat, biomass, and MSW. Morethan 30 years of intensive R&D has led to building andoperation of several industrial-scale gasifiers producingsyngas on a commercial basis for many years.

Recent developments made by ThyssenKrupp IndustrialSolutions AG are focused on widening of the feedstockportfolio and improving the design by an intensive researchand development programme based on both experiments atthe HTW pilot plant and numerical simulations using thenewest achievements in modelling of reacting fluidized bedprocesses.

ReferencesABB LTD. Not dated. The State of Global Energy Efficiency. Global and Sectional

Energy Efficiency Trends. Corporate Communications Report, ABB Ltd,Zurich, Switzerland. www.abb/energyefficiency

ADLHOCH W., SATO, K., WOLFF, J., and RADTKE, K. 2000. High Temperature

Winkler gasification of municipal solid waste. Gasification TechnologiesConference, San Francisco, CA, 8–11 October 2000.

ANDRULEIT H., BAHR, A., BABIES, H-G., FRANKE, D., MESSNER, J., PIERAU, R.,

SCHAUER, M., SCHMIDT, S., and WEIHMANN, S. 2013. Reserves, Resources and

Availability of Energy Resources 2013. Energy Study, Bundesanstalt für

Geowissenschaften und Rohstoffe (BGR), Hannover.

DERSCH J., and FETT, F. 1997. Anleitung zur Benutzung des

Simulationsprogramms PFBG „Pressurised Fluidised Bed Gasifier“.

University of Siegen.

GOMEZ-BAREA, A. and LEKNER, B. 2010. Modelling of biomass gasification in

fluidised bed. Progress in Energy and Combustion Science, vol. 36. pp.

444–509.

HAMEL, S. 2001. Mathematische Modellierung und experimentelle

Untersuchung der Vergasung verschiedener fester Brennstoffe in

atmosphärischen und druckaufgeladenen stationären Wirbelschichten.

PhD thesis, University of Siegen.

INTERNATIONAL ENERGY AGENCY. 2010. World Energy Outlook 2010.

KLAUKE, F. 2006. Moderne und umweltfreundliche Kohlekraftwerke als

essentieller Baustein zur globalen CO2-Reduktion. Workshop at RWTH

Aachen University, 13 July, 2006.

RENZENBRINK W., WISCHNEWSKI, R., ENGELHARD J., and MITTELSTADT, A. Not dated.

High Temperature Winkler (HTW) Coal Gasification – A Fully Developed

Process for Methanol and Electricity Production. Rheinbrawn AG.

VON ALBERTI, H-J. and RAMMLER, E. 1962. Technologie und Chemie der

Braunkohleverwertung. er Deutsche Verlag für Grundstoffindustrie,

Leipzig. ◆



Figure 11 – Example of CFD simulations of the HTW Berrenrath; particle (left) and gas temperature distribution in the middle plane (right)

IntroductionA coal seam with a large dip angle (CLDA) is aseam that dips at 35°–55°. CLDAs account forabout 15–20% of China’s coal reserves and5–10% of the output. More than 50% of thecoal seams comprise scarce coal varietiesunder protective mining. CLDA mining ischallenging due to the difficulty in controllingthe stability of the roof, floor, and the coal faceequipment, the difficult operating environmentfor workers, frequent accidents, and lowextraction rates. CLSD refers to a coal facehaving a large underhand angle or largeoverhand angle in the strike direction, as wellas a large dip angle. To ensure safe productionat the coal face, more stringent requirementsfor the stability of the support and otherequipment, as well as roof control, have beenproposed.

The mining of CLDA occurs mainly in theregion of the former Soviet Union, andrelevant reports are also available fromGermany, France, Spain, and India. The study

of CLDA mining includes the mining method,strata control, and equipment development. Inrecent years, the study of CLDA mining hasfocused mainly on surface subsidence and itsprediction. Kulakov (1995a and 1995b) madea systematic study of the rock pressure in acoal face with steep dip (large dip angle).Rafael and Javier (2000) investigatedsubsidence phenomena caused by CLDAmining and established a subsidenceprediction model for CLDA mining.

Chinese scholars have focused on thesupport stability control mechanism in a coalface with large dip angle. Wu (2006, 2005)analysed the varieties of instability of a roof-support-floor (R-S-F) system under diverseconditions, established a R-S-F systemdynamic model, and determined the controlmode for R-S-F system dynamic stability. Linet al. (2004) analysed anti-topple, anti-slip,and skew stability of hydraulic support forfully mechanized caving mining under thecondition of large dip angle based on statics.Combined with field investigations, theystudied three kinds of stability of hydraulicsupport for fully mechanized caving miningwith large dip angles.

However, all the research findings to datethat analyse the support stability start from thepoint of view of the coal face dip angle. Theresearch involving underhand/overhandmining has also focused merely on the charac-teristic analysis of roof-breaking (Zhang et al.,2010; Tian et al., 1994). There are few reportsin the literature on research into supportstability for CLSD.

Support stability mechanism in a coalface with large angles in both strikeand dip by L.Q. Ma*†, Y. Zhang*†, D.S. Zhang*†, X.Q. Cao*†, Q.Q. Li*†, and Y.B. Zhang‡

SynopsisTo solve the support stability control problem for a coal face with largeangles along both strike and dip (CLSD), the ’support-surrounding rock’mechanical model has been developed, which takes into account theimpact of the dip angle of the seam on the stability of the support in thestrike direction. The mechanical relationships of the critical topple angleand critical slip angle of the support along the strike of the coal face withlarge dip angle and the support height, support resistance, frictioncoefficient, and other factors have been derived through the mechanicalanalysis of support stability in the strike direction of CLSDs in the freestate, the operating state, and the special state. The research findings wereapplied to a fully mechanized CLSD in Xinji Coal Mine. The maximumunderhand angle and overhand angle in strike are 42° and 25° respec-tively, and the maximum dip is 39°. It is calculated that during underhandmining and overhand mining, the critical support resistances for avoidingsupport toppling are 3723 kN and 1714 kN respectively, and the criticalsupport resistances for avoiding slipping of the support are 7405 kN and6606 kN respectively. Thus, the selection of type ZZ7600/18/38 hydraulicroof support for the coal face is justified. Measures to prevent sliding ofthe support and the installation of a limiting stop maintain the supportruns in good condition and ensure safe and efficient mining of CLSD.

Keywordscoal mining, support stability, dip angle, strike angle, critical topple,critical slip.

* School of Mines, China University of Mining &Technology, Xuzhou, China.

† Key Laboratory of Deep Coal Resource Mining,Ministry of Education of China, Xuzhou, China.

‡ State Development & Investment Corporation XinjiEnergy Company Limited, Huainan, China.

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. Paper receivedJune 2012 and revised paper received May 2015.



Support stability mechanism in a coal face with large angles in both strike and dip

The current investigation examines the underhand andoverhand mining of CLSD at State Development & InvestmentCorporation’s Xinji Coal Mine, in the Anhui Province ofChina. The impact of the dip angle of the coal seam onsupport stability in the strike direction is considered for thefirst time, based on the actual conditions for underhandmining and overhand mining. The dip angle is introduced asan important parameter in the ‘support-surrounding rock’mechanical model to study the instability of the support inthe strike direction. Mechanical parameters of the support inthe free state, the operating state, and the special state arecalculated for underhand mining and overhand mining. Thefactors that impact on support stability in the strike directionare analysed, and methods are proposed to solve the stabilitycontrol problem in CLSD.

CLSD ‘support-surrounding rock’ mechanical model

Support stability in the strike direction in the freestateSupport stability depends on the interaction between theangle of strike and dip angle of the coal face. The analysis ofdeadweight of the support is shown in Figure 1 (Cao et al.,2010; Ma et al., 2010; Zhang, 2010; Li, 2009; Ostayen et al.,2004).

Underhand mining stageThe mechanical model of the support in the free state isshown in Figure 2.

The topple mechanical model of the support in the freestate is shown in Figure 2(a), and its stress state is given by:

[1]

where

The critical topple angle β1 is:

[2]

where G is the support deadweight (kN), G2 is the componentof the support deadweight perpendicular to the floor (kN), G3is the component force of the gravity of support along the

strike direction of the coal face (kN), α is the dip angle of thecoal face, L is the base length of the support (m), h is thesupport height (m), λ1 is the height coefficient of the gravita-tional centre of the support (the ratio between the height ofgravitational centre y and the support height h), and λ2 is thelength coefficient of the gravitational centre of the support(the ratio between the tail length of the support base awayfrom the gravitational centre x and the base length of thesupport L).

The slip mechanical model of the support in the free stateis shown in Figure 2 (b), and its stress state is analysed inEquation [3]:

[3]

The critical slip angle β2 is:

[4]

where f21 is the frictional resistance provided by the floor tothe support (kN), R21 is the reaction force between the floorand the support (kN), and μ is the frictional coefficientbetween the support and the roof/floor.

Overhand mining stageThe mechanical model of the support in the free state isshown in Figure 3.

The topple mechanical model of the support in the freestate is shown in Figure 3 (a) and its stress state is analysedin Equation [5]:

[5]

The critical topple angle β1 is shown by Equation [6]:

[6]

The slip mechanical model of the support in the free stateis shown in Figure 3 (b), and its stress state is analysed inEquation [7]:

▲


Figure 1 – Component analysis of deadweight of the supportFigure 2 – Mechanical model for support in the free state duringunderhand mining

[7]

The critical slip angle β2 is given by Equation [8]:

[8]

Support stability in the strike direction in theoperating state

Underhand mining stageThe mechanical model of the support in the operating state isshown in Figure 4.

The topple mechanical model of the support in theoperating state is shown in Figure 4 (a), and its stress stateis analysed in Equation [9]:

[9]

where f22 is the frictional resistance provided to the roof bythe support (kN), Le is the distance between the tail of thesupport canopy and the tail of the support base (m), and R22is the reaction force between the roof and the support (kN).

The critical topple angle β1 is given by:

[10]

where M1 = LR22 + R22μh – R22LeThe slip mechanical model of the support in the operating

state is shown in Figure 4 (b) and its stress state is analysedin Equation [11]:

[11]

The critical slip angle β2 is given by:

[12]

Overhand mining stageThe mechanical model of the support in the operating state isshown in Figure 5.

The topple mechanical model of the support in theoperating state is shown in Figure 5 (a) and its stress state isanalysed in Equation [13]:

[13]


[14]



Figure 3 – Mechanical model of support in the free state duringoverhand mining

Figure 4 – Mechanical model of support in the operating state duringunderhand mining

Figure 5 – Mechanical model of support in the operating state duringoverhand mining


The slip mechanical model of the support in the operatingstate is shown in Figure 5 (b) and its stress state is analysedin Equation [15]:

[15]


[16]

Support stability in the strike direction in the specialstateThe stress state of the support in the strike direction for thecoal face in the special state is shown in Figure 6. When roofweighting occurs in the coal face, the support stability issubjected to a larger lateral force, due to roof fracture, whichis generated in the upper part of the support along the strikedirection. The impacts of faulting, roof falls, and other factorson the support are consistent with that of roof weighting inthe coal face in the strike direction.

The caving zoneAfter the coal has been extracted, the roof strata will fail frombottom to top, layer by layer. When a stable geometry isformed in the strata above the caving zone, the lateral forceon the support is mainly from the weight of the rock withinthe caving zone (Qian and Miao, 1995).

The theoretical thickness of strata in the caving zone, hK(m), is shown in Equation [17]:

[17]

where M is the mining height (m), KK is the bulking factor,and α is the dip angle of the coal face.

The thickness of each layer of the immediate roof and themain roof is accumulated from bottom to top to evaluate hK.

When hK is reached or exceeded, the last layer will be the toplayer in the caving zone. The total thickness from the bottomlayer to the top layer is the actual thickness of strata in thecaving zone, as shown in Figure 7 (Dou et al., 2009).

The total weight of the rock in the caving zone P (kN) isgiven by:

[18]

where b is the support width (centre to centre) (m)γz is the average body force of the immediate roof in thecaving zone (kN/m3)hz is the thickness of the immediate roof (m)Lz is the rock canopy length of the immediate roof (m).Lz=Ld+Lh+Lzx (Ld is the tip-to-face distance, which isabout 1.0 m, Lh is the sum of the support canopy andfront canopy lengths (m), Lzx is the maximum hanginglength of the immediate roof behind the support (for themudstone, Lzx is about 1.0 m)n is the number of the main roof layers in the cavingzoneγi is the average body force of the ith layer of the mainroof in the caving zone (kN/m3)hi is the thickness of the ith layer of upper roof in thecaving zone (m)Li is the length of the ith layer of rock in the main roof inthe caving zone.The actual measured data for rock length of each main

roof should be used for the calculation. If there is no actualdata, the rock length of each main roof can be assumed to bethe same as the rock length of the first layer of the main roof,due to few main roof layers collapsed in the actual cavingzone. The rock length of the first layer of the main roof is theaverage periodic weighting interval at the coal face (m).

The lateral force on the support from the rock in thecaving zone F1 (kN) is shown in Equation [19].

[19]

Underhand mining stageThe mechanical model of the support in the special state isshown in Figure 8.

▲


Figure 6 – The stress state of the support in the strike direction for thecoal face in the special state

Figure 7 – Schematic of the caving zone

The topple mechanical model of the support in the specialstate is shown in Figure 8 (a) and its stress state is analysedin Equation [20]:

[20]


[21]

The slip mechanical model of the support in the specialstate is shown in Figure 8 (b) and its stress state is analysedin Equation [22]:

[22]


[23]

Overhand mining stageThe mechanical model of the support in the special state isshown in Figure 9.

The topple mechanical model of the support in the specialstate is shown in Figure 9 (a) and its stress state is analysedin Equation [24]:

[24]


[25]

where

The slip mechanical model of the support in the specialstate is shown in Figure 9 (b) and its stress state is analysedin Equation [26]:

[26]


[27]

Increasing the support resistance, increasing the frictionalcoefficient between the support and the roof/floor, andreducing the support deadweight (while ensuring the supporthas sufficient strength) can be conducive to preventing thesupport from slipping in the strike direction.

Engineering projects

Mining geological conditionsIn Xinji Coal Mine, the E1108 coal face has a length of 877 min the strike direction and 115 m in the dip direction. Thethickness of the seam is 2.2–3.6 m, with an averagethickness of 2.83 m. The conditions in the roof of the seamare shown in Table I. The dip angle of the coal seam is22–39°, with an average of 30°. Because the strike directionof the coal seam at the coal face varies, underhand mining isadopted in the inner segment of the coal face, and overhandmining in the outer segment. The dip angle of the seam in theunderhand mining section is 22–35° and the maximumunderhand mining angle is 42°. The dip angle of the seam inthe overhand mining section is 28–39° and the maximum



Figure 8 – Mechanical model of support in the special state duringunderhand mining

Figure 9 – Mechanical model of support in the special state duringoverhand mining


overhand mining angle is 25°. The layout and the cross-section of the E1108 coal face are shown in Figure 10.ZZ7600/18/38 chock-shield support is utilized for the coalface. At the mining height of 2.8 m, the height coefficient andthe length coefficient of the gravitational centre of the supportare 0.5 and 0.4 respectively. The distance between the tail ofthe support canopy and the tail of the support base is 0.95 m.The main technical parameters of the support are shown inTable II.

Back-analysis of the support selectionThe friction coefficient is considered to be 0.3 (Hu et al.,2008) and technical parameters of the support are put intothe mechanical model in the free state. The critical toppleangle and the critical slip angle of the support for the coalface in the free state can be calculated for the maximum dipangle, which is 35° during underhand mining and 39° duringoverhand mining. The results are shown in Table III.

If the working resistance of the support (7600 kN) andthe maximum dip angle of the coal face are entered intoEquations [10], [12], [14], and [16], neither the criticaltopple angle nor the critical slip angle of the support in theoperating state are reached at the stage of underhand miningor overhand mining.

If the maximum dip angles of 35° during underhandmining and 39° during overhand mining are entered intoEquation [17], the corresponding thicknesses of thetheoretical caving zone are 8.54 m and 9 m. The immediateroof and the first layer of the main roof slice can completelyfill the gob area. Therefore, the highest slices of the cavingzone during underhand mining and overhand mining areboth the first layer of the main roof slice. If the parametersare entered into Equation [18], the total weight of the rock inthe caving zone, P, can be calculated as 5410 kN.

If the working resistance of the support is 7600 kN, and Pand the maximum dip angle of the coal face duringunderhand mining and overhand mining are entered intoEquations [21], [23], [25], and [27], the critical topple angleand the critical slip angle of the support in the special statecan be obtained for underhand mining and overhand mining.The results are shown in Table IV.

The critical support resistance at the maximumunderhand mining angle and maximum overhand mining

angle of the coal face in the special state can be calculatedusing Equations [21], [23], [25], and [27]. The results areshown in Table V.

It can be seen from the results that the working resistanceof the support meets the requirements not only in theoperating state, but also in the special state. During

▲


Figure 10 – Layout of the E1108 coal face

Table I

Characteristics of the roof of the coal seam

Roof Lithology Thickness (m) Body force (kN.m-3) Bulking factor

2nd layer of the main roof Medium sandstone 6.2 25 1.41st layer of the main roof Sandstone 7.5Immediate roof Mudstone 2.0

Table II

Main technical parameters of the support

Model number Nominal working Support height Canopy length Front canopy Base length Centre distance Weight (t)resistance (kN) (mm) (mm) length (mm) (mm) (mm)

ZZ7600/18/38 7600 1800-3800 2767 1740 3050 1500 25

Table III

Support stability in the strike direction in the freestate

Critical topple Critical slip Maximum angle (°) angle (°) dip angle

Underhand mining 42.15 0 35°Overhand mining 14.12 0 39°

Table IV

Support stability in the strike direction in thespecial state

Critical topple Critical slip Maximum angle (°) angle (°) dip angle

Underhand mining - 44.78 35°Overhand mining - 41.57 39°

underhand mining and overhand mining at large angles, thesupport is prone to slip and topple in the free state, and thisis much more likely to occur during overhand mining thanduring underhand mining. Therefore, some auxiliarymeasures are required in order to ensure support stability inthe free state.

Auxiliary measures

Prevention of sliding of the supportIt can be seen from mechanical analysis that the support caneasily slip in the free state. Therefore, measures must beintroduced to increase the frictional force between the supportbase and the floor and between the support canopy and theroof at all times, thus changing the support action from thefree state into the operating state.

Installation of the limiting stopIn order to prevent the sliding of the scraper conveyor andthe swinging of the support during support advance, limiting

stops are installed on the support base (see Figure 11) torestrict the swing range of the push-pull rod and to ensuresupport stability.

Practical effectDuring underhand mining at a large angle, the support has amaximum support resistance of 7489 kN, reaching 98.5% ofthe working resistance; and the average support resistanceduring roof weighting is 4354 kN, and 3175 kN without roofweighting.

During overhand mining at a large angle, the support hasa maximum support resistance of 6535 kN, reaching 86.0%of the working resistance; and the average support resistanceduring roof weighting is 3461 kN, and 2676 kN without roofweighting.

It can be seen that the support resistance meets therequirements for roof control. Assisted by technicalenhancements, including anti-topple and anti-slip measures,the support provides good operating conditions, and ensuresnormal mining of CLSD. The daily average coal cuttingproduction is 2080 t. The operating states of the support forthe coal face during underhand mining are shown in Figure12, and during overhand mining in Figure 13.

DiscussionAccording to the mechanical model, by reducing the height ofthe gravitational centre of the support, reducing the weight ofthe support, increasing the base length of the support, andincreasing the support’s resistance, the anti-topple and anti-slip capacity of the support can be significantly improved.Therefore, in the support design, the established mechanicalmodel can be adjusted to optimize the structure of thesupport and the dimensions of each part of the support, so asto enable it to be best suited for CLSD and to have a stronganti-topple and anti-slip capacity. This mechanical model is



Table V

Critical support resistance of support stability in the strike direction in the special state

Strike angle Critical support resistance

Avoid toppling (kN) Avoid slipping(kN) Maximum dip angle

42° (underhand mining) 3732 7405 35°25° (overhand mining) 1714 6606 39°

Figure 12 – Operating state of the support

Figure 11 – Location of the limiting stop


applicable not only to support selection for Xinji Coal Mine’sE1108 coal face, but also to other CLSD situations.

The established mechanical model described above doesnot cater for some special cases yet. For example, duringunderhand mining at a large angle, the waste within the gobarea may surge onto the support, causing a thrust on thesupport, and increasing the possibility of slipping andtoppling.

The limiting stops installed on the support restrict theswing range of the push-pull rod. As the push-pull rod isrigidly connected, it may break when the sliding impulsiveforce of the scraper conveyor is very large. To avoid this, ahydraulic jack can substitute for the limiting stop, with asafety valve installed to ensure the normal use of the push-pull rod.

Key conclusions➤ The ‘support-surrounding rock’ mechanical model has

been developed considering the impact of the dip angleof the coal seam on the support stability in the strikedirection. The mechanical analysis of the support isdivided into the free state, the operating state, and thespecial state. Various forces have been adopted as theboundary conditions, so as to obtain the critical toppleangle and the critical slip angle of the support indifferent states. The key factors that impact on thesupport stability of the coal face with a large dip anglehave been analysed

➤ The research findings have been applied to the E1108fully mechanized coal face in Xinji Coal Mine, and thecritical support resistances required to ensure that thesupport neither topples nor slips during underhand andoverhand mining have been calculated. It has beenverified that the working resistance of the supportmeets requirements for support in the special state

➤ During underhand mining and overhand mining of theE1108 coal face with a large angle, the resistance of thesupport meets the requirements for roof control and theselection of the coal face support is relativelyreasonable. Assisted by technical enhancements, suchas prevention of sliding of the support and the instal-lation of a limiting stop, the support has achieved goodoperating conditions, and ensured normal mining ofCLSD.

AcknowledgementsWe thank State Development & Investment Corporation XinjiEnergy Company Limited for their assistance. We also thankDr. F.T. Wang and the Fundamental Research Funds for theCentral Universities (2014YC01). This work was supportedby Qing Lan Project, and the Priority Academic ProgramDevelopment of Jiangsu Higher Education Institution.

ReferencesCAO, S.G., XU, J., LEI C.G., PENG, Y., and LIU, H.L. 2010. The stent adaptability of

steep fully mechanized working face under the complex conditions.Journal of China Coal Society, vol. 35, no. 10. pp. 1599–1603 (in Chinese).

DOU, L.M., LU, C.P., and MOU, Z.L. 2009. The stope roof control and monitoringtechnology. China University of Mining and Technology Press, Xuzhou (inChinese).

HU, M. and CAO, B.D. 2008. Analysis for lateral stability of ZY10800/28/63powered support. Coal Mine Machinery, vol. 29, no. 8. pp. 61–63 (inChinese).

KULAKOV, V.N. 1995a. Stress state in the face region of a steep coal bed. Journalof Mining Science, vol. 31, no. 3. pp. 161–168.

KULAKOV, V.N. 1995b. Geomechanical conditions of mining steep coal beds.Journal of Mining Science, vol. 31, no. 2. pp. 136–143.

LI, H.C. 2009. Force analysis of hydraulic support. Coal Mine Machinery, vol.30, no. 7. pp. 69–71 (in Chinese).

LIN, Z.M., CHEN, Z.H., XIE, J.W., and XIE, H.P. 2004. Stability analysis andcontrol measures of powered supports in greater inclined full mechanizedcoal seam. Journal of China Coal Society, vol. 29, no. 3. pp. 264–268 (inChinese).

MA, L.Q., ZHANG, D.S., REN, T.X., ZHANG C.G., and LI, Y.S. 2010. Support designand strata control of coal face with deep dip angle and large miningheight. ICMHPC-2010 International Conference on Mine HazardsPrevention and Control, Qingdao, China, 15–17 October 2010. AtlantisPress, Paris, France. pp. 460–467.

OSTAYEN, R.A.J.V., BEEK, A.V., and ROS, M. 2004. A parametric study of thehydro-support. Tribology International, vol. 37, no. 8. pp. 617–625.

QIAN, M.G. and MIAO, X.X. 1995. Theoretical analysis on the structural fromand stability of overlying strata in longwall mining. Chinese. Journal ofRock Mechanics and Engineering, vol. 14, no. 2. pp. 97–106 (in Chinese).

RAFAEL, R.D. and JAVIER, T.A. 2000. Hypothesis of the multiple subsidencetrough related to very steep and vertical coal seams and its predictionthrough profile functions. Geotechnical and Geological Engineering, vol.18, no. 4. pp. 289–311.

TIAN, Q.Z., LIU, J.C., and ZHANG, Y.Q. 1994. Fracturing characteristics of mainroof strata when mining uphill and downhill and its effect on stability ofimmediate roof. Journal of China Coal Society, vol. 19, no. 2. pp. 140–150(in Chinese).

WU, Y.P. 2006. Keys to dynamic equations of system R-S-F and determinationon working resistance of face support in steeply dipping seam mining.Journal of China Coal Society, vol. 31, no. 6. pp. 736–741 (in Chinese).

WU, Y.P. 2005. Dynamic equation of system ‘roof (R)-support (S)-floor (F)’ insteeply dipping seam mining. Journal of China Coal Society, vol. 30, no. 6.pp. 685–689 (in Chinese).

ZHANG, Y.D., CHENG, J.Y., WANG, X.X., FENG, Z.J., and JI, M. 2010. Thin platemodel analysis on roof break of up-dip or down-dip mining stope. Journalof Mining and Safety Engineering, vol. 27, no. 4. pp. 487–493 (inChinese).

ZHANG, W.D. 2010. Research on the stability of hydraulic support with the largemining height under conditions of the big inclination and the largedepression angle. Mining and Processing Equipment, vol. 38, no. 1. pp.21–24 (in Chinese). ◆

▲


Figure 13 – Operating state of the support in dip direction

IntroductionThe open pit mine design process seeks todefine the optimum pit limits and sequence ofmining, in order to derive the maximumbenefit from the exploitation of a mineralresource given its spatial distribution and theparticular geological, economic, and minesettings. Pit slope angles are determined usingthe conventional approach, whereby slopestability indicators such as the factor of safety(FS) or the probability of failure (PF) arecalculated and compared with generic accept-ability criteria to define the values to be usedin the mine design process. The maindrawback of this approach is that in spite ofthe effect that the slope angle has on theeconomics of the mine plan, its definition isbased on criteria not directly related to thisaspect of the design. The pit slope designprocess described in this paper attempts to

avoid this drawback. The methodology isbased on a quantitative risk evaluation of theslopes, which has as a central element theconstruction of a risk map that relates theprobability of the impact to its magnitude. Inthis process the economic impacts of slopefailure are calculated and used as the elementson which to apply the acceptability criteria fordesign.

The proposed methodology is an evolutionof the approach described by Tapia et al.(2007) and Steffen et al. (2008), where eventtree analysis similar to that used for safety riskevaluations was applied to the economicassessment of slope failures. This approachwas superseded by a probabilistic method witha less subjective basis, as described byContreras and Steffen (2012). The method wasstill in a development phase at the time of thelatter publication, and was due to be applied toactual projects. Since then, the methodologyhas been used to evaluate two open pit mineprojects, and as a result of that work someimprovements have been implemented, partic-ularly in terms of the concepts of probabilityused for the construction of the risk map. Thegraphs and data used in this paper to presentthe methodology are derived from these twoprevious studies.

BackgroundThe optimum design of a pit requires thedetermination of the most economic pit limit,which normally results in steep slope angles asin this way the excavation of waste isminimized. In general, as the slope anglebecomes steeper, the stripping ratio (waste toore ratio) is reduced and the mining economicsimprove. However, these benefits arecounteracted by an increased risk to the

An economic risk evaluation approachfor pit slope optimizationby L.F. Contreras*

SynopsisIn open pit mine design, it is customary for geotechnical engineers todefine the appropriate slope design angles within practical limits. Theconventional approach to slope angle design is based on the comparison ofcalculated stability indicators, such as the factor of safety (FS) and theprobability of failure (PF), with generic acceptability criteria not directlyrelated to the impacts of failure. A major drawback of this type ofapproach is related to the difficulty of defining meaningful acceptabilitycriteria. An alternative methodology of pit slope design is proposed, wherethe economic impacts of potential slope failures are calculated and used asthe elements on which to apply the acceptability criteria for design. Themethodology is based on the construction of a graph, referred to as a riskmap, that relates the probability of exceeding the economic impact of slopefailure to the magnitude of the impact measured in monetary terms. Theprocess includes the analysis of a selected number of representative yearsof the mine plan and slope sections of the pit areas to define the requiredinputs for the construction of the risk map. The paper discusses theconcepts used in interpreting the probability of slope failure, and describesthe approach followed for the estimation of the economic impacts of slopefailure and the construction of the risk map. Finally, the two main uses ofthe risk map are discussed, including the comparison with acceptabilitycriteria for the evaluation of a specific open pit design and the comparativeanalysis of open pit design options in terms of value and risk to identifyoptimum pit layouts.

Keywordsrisk evaluation, economic risk map, slope design, slope failure, probabilityof failure.

* SRK Consulting, Johannesburg, South Africa.© The Southern African Institute of Mining and

Metallurgy, 2015. ISSN 2225-6253. Paper receivedJul. 2014 and revised paper received Jan. 2015.



An economic risk evaluation approach for pit slope optimization

operation. Thus, the determination of the acceptable slopeangle is a key aspect of the mining business.

The difficulty in determining the acceptable slope anglestems from the uncertainties associated with slope stability.Typical uncertainties encountered in the pit slope designprocess are discussed by Tapia et al. (2007) with reference tothe Chuquicamata open pit. There are three main approachescommonly used to account for the uncertainties in slopedesign: factor of safety, probability of failure, and riskanalysis.

Factor of safety approachThe oldest approach to slope design is based on thecalculation of the factor of safety (FS). The FS can be definedas the ratio between the resisting forces (strength) and thedriving forces (loading) along a potential failure surface. Ifthe FS has a value of unity, the slope is said to be in a limitequilibrium condition, whereas values larger than unitycorrespond to stable slopes. The FS approach is adeterministic design technique as a point estimate of eachvariable is assumed to represent the variable with certainty.The uncertainties implicit in the stability evaluation areaccounted for through the use of a FS for design larger thanunity. This acceptability criterion is intended to ensure thatthe slope will be stable enough to ensure a safe miningoperation. Acceptable FS values in mining applications rangebetween 1.2 and 2.0 according to Priest and Brown (1983),as indicated in Wesseloo and Read (2009). Acceptable valuesare based on observations of the performance of slopes atspecific sites and experience accumulated over time.

There are two main disadvantages in the FS approach forslope design. Firstly, the acceptability criterion is based on alimited number of cases and combines the effect of manyfactors that make it difficult to judge its applicability in aspecific geomechanical environment. Secondly, the FS doesnot provide a linear scale of the likelihood of slope failure.

Probability of failure approachIn recent years, probabilistic methods have been increasinglyused in slope design. These methods are based on thecalculation of the probability of failure (PF) of the slope. Aprobabilistic approach requires that a deterministic modelexists. In this case the input parameters are described asprobability distributions rather than point estimates of thevalues. By combining these distributions within thedeterministic model used to calculate the FS, the probabilityof failure of the slope can be estimated. A techniquecommonly used to combine the distributions is the MonteCarlo simulation. In this case, each input parameter value issampled randomly from its distribution, and for each set ofrandom input values a FS is calculated. By repeating thisprocess many times, a distribution of the FS is obtained. ThePF can be calculated as the ratio between the number of casesthat represent failure (FS<1) and the total number ofsimulations.

The advantage of the PF over the FS as a stabilityindicator is based on the fact that there is a linearrelationship between the PF value and the likelihood offailure1, whereas the same is not true for the FS. A larger FSdoes not necessarily represent a safer slope, as the magnitudeof the implicit uncertainties is not captured by the FS value. Aslope with a FS of 3 is not twice as stable as one with a FS of

1.5, whereas a slope with a PF of 5% is twice as stable as onewith a PF of 10%.

Some drawbacks of the FS methodology that persist inthe PF approach are the difficulties in defining an adequateacceptability criterion for design and the limitations inpredicting failure with the underlying deterministic model.

Acceptability criteria for PF have been defined bydifferent authors and organizations, and a summary of thisinformation is presented in Wesseloo and Read (2009).However, the actual criteria to be used in a specific minecannot be determined from general guidelines like these, andshould be subjected to a more thorough analysis of theconsequences of failure (Sjoberg, 1999).

Risk analysis approachThe risk analysis approach tries to solve the main drawbackof the previous methodologies with regard to the selection ofthe appropriate acceptability criteria. Risk can be defined asthe probability of occurrence of an event combined with theconsequence or potential loss associated with that event:

Risk = P(event) × Consequence of the event

In the case of slopes, the P(event) is the PF of the slope andthe consequences can be two-fold: personnel impact andeconomic impact.

The PF calculated as part of the design process isnormally based on a slope stability model calculation andaccounts only for part of the uncertainties of the slope.Because risk analysis sets the acceptability criteria on theconsequences rather than on the likelihood of the event, athorough evaluation of the PF of the slope is required,incorporating other sources of uncertainty not accounted forwith the slope stability model. For this purpose and for theanalysis of consequences of slope failure, non-formal sourcesof information (engineering judgment, expert knowledge) areincorporated into the process with the aid of methods such asdevelopment of logic diagrams and event tree analysis. Thesetechniques are described in detail by Baecher and Christian(2003) with reference to geotechnical engineering problems,more commonly in the disciplines of dam and foundationengineering. However, the use of risk methods in open pitmining focuses on safety applications, based on qualitativeapproaches to assess operational aspects.

In the following sections, a description of the proposedrisk methodology for slope design optimization is presented.

MethodologyThe proposed methodology uses the framework described bythe Australian Geomechanics Society (2000) with referenceto the landslide risk management process, characterized bythe following main steps:

➤ Identify the event generating hazards

▲


1 In non-technical literature, ‘likelihood’ is usually a synonym for’probability’, but in statistical usage, a clear technical distinction ismade. Here, probability of failure refers to the estimated frequencyof FS<1 cases with the model assuming that this conditionsrepresents failure. PF can take values only between 0 and 1. Thelikelihood of failure is a quantity not constrained and refers to thechances of actual failure, given the results of the stability analysis.

➤ Assess the likelihood or probability of occurrence ofthese events

➤ Assess the impact of the hazard➤ Combine the probability and impact to calculate the risk➤ Compare the calculated risk with benchmark criteria to

produce an assessment of risk➤ Use the assessment of risk as an aid to decision-making.

The methodology described in this paper refers mainly tosteps 2 to 5 as applied to the risk evaluation of pit slopes.

The proposed risk evaluation process for slope design isintended to quantify the impact of potential slope failures onthe economic performance of open pit mines. Figure 1illustrates the risk evaluation process and depicts the mainelements of the methodology, which are described in detail inthis paper. The diagram includes the main components of theconventional geotechnical slope design process as describedin Stacey (2009) and incorporates the additional elementsrequired from the mine design process.

The main objective of the methodology is the definition ofthe pit slope angles for mine design by applying projectspecific criteria to the quantified risk costs. The approachincludes the following main steps:

➤ Definition of the set of slope sections for analysiscovering key and critical pit areas during the mine lifeto provide representative cases of potential risks ofslope failure within the mine plan

➤ Calculation of the probability of failure (PF) of theslopes from the analysis of stability of the selectedslope sections

➤ Quantification of the economic impacts of slope failurewith reference to the loss of annual profit or totalproject value as measured by the NPV

➤ Integration of the results of probability of failure andeconomic impact on an annual basis to define theeconomic risk map per year and for the life of mine

➤ Comparison of the risk map with criteria to assessacceptability of the design and to define risk mitigationoptions as required

➤ If the analysis is intended for the comparison ofalternative slope design options, the process is repeatedfor each alternative pit layout and the results arecollated in a graph of slope angle versus value and riskcost where the optimum slope angles can be defined.

A complete risk evaluation process should also includethe evaluation of the safety impact of slope failures. Safetyrisk evaluation is discussed by Contreras et al. (2006),Terbrugge et al. (2006), Tapia et al. (2007), and Steffen etal. (2008), and is not covered in this paper.

Slope sections for analysisThe risk evaluation process requires a programme of slopestability analyses, including the critical pit areas and years interms of potential economic impacts of eventual slopefailures. This means that besides adequate information ongeotechnical conditions defining the likelihood of failures, agood understanding of the mine plan is required to identifythose areas and years in which the impacts of failure arelikely to be greater.

The selection of the sections for stability analysis startswith the selection of the years of the mine life that representdevelopment periods in the mine plan with similar character-istics in terms of pit geometry, production profile, andeconomic scenario. Figure 2 shows an example of thecumulative discounted profit of a mine plan, which is arepresentation of the realization of value with time. Thisgraph facilitates the definition of the appropriate periods andrepresentative years of mine development for the risk modelanalysis, which in this example corresponds to the six yearsdefining the stepped curve.



Figure 1 – Risk-based slope design approach


In general, probabilities of failure increase through themine life, whereas impacts tend to maintain their levels oreven decrease as mining progresses. The assumption thatrisk conditions of a later year (2027) represent those of earlyyears (2025/2026) is therefore reasonable, with a minor

effect on the results or (more commonly) on the conservativeside. The graph in Figure 2 implies that there is a trade-offbetween rigour and practicality when selecting the years foranalysis. Ideally, every year would have to be analysed,although this would not be practical and is probablyunnecessary in the majority of cases.

The appropriate slope sections for analysis can be selectedby examination of the mine plan in the identified key years.The criterion used for this selection is based on covering theanticipated higher risk areas of the mine, which includelocations where the likelihood of slope failure or theassociated impact is expected to be high. Examples of thepreferred locations for analysis include areas with higher orsteeper slopes, sites with unfavourable geological conditions,areas with distinct characteristics such as those defined bythe geotechnical domains, critical access points to miningfaces, areas close to key infrastructure, and so forth. The pitdevelopment plan sketched in Figure 3 shows an examplewith the selection of 42 sections used in this paper toillustrate the risk process.

Slope stability analysisThe results of the slope stability analyses are reported interms of PF values, which are calculated with the appropriate

▲


Figure 3 – Example of selection of slope sections for risk analysis

Figure 2 – Realization of value with time as a criterion for defining yearsof risk model analysis

slope stability models in accordance with the relevant failuremechanisms in each domain. The methodology used for thecalculation of the PF is in part determined by the type ofdeterministic model used for the calculation of the FS of theslope. A compilation of the methods commonly used in slopedesign can be found in Lorig et al. (2009). In a probabilisticstability analysis, the input parameters that represent theuncertainties are described by probability distributions. Thesedistributions are combined within the deterministic model todefine the distribution of the FS, which is used to estimatethe PF of the slope. The PF is calculated as the ratio betweenthe number of cases representing failure (FS<1) and the totalnumber of cases of FS described by the distribution. Simplemodels, such as those based on the limit equilibrium method,can incorporate built-in routines to perform Monte Carlosimulations that enable the PF to be calculated relativelyquickly. However, the use of more elaborate models based onstress-deformation analysis, with higher computationaldemands, restricts the calculation of the PF to those methodsrequiring a reduced number of FS entries to define itsvariability. Examples of such methods include those based onTaylor series expansions, the point estimate method, and theresponse surface methodology. Descriptions of these methodsin terms of their conceptual basis are given by Baecher andChristian (2003) and Morgan and Henrion (1990). Theresponse surface method has been used in risk-based slopedesign applications as described by Steffen et al. (2008). Thisapproach has the advantage of combining the rigour of aMonte Carlo simulation with the practicality of requiringfewer FS calculations with the geotechnical model toconstruct the response surface used as a surrogate model inthe process.

Due to practical limitations, the PF values calculated withslope models are typically the result of considering theuncertainty of the strength properties of rock masses andstructures, without consideration of any other potentialfactors contributing to slope instability. Therefore, these PFvalues are incomplete representations of the likelihood offailure, and need to be adjusted as discussed later for thepurpose of a risk consequence analysis.

Interpretation of probability of failure (PF) of the slopeA slope failure event could be regarded as a Bernoulli trial(also called binomial trial), which is defined as a randomexperiment with only two possible outcomes, success orfailure, and in which the probability of success (or failure) isthe same every time the experiment is conducted. Accordingto this definition, and considering failure as the target eventof analysis, if p is defined as the probability of failure, then q= (1-p) corresponds to the probability of no failure. Examplesof Bernoulli trials include a ‘head’ after tossing a coin(p=50%, q=50%), a ‘one’ after rolling a dice (p=16.7%,q=83.3%) and, under certain assumptions as explainedbelow, a failure after excavating a slope (p=PF, q=1-PF). Thesuccessive repetition of Bernoulli trials constitutes a Bernoulliprocess. The probability of success (or failure) is revealed ina Bernoulli process with a large number of trials. It ispossible to verify that after rolling the dice a hundred times,the number of ‘one’ cases will be close to 17 and as moretrials are considered, the better the approximation will be tothe ‘one in six’ probability of getting a ‘one’.

Strictly speaking, a Bernoulli trial refers to a discreteindependent event, which is not exactly the case of thecontinuous process in time or space that characterizes theexcavation of a pit slope. However, the consideration of theslope excavation process as a series of discrete situations, forexample, excavation of consecutive slope lengths along a pitwall or annual exposure of slopes through the mine life, is avalid assumption within the framework of the risk model forslope failure, as failure events are associated with specificslope sections that are selected precisely to represent distinctconditions in terms of time of exposure and location withinthe pit.

The association of open pit slope failure events with aBernoulli process enables the following interpretations basedon the number of trials of the process.

Bench slope failure in a homogeneous domainA bench slope failure in an open pit situation could be seenas a Bernoulli process involving many trials. The probabilityof bench failure in a benched slope within a homogeneousstructural domain corresponds approximately to the ratiobetween the cumulative length of failed benches and the totallength of constructed benches in that domain. In this case,the entire slope could be considered as a series of consecutiverealizations of a unitary slope with a length given by thetypical failure width. This case is illustrated in the sketch inFigure 4 and is comparable with the situation of rolling a dicemany times to verify the probability of getting a ‘one’. In fact,the bench slope case can be seen as if a bench of length ‘b’ isconstructed many times, with a percentage of thosecorresponding with failure situations.

Inter-ramp slope failure in a homogeneous domainThe case of a hangingwall in an open pit mine located withina homogeneous geotechnical domain could be looselyassociated with a Bernoulli process with several trials. In thiscase the probability of failure of the inter-ramp slopes for thelife of mine could be approximated by the ratio between thecumulative volume of inter-ramp slope failures having



Figure 4 – Interpretation of the bench slope failure case as a Bernoulliprocess with many trials


occurred and the total volume of rock excavated during thelife of mine in the hangingwall, as illustrated in Figure 5.

Overall slope failure in a heterogeneous domainThe case of overall slopes in open pits in heterogeneousgeotechnical domains could be associated with a Bernoulliprocess with few trials or even with a single Bernoulli trial.The probability of failure of these slopes is not revealed in aphysical manner and the estimation can be based only onsimulation trials with geomechanical models representing theslopes. In this case the slope could be seen as a uniquerealization or trial that is not repeated in time or space,similar to the situation of a dice rolled once with two possibleoutcomes in terms of getting a ‘one’, success or failure. Theoverall slope failure case as a Bernoulli trial is illustrated inFigure 6.

Estimation of PF values for risk analysisThe PF values to be used in a risk evaluation process need torepresent all the exposed areas of the pit in the year ofanalysis, and to account for all possible uncertain factors thatmay lead to slope failures. The PF values calculated withslope stability models refer to specific sections of the slopesand typically account only for the uncertainties associatedwith variability of geotechnical properties. Therefore, theselimitations need to be accounted for in the set of PF valuesresulting from the geotechnical analysis, such that they aretruly representative of the likelihood of failures in the pitareas and mine plan years of analysis. For this purpose, twotypes of adjustments are required to the PF values calculatedwith the geotechnical models: one related to the estimation ofthe PF of the pit wall as opposed to that of the section ofanalysis; and the other to the estimation of the total PF asopposed to the model PF.

Section and slope wall PFFigure 7 shows the difference between the PF resulting froma stability analysis with a representative section of the slopeand the PF value reflecting the likelihood of slope failure in apit wall with a length greater than the expected width of thefailure.

It is clear that the PF of the longer slope wall in Figure 7is greater than that of the shorter slope shown. Consideringthe shorter slope as a unitary slope with a length comparableto the expected width (d) of the failure, then the longer wallwith length (L) could be seen as a series of consecutiverealizations of the unitary slope (Bernoulli trials). If the PF ofthe shorter slope is given by the probability of failure (ps)resulting from the analysis of a typical section of the slope,then the PF of the longer wall (PFw) can be estimated withthe following expression:

PFw ≈ 1 – (1 – ps)L/d [1]

This consideration is useful to ensure that the possibilityof failure of every exposed slope in the pit is included in the

▲


Figure 6 – Interpretation of the overall slope failure case as a Bernoullitrial

Figure 5 – Interpretation of the inter-ramp slope failure case as aBernoulli process with several trials

Figure 7 – Interpretation of the probability of failure of a slope wall in ahomogeneous rock mass as a function of its length

risk analysis. However, the applicability of this adjustment isrestricted to those situations where the assumption ofhomogeneity of the wall represented by the section analysedis reasonable; otherwise, the analysis of additional sectionsneeds to be implemented.

This consideration is also important in conventionalgeotechnical design procedures where PF values fromgeotechnical analysis are compared with acceptability criteriaof reference, since complying with the specified criteria on asection basis does not necessarily guarantee that the criteriaare met for the slope wall.

Model and total PFThe analysis of consequences of slope failure requires thatthe PF of the slopes be a true reflection of the likelihood ofoccurrence; therefore, all possible situations leading to slopefailures need to be included in the analysis. Due to practicallimitations, the PF calculated with the geotechnical slopemodel typically accounts only for the uncertainty of thematerial properties, and is referred to as the model probabilityof failure (PFmodel) in the following discussion.

The estimation of the PF incorporating other sources ofuncertainty not accounted for with slope models wasdiscussed by Contreras et al. (2006) and by Steffen et al.(2008) using a methodology based on the analysis of sourceresponse diagrams (SRDs). The methodology is based onconcepts presented by Chapman and Ward (2003) withreference to project risk management processes used in awide range of industries. The method enabled the quantifi-cation of the contributions to the PF caused by departuresfrom the normal conditions assumed for the design of theslopes. These variations were evaluated within variouscategories such as groundwater conditions, geologicalfeatures, operational factors, or occurrence of seismic events.The estimated contributions were added to the PF valueresulting from assuming normal conditions of design tocalculate the total probability of failure (PFtotal) to be used ina risk analysis.

The methodology presented in this paper is analogous tothe SRD approach described by Steffen et al. (2008), butadds some considerations regarding time in order to reflectthe gradual increase, with time, of exposure to the atypicalconditions evaluated. The method is appropriate for theassessment of types of uncertainties characterized by analeatory nature. Other uncertainties not associated withfrequency of events would be better treated with an expertopinion approach, with a greater reliance on experience andintuition.

There are two main types of uncertainty in geotechnicalengineering – aleatory and epistemic. The former is due tothe random variation of the aspect under analysis, and thelatter to the lack of knowledge of the aspect. Uncertainties arequantified with probabilities, which in turn can be interpretedas frequencies in series of similar trials or as degrees ofbelief. Baecher and Christian (2003) provide a detaileddiscussion on the topic of this duality in the interpretation ofuncertainty and probability in geotechnical engineering,indicating that both types of probabilities are present in riskand reliability analysis and pointing out that the separationbetween them is a modelling artifact rather than animmutable property of nature. Some aspects of geotechnicalengineering can be treated as random entities represented by

relative frequencies, and others may correspond to uniqueunknown events better treated as a degree of beliefrepresented by expert opinion.

Subjectivity associated with probability estimates is a wayof capturing and integrating expert judgment, only some ofwhich may be based on hard data, and is what formalmodeling of uncertainty and risk is about. Analysis, whichmust be based on hard data, is inherently partial and weak.The topic of subjectivity and expert opinion as a key elementof risk and reliability analysis in geotechnical engineering isdiscussed in detail by Vick (2002) and by Baecher andChristian (2003).

The atypical conditions treated with this methodology areanalysed on an annual basis, therefore each year they eitheroccur or do not, and their annual occurrence is determined bythe same underlying probability derived from a common setof conditions judged for the life of mine, either from harddata or from expert opinion or from a combination of both.These conditions suit those of a Bernoulli process andsupport the gradual increase of likelihood of occurrence withtime estimated with the approach.

Given the probability of occurrence of a particularuncertain atypical situation leading to slope failure (Patypical)associated with a defined mine life duration in years (n), theannual probability of occurrence of this situation (patypical)can be calculated with the following expression:

patypical = 1 – (1 – Patypical)1/n [2]

The probability of failure of the slope, given that theatypical conditions occur (PFmodel│atypical), could be evaluatedwith the slope stability model. The results of such analysiscould be expressed as a factor (fatypical) of the modelprobability of failure evaluated under normal conditions. Thisfactor could be the result of sensitivity analysis wheredifferent scenarios of the atypical condition are evaluated.Therefore:

PFmodel│atypical = PFmodel × fatypical [3]

Finally, the probability of failure of the slope due toatypical conditions (PFatypical) can be calculated for aparticular year (i) of the mine plan as follows:

(PFatypical)i = PFmodel│atypical × (1 – (1 – patypical)i ) [4]

The probability of failure of the slope due to atypicalconditions (PFatypical) is added to the model probability offailure (PFmodel) from the geotechnical analysis under normalconditions of design to define the total probability of failure(PFtotal) appropriate for the risk evaluation process. Theaddition of the probability values is carried out with thefollowing generic expression, which is based on the conceptof system reliability:

PFtotal = 1 - (1 – PFmodel) × (1 – PFatypical) [5]

The method of calculation of PFtotal from PFmodel isillustrated with an example where the contributions fromuncertainties related to groundwater, geology, and mining areadded to the PF calculated with the geotechnical model for theslope represented by Section 3 of the mine case shown inFigure 3. Equation [5] can be extended to account for thesethree aspects as follows:

PFtotal = 1 - (1 – PFmodel) × (1 – PFgroundwater) � (1 –PFgeology) × (1 – PFmining) [6]




The results of this analysis are indicated in Table I andFigure 8, and the input probabilities and factors are indicatedin the footnotes to the table. Equations [2], [3], and [4] areused to calculate the terms in Equation [6]. The calculatedcontributions of each uncertain aspect to the PFtotal are shownby the curves in Figure 8.

The uncertainties considered in the example of Table Iand Figure 8 are intended to present the concept of addinguncertainties of random character not included in thegeotechnical models for slope analysis. However, the relevantuncertainties not included in the models need to be identifiedand assessed on a project-specific basis. It may be thatfactors such as unknown stress conditions, actual pitgeometry variations, or other specific situations are the morerelevant aspects that would contribute to the overall PF in agiven project. Also, the best way to treat a particularuncertainty needs to be defined based on its prevalent nature(i.e. aleatory or epistemic).

In the slope stability evaluation process, the considerationof the potential effect of atypical situations leading to failuremeans that no matter how stable a slope might appear interms of the calculated stability indicators, the probability offailure for the risk analysis is never zero and therefore therisk of failure is always present.

Model uncertaintyModel uncertainty in the slope stability analysis can beevaluated through the critical FS value (FScritical) used todefine failure with the model. This type of uncertainty arisesthrough systematic biases in input parameter determinationsand idealizations in the calculation process, leading to theresult that failure occurs for some FScritical value that may notbe unity, as commonly assumed. Bias in parameter determi-nation is inevitable, and is handled by calibration to slopeperformance. Model idealizations arise from simplificationsrequired to represent the geometry, material behaviour, etc.Some aspects of model idealizations will tend to reduceFScritical, while others might raise FScritical. The effect of theparameter bias and model uncertainty is to produce anuncertainty band that is centred on the underlying bias. Anevaluation of FScritical based on the comparison of actuarialfailure rates versus nominal factor of safety was carried out

for the risk study of the Chuquicamata pit as described byTapia et al. (2009). Unfortunately, this approach requireslocal historic records, which are not always available;therefore, judgement as well as reference to similar projects isthe only practical option left to account for this uncertainty.

Estimation of economic impact of slope failure eventsThe economic impact of a slope failure can be measuredthrough the quantification of the effect of this event on thevalue of the mine plan as measured by the NPV. The NPVcorresponds to the cumulative discounted annual profitsduring the life of the mine and is normally defined as theresult of a mining scheduling and optimization processcarried out with specialized software. In general, theeconomic impact of a slope failure is a result of the disruptionof the planned ore feed during the time required to restore thesite, and the additional costs caused by these activities.Figure 9 illustrates the conceptual basis for the estimation ofimpacts of slope failures. The economic impact of a slopefailure is defined as the difference between the NPV ofreference (mine plan without failures) and the re-calculatedNPV incorporating the effects of the failure on production andcost components.

▲


Table I

Example of estimation of PFtotal

Year Mine plan 2 4 6 8 11 14

Year 2015 2017 2019 2021 2024 2027

Section 3 PFmodel 1.0% 1.3% 1.8% 2.3% 2.8% 3.0%PFgroundwater 0.0% 0.1% 0.2% 0.4% 0.6% 0.8%PFgeology 0.1% 0.3% 0.6% 1.0% 1.6% 2.1%PF mining 0.0% 0.0% 0.1% 0.1% 0.2% 0.3%PFtotal 1.2% 1.7% 2.6% 3.7% 5.1% 6.1%

Notes:Input data on uncertainties:Groundwater: P 10% in 15 years (p annual = 0.70%)

fgroundwater = PFmodel⏐ groundwater/PFmodel = 3Geology: P 15% in 15 years (p annual = 1.08%)

fgeology = PFmodel⏐ geology/PFmodel = 5Mining: P 5% in 15 years (p annual = 0.34%)

fmining= PFmodel⏐ mining/PFMODEL = 2

Figure 8 – Calculated contributions to the PFtotal due to uncertainatypical conditions not included in PFmodel

Production may be disrupted by different factors such asinterrupted access to the mining faces, covered ore, variationsof grade when alternative sources of ore are used to mitigatethe effects of the failure, and so forth. The additional costsare caused by the additional material handling and re-scheduling of equipment required to restore the site affectedby the failure.

A simplified approach to quantifying the impact of afailure consists of calculating the differential NPV due to thefailure, using a cash flow model that includes the estimatedeffects of the failure on production and costs. The impact onproduction is simulated by means of a reduction factor of themined tons, which is estimated by considering aspects suchas the magnitude, location, and time of occurrence of thefailure and the flexibility of the mine plan to providealternative ore feed sources. Engineering judgment andsupporting reference data are normally used to estimate theimpact factors from each failure event.

The simplified cash flow model should include productiondata per mining phase, revenue calculations, as well asoperating and capital costs, and needs to be calibratedagainst the reference NPV in the mine plan. An example ofthe structure of the simplified cash flow model used for thecalculation of economic impact of slope failures is shown inFigure 10. The example illustrated shows that the impact onproduction affects the plant product tons and the revenue,which, together with the additional costs of restoring the site,ultimately reduces the net benefit and consequently the NPV.

One drawback of the simplified approach is that thecomplex effect of variations of the planned grade feed whendrawing from stockpiles cannot be simulated accurately. Forthis reason, the calculated impacts need to be validated withresults derived from a thorough evaluation of selected key



Figure 9 – Conceptual basis for estimation of the economic impact ofslope failure

Figure 10 – Structure of simplified cash flow model for slope failure impact assessment


events in a similar manner as they would be evaluated in areal-life situation, where specific re-designs of the plan wouldbe carried out to minimize the impact of the slope failure.

Risk map for economic impact analysis of slopefailureThe results of probability of failure and economic impactcalculations for individual failure events are used to constructthe economic risk map per year and for the mine life. The riskmap defines the relationship between the probability of aparticular economic impact and the magnitude of that impact;and accounts for different situations of occurrence of eventsin a year, including isolated occurrences, concurrentoccurrences of the different possible combinations of theevents, and no occurrence of any event.

The risk map construction process is based on the conceptof event tree analysis. The event tree is a diagram thatconnects a starting event with the ultimate consequenceunder evaluation through a series of intermediate eventsbased on a cause-effect relation. The events are quantified interms of their likelihood of occurrence, thus enabling theassessment of the final outcomes in terms of theirprobabilities of occurrence. The event tree methodology foreconomic impact, originally described by Tapia et al. (2007)with reference to the case of the Chuquicamata mine and laterdiscussed by Steffen et al. (2008) and in Wesseloo and Read(2009), relies on subjective inputs of probability for theevents in the tree to produce an assessment of the expectedlikelihood of three categories of economic impact (forcemajeure, loss of profit, and minor impact). The maindrawbacks of this methodology are that there is no consid-eration of the possible occurrence of various events in a yearand that the impacts are assessed only in terms of likelihood,without a clear definition of the magnitude of these impacts.

The combined analysis of probability and economicimpacts with event trees is discussed in detail by Baecher andChristian (2003), including examples of consequenceanalysis where the probabilities of events and the respectiveimpacts in monetary terms are multiplied to produce risk costvalues used as a measure of the risks. One drawback of thisapproach is that the outcomes of the analysis do notrepresent actual possible impacts, but rather amountsweighted by the respective probabilities. This characteristic ofthe risk calculation is referred to by Baecher and Chirstian(2003) as ‘risk neutrality’, where high-probability low-consequence outcomes are treated as equivalent to low-probability high-consequence outcomes, as long as theproduct is the same. The reality is that the events either do ordo not occur and consequently the impacts will either becaused or not – intermediate results are not possible.

The proposed risk evaluation approach is carried out witha separate accounting for probabilities and impacts and theend results from the event tree branches are used to constructthe risk map. The method is illustrated in Figure 11 for thesimple case of a pit with two major slopes named East andWest, with PF values of 5% and 10% and impacts of 100 and50, respectively. The sum of the probabilities of the fourpossible outcomes depicted with the tree is 100%, indicatingthat all the possible combinations of events have beenadequately accounted. The risk map constructed with theresults of the event tree is shown in Figure 12. The

cumulative probability curve of particular impacts constitutesthe economic risk envelope of the pit.

The risk map of a more realistic case, such as the mineplan described in Figures 2 and 3, is constructed for theindividual key years selected to represent the various periodsof the mine plan, which are then used to define the overallrisk map for the life of mine, as shown in Figure 13. Thegraph at the top shows the various risk envelopes and thegraph at the bottom shows the details of the failure events ofyear 2019 used to construct the envelope. The risk envelopesare cumulative probability distributions of impacts and areinterpreted as indicated in the graph at the top of Figure 13for the case of impacts with a 10% probability of exceedance.The result for the indicated case would be a 20% probabilityof having an annual impact of at least $160 million over aperiod of 15 years. The display of the individual events in therisk chart is useful to identify critical events causing anincrease of the risk level as measured by the envelope, asdepicted in the example shown on the graph at the bottom ofFigure 13.

Probability concepts for construction of risk mapThe slope failure events considered for the construction ofrisk maps correspond to large-scale failures and are analysed

▲


Figure 12 – Risk map from results of event tree analysis in Figure 11

Figure 11 – Event tree for economic impact of slope failure of pit withtwo major slopes

on a year-by-year basis. The events are treated as Bernoullitrials and are characterized by a probability of occurrence (p)given by the calculated probability of failure of the slope (PF)and the respective impact (i) estimated in monetary terms.The risk map construction is based on the calculation of theprobability (P) of having an economic impact (I) consideringdifferent possible situations of occurrence of the events asexplained below.

In the following expressions, the terms with sub-indices i,j, and k (in bold) represent the occurring events, and thosewith sub-indices r, s, and t (in italic), refer to the non-occurring events:

➤ Occurrence of single events:Pi = pi × (1 - pr) × … × (1 – pt)Ii = ii

➤ Multiple occurrence of events:Pi…k = pi × … × pk × (1 - pr) × … × (1 – pt)Ii…k = ii + … + ik

➤ A particular case of the multiple occurrence describedin (2) is the occurrence of all the events in a year:

Pi…k = pi × pj × … × pkIi…k = ii + ij + … + ik

➤ No occurrence of any of the events:Pr…t = (1 – pr) × (1 – ps) × … × (1 – pt)Ir…t = 0

The total number of possible cases of occurrence ofevents (T) for (n) independent events in a year effectivelycorresponds to the number of branches of the respective

event tree, and is given by the following expression:

T = 2n [7]

From this number, n cases correspond with theoccurrence of isolated events and one case to the non-occurrence of any of the events. The remaining N casescorrespond with the occurrence of combinations of two ormore events. The generic expression to calculate the numberof combinations (N) of 2 or more events that can be obtainedwith (n) events is:

[8]

or

N = 2n – (n + 1) [9]

The calculation of all possible probability and impactpairs can be done without constructing the respective eventtree, which would be a cumbersome task as the number ofbranches of the tree increases exponentially with the numberof annual events. A summary of the probabilities and impactsof the different possible combinations of 7 events per year ispresented in Table II. In this table, p corresponds with theprobability of occurrence (failure) and q with the probabilityof no occurrence (no failure) of the respective events. Thenumber of cases in Table II is calculated with Equation [8]and the total number of possible occurrences of the 7 eventsis 128. This is the number of data points available toconstruct the risk map as described in the following section.

Construction of the risk mapAn example of the input data required for the construction ofthe risk map is presented in Table III. The data includes theprobability of slope failure and the associated impact of sevensections per year and six years of analysis, on the mine planof 15 years’ duration, as described in Figures 2 and 3. The PFvalues in Table III are based on the results of the geotechnicalanalysis of the respective sections and cater for the atypicalconditions leading to failure discussed previously.

The data in Table III is shown in graphic form in Figure14 to illustrate the variations of the probability of failure andassociated impacts with pit development. The graph at the leftof Figure 14 is consistent with the increasing likelihood offailure of the slopes expected as the pit grows deeper. Thecurves in the graph at the right of Figure 14 do not show aunique trend in the variation of impact with pit growth, asimpacts are dependent on the particular characteristics of oreexposure and ore access during the development of themining phases.

The risk map construction is carried out per year and thedata is used to calculate the pairs of values of probability andimpact associated with all possible combinations of failureevents using the expressions in Table II. The 128 data pairsfor each year of analysis are sorted and used to construct therespective probability distribution graphs of impacts. Thesegraphs include a frequency distribution histogram and thecorresponding cumulative frequency curve as shown in thegraph at the top of Figure 15 for the year 2019 of theexample in Table III.

The risk map result is shown in the graph at the bottomof Figure 15. The graph contains the probability distributionplots with the axes swapped to conform with the typical way



Figure 13 – Example of a risk map for economic impact of slope failure,showing risk envelopes for key years and for life of mine (top), anddetails of events shaping the risk envelope for year 2019 (bottom)


in which risk acceptability criteria is presented, as discussedin the following section. The graph also includes the datapoints representing the various possible occurrences of theevents. The blue data points correspond with isolated events,the green points with the concurrent occurrence of

combinations of events, and the red point on the horizontalaxis represents the particular situation of no occurrence ofany of the events. Not all the data points are visible becausemany of them correspond with low probability values outsidethe range of the logarithmic scale used in the graph.

▲


Table III

Example of data for construction of the risk map for impact on NPV

Year 2015 2017 2019 2021 2024 2027

LOM year 2 4 6 8 11 14

Section PF % Impact PF % Impact PF % Impact PF % Impact PF % Impact PF % Impact M$ M$ M$ M$ M$ M$

1 0.1 109 0.2 72 0.4 55 0.7 52 1.7 54 3.7 592 0.1 78 0.9 96 5.8 26 11.1 64 17.6 62 23.0 523 1.2 25 1.7 70 2.6 34 3.7 27 5.1 35 6.1 294 3.7 41 5.9 36 8.0 12 10.3 15 13.9 43 16.1 605 0.6 16 5.2 166 9.5 155 12.0 65 15.4 68 19.4 446 0.1 18 0.4 92 1.2 47 2.9 14 7.3 48 10.1 407 0.1 14 0.8 83 2.9 42 6.4 11 9.4 43 12.0 34

Note:– NPV of reference M$ 5.000

Table II

Number of possible cases of occurrence for the situation of 7 events per year

Probabilities and impacts of combination of events

Description No cases P I

Isolated events 7 p1.q1.q2.q3.q4.q5.q6 i12 events 21 p1.p2.q1.q2.q3.q4.q5 i1+i23 events 35 p1.p2.p3.q1.q2.q3.q4 i1+i2+i34 events 35 p1.p2.p3.p4.q1.q2.q3 i1+i2+i3+i45 events 21 p1.p2.p3.p4.p5.q1.q2 i1+i2+i3+i4+i56 events 7 p1.p2.p3.p4.p5.p6.q1 i1+i2+i3+i4+i5+i67 events 1 p1.p2.p3.p4.p5.p6.p7 i1+i2+i3+i4+i5+i6+i7No event 1 q1.q2.q3.q4.q5.q6.q7 0Total 128Notes:Numbers identifying the p, q and i terms in the expressions to calculate P and I should be interpreted as indices that are cycled through the 7 individualevents to generate the number of cases indicated in column 2.p = probability of failureq = probability of no failure = (1 –p)i = economic impact of individual eventP = probability of occurrence of combination of eventsI = cumulative impact of combination of events

Figure 14 – Input data for construction of economic risk map, probability of failure of the slopes (left) and impact on NPV (right)

Nevertheless, these low-probability events have an influenceon the final result, which is captured by the cumulative distri-bution curve. Typically the risk map excludes the frequencydistribution histogram in order to avoid an overcrowdedgraph. A practical way of defining the cumulative distributioncurve of impacts is through a Monte Carlo simulation wherethe seven failure events are modelled with Bernoulli distrib-utions (also called yes-no distributions) and the impactscalculated accordingly.

The probability values given by the risk envelope shouldbe interpreted as probabilities of exceedance of the respectivevalue, as this curve corresponds to a cumulative probabilitydistribution associated with all possible combinations ofevents considered. The risk envelope defines the economicrisk profile for the respective year. The analysis of thepatterns shown by the data points representing theoccurrence of individual events is valuable for identifyingcritical events that push the risk envelope towards the upperright side of the graph. One example of such an event wouldbe the slope failure associated with Section 5 in year 2019 asshown in Figure 15.

The risk envelopes of the six representative yearsincluded in Table III were used to construct the economic riskmap for the life of mine as shown in the graph at the top ofFigure 13. The procedure is based on compounding theprobabilities of the various years for fixed values of impact,considering the periods of the mine life represented by eachyear as shown in Figure 2. The probability values are addedusing the concept of reliability of a system. In this particularexample the probability of an economic impact for the life of

mine (PLOM) for a given impact is calculated from thecorresponding annual probabilities using the followingexpression:

PLOM = 1 - (1-P2015)2 x (1-P2017)2 x (1-P2019)2 x(1-P2021)2 x (1-P2024)3 x (1-P2027)4 [10]

The exponents in this equation correspond to the numberof years represented by the probability value in the respectiveterm. The sum of these exponents is 15 and corresponds withthe total number of years of the mine plan.

A different perspective of the economic risk could beprovided by the analysis of impacts on annual profits,because in this way, future amounts are not discounted topresent values, which in some cases causes a perceiveddistortion of value. Risk maps based on the impacts onannual profits can be calculated following a similar process tothat described for impacts on NPV. Furthermore, the analysiscan be carried out with impacts measured in terms ofcommodity product rather than monetary units, in order toavoid possible distortions caused by the assumptions oncommodity prices.

Uses of the risk mapThere are two main uses of the risk map described in thispaper; one is for the evaluation of a specific open pit designin terms of economic risk by comparing the result withacceptability criteria, and the second refers to the comparativeanalysis of open pit design options, in terms of value andrisk, to identify optimum pit layouts.

Comparison with acceptability criteriaThe risk map can be used to assess a specific pit design bycomparing this result with acceptability criteria specificallydefined for the project. The result of this analysis enables theidentification of the more appropriate risk treatmentstrategies to advance the project. In particular, thecomparison with acceptability criteria is useful for the identi-fication of those years of more relevance in terms of potentialeconomic impacts and the respective critical pit areas causingthose risks. This information is valuable for the definition ofthe areas requiring more investigation in further stages ofstudy and for the evaluation of mitigation strategies to reducethe risks.

Risk acceptability criteria are normally described in theform of a matrix in which risk is categorized in terms oflikelihood of occurrence along the horizontal axis andseverity of the impact up the vertical axis, to define high (H),medium (M), and low (L) risk levels. This type of matrix wasoriginally developed for use in qualitative methods of riskanalysis, with the scales adapted or adjusted to suit differenttypes of application (Joy and Griffiths, 2005). However, amore precise definition of the scales of likelihood and severityresults in acceptability matrices especially suited for the usein quantitative risk evaluation methods such as that based onthe risk map construction described in this paper. Anexample of a risk acceptability matrix is shown in Figure 16,where likelihood and impact categories are defined specif-ically for the project setting at hand. The risk matrix alsoprovides guidelines for risk treatment actions to follow, basedon the risk results.

The use of the risk acceptability matrix in Figure 16 isillustrated in Figure 17, where the risk map results shown in



Figure 15 – Construction of the economic risk envelope for year 2019 inexample from Table III; probability distribution graphs (top) and riskmap result (bottom)


Figure 13 are compared with the acceptability criteria. Thecriteria presented in Figure 16 are intended to adjudicate riskenvelopes of individual years and need to be converted to theappropriate values for the analysis of the LOM envelope. Theconversion is carried out with the same approach used tocalculate the LOM envelope from the annual curves. Thisinvolves adding the annual probabilities using the concept ofsystem reliability, considering a 15-year time span.

In the example presented in Figure 17 the grey curves areincluded for reference but are not intended to be comparedwith the displayed risk zone categories. The evaluation of theindividual years (top graph) indicates a low to moderate riskprofile for all years, with the envelope of year 2019 showinga local elevated risk associated with conditions of Section 5,as depicted in Figure 15. This finding constitutes a pitoptimization opportunity and illustrates the way in which therisk envelopes can be used to identify areas requiringattention in further stages of study. The evaluation of theLOM risk envelope illustrated in the graph at the bottom ofFigure 17 suggests a moderate risk level of the overall mineplan.

Value and risk analysis of design optionsThe risk map can also be used to define risk cost values ofalternative pit slope design options that need to be comparedin terms of economic risk performance. Risk cost values areused to construct the value and risk profile for changing slopegeometries, which provides the elements for screening ofoptions in an early design stage and facilitates the identifi-cation of the main features of pit geometry for an optimumdesign.

Generally, a base case pit slope design is available, whichis the result of conventional slope design methods based onFS or PF criteria, or local experience in terms of slopeperformance in particular geological settings. The base casemine plan typically corresponds with a balanced riskcondition, therefore slope design options on both sides of thebase case are required to define the relationship between theslope angle and the value and risk condition of the pit layout.An example of the construction of alternative pit slopegeometries for the risk analysis from the base case layout isillustrated in Figure 18. In this case the alternative slopedesigns are generated by flattening the base case by 5° andsteepening by 5° and 10°, resulting in nominal slope designangles of 35°, 40°, 45°, and 50° for the east wall and 40°,45°, 50°, and 55° for the west wall.

▲


Figure 16 – Example of risk acceptability matrix for economic impact(top) and the appropriate risk treatment options (bottom)

Figure 17 – Comparison of risk map in Figure 13 with acceptabilitycriteria in Figure 16, for the evaluation of results of individual years (top)and LOM (bottom) Figure 18 – Example of definition of alternative slope design options

The risk maps for the four alternative pit design optionsare constructed using the respective slope stability resultsand economic impact assessment of slope failures. Anexample of the risk envelopes for the life of mine of the fourslope design options shown in Figure 18 is presented inFigure 19. The risk envelopes are compared with the accept-ability criteria (Figure 16), adjusted for a life of mine of 15years. The graph also includes the risk cost values read fromthe envelopes for probabilities of exceedance of 10%, 50%,and 90%, which are used to assess the options in terms ofvalue and risk.

The risk envelopes in Figure 19 indicate that the basecase -05° (BC-05) is in the low to moderate risk threshold,the base case (BC) and base case +05° (BC +05) are in themoderate risk area, and the base case +10° (BC+10) optionfalls in the high risk area. The comparison with the accept-ability criteria does not provide sufficient elements toestablish a clear contrast between the options in terms oftheir risk performance.

The risk cost values indicated in Figure 19 are used toconstruct the value and risk profiles of the slope design asshown in Table IV and Figure 20. These results show thevariation of value in terms of NPV and risk cost for thevarious slope design angles. The design options have beencategorized in terms of the risk results as conservative,balanced, aggressive, and maximum, for the slope designcases of BC -05, BC, BC +05, and BC +10, respectively. Therisk cost or costs of impact of slope failures have an inverserelationship with the probability of incurring those costs,with higher probabilities of small impacts and lowerprobabilities associated with large impacts.

The graph at the top of Figure 20 shows the typicalincrease of risk cost with increasing slope angle for variouslevels of likelihood of impacts. The risk cost values were usedto construct the NPV with risk curves shown in the graph atthe bottom of Figure 20. This graph shows a steady increasein NPV with increasing slope angle when no risk aspects areconsidered. However, once the risk cost is included in theanalysis, the curve of value shows an inflexion point as theslope steepens, defining the angle that represents theoptimum balance between value and risk. The results inFigure 20 would serve to confirm the adequacy of the basecase design, and would suggest a possible optimizationopportunity by steepening the slopes by up to 3 degrees.

Information such as that included in Figure 20 constitutesa valuable tool to optimize the pit design and to bracket theoverall slope angles for further phases of study.

ConclusionsThe methodology presented provides a rational approach todefining, at an early stage of a mine, the main features of pitgeometry reflecting the appropriate balance between value



Figure 19 – Example of risk envelopes of alternative pit design optionscompared with acceptability criteria and risk cost values indicated forthree levels of likelihood

Table IV

Value and risk cost of the pit design options

Case no. Slope angle Design class NPV (M$) Risk costs (M$) NPV with risk (M$)

option (°) P 10 % P 50 % P 90 % P 10 % P 50 % P 90 %

1 BC –05 conservative 4.935 157 88 53 4.778 4.847 4.8822 BC balanced 5.000 170 115 70 4.830 4.885 4.9303 BC +05 aggressive 5.050 205 160 112 4.845 4.890 4.9384 BC + 10 maximum 5.090 275 230 198 4.815 4.860 4.892

Figure 20 – Risk cost (top) and project value (bottom) variations withslope design angle for risk levels of 10%, 50%, and 90%


and risk, in accordance with the specific conditions of theproject. The process considers both the likelihood ofoccurrence of individual slope failure events and the resultingeconomic impacts from all possible combinations ofoccurrence of these events on an annual basis and for themine life. The economic risk map constructed for a particularpit slope layout can be used in an optimization process bycomparing this result with project-specific acceptabilitycriteria. When the process is used for the evaluation ofalternative design options, the risk maps can be used to makerisk cost estimations to calculate the variation of project valuewith slope angle. These results enable the definition of themain features of pit geometry reflecting the appropriatebalance between value and risk, in accordance with thespecific conditions of the mine, which allows the rational-ization of requirements of geotechnical information atdifferent stages of project development, once risk criteriahave been defined.

ReferencesAUSTRALIAN GEOMECHANICS SOCIETY. 2000. Landslide risk management concepts

and guidelines. AGS Sub-Committee on Landslide Risk Management,Sydney, Australia.

BAECHER, G.B., and CHRISTIAN, J.T. 2003. Reliability and Statistics inGeotechnical Engineering. Wiley, Chichester, UK.

CHAPMAN, C. and WARD, S. 2003. Project Risk Management: Processes,Techniques and Insights. 2nd edn. Wiley, Chichester, UK. pp. 148–150.

CONTRERAS, L.F., LESUEUR, R., and MARAN, J. 2006. A case study of riskevaluation at Cerrejon Mine. Proceedings of the International Symposiumon Stability of Rock Slopes in Open Pit Mining and Civil EngineeringSituations, Cape Town, South Africa, 3-6 April 2006. Symposium SeriesS44. Southern African Institute of Mining and Metallurgy, Johannesburg.

CONTRERAS, Lf. and STEFFEN, O.K.H. 2012. An economic risk-based methodologyfor pit slope design. Newsletter of the Australian Centre for Geomechanics(ACG), vol. 39, December 2012.

HARR, M.E. 1996. Reliability-based Design in Civil Engineering. DoverPublications, Mineola, New York.

JOY, J. and GRIFFITHS, D. 2005. National Minerals Industry Safety and HealthRisk Assessment Guideline. Minerals Council of Australia. Version 4, Jan.2005.

LORIG, L., STACEY, P., and READ, J. 2009. Slope design methods. Guidelines forOpen Pit Slope Design. Read, J. and Stacey, P. (eds). CSIRO Publishing,Collingwood, Victoria. pp. 237–264.

MORGAN, M.G. and HENRION, M. 1990. Uncertainty: a Guide to Dealing withUncertainty in Quantitative Risk and Policy Analysis. CambridgeUniversity Press.

READ, J. 2009. Data Uncertainty. Guidelines for Open Pit Slope Design, Read, J.and Stacey, P. (eds). CSIRO Publishing, Collingwood, Victoria. pp.214–220.

ROSS, S. 2010. A First Course in Probability. 8th edn. Pearson Prentice Hall,New Jersey.

SJOBERG, J. 1999. Analysis of large scale rock slopes. Doctoral thesis, Division ofRock Mechanics, Lulea University of Technology, Lulea, Sweden.

STACEY, P. 2009. Fundamentals of slope design. Guidelines for Open Pit SlopeDesign. Read, J. and Stacey, P. (eds). CSIRO Publishing, Collingwood,Victoria. pp. 1–14.

STEFFEN, O.K.H. 1997. Planning of open pit mines on a risk basis. Journal of theSouthern African Institute of Mining and Metallurgy, vol 97, no. 1. pp.47–56.

TERBRUGGE, P.J., WESSELOO, J., and VENTER, J. 2006. A risk consequence approachto open pit slope design. Journal of the South African Institute of Miningand Metallurgy, vol. 106, no. 7. pp. 503–511.

STEFFEN, O.K.H., CONTRERAS, L.F., TERBRUGGE, P.J. and VENTER, J. 2008. A riskevaluation approach for pit slope design. Proceedings of the 42nd US RockMechanics Symposium, 2nd US-Canada Rock Mechanics Symposium,ARMA, San Francisco, USA, June 30 – July 2, 2008.

TAPIA, A., CONTRERAS, L.F., JEFFERIES, M., and STEFFEN, O.K.H. 2007. Riskevaluation of slope failure at the Chuquicamata Mine. Proceedings of the2007 International Symposium on Rock Slope Stability in Open PitMining and Civil Engineering, Perth, Australia, 12-14 September 2007.Potvin, Y. (ed.). Australian Centre for Geomechanics.

VICK, S.G. 2002. Degrees of Belief: Subjective Probability and EngineeringJudgment. American Society of Civil Engineers, Reston, Virginia.

WESSELOO, J. and READ, J. 2009. Acceptance criteria. Guidelines for Open PitSlope Design. Read, J. and Stacey, P. (eds). CSIRO Publishing,Collingwood, Victoria. pp. 221–236. ◆

▲


IntroductionDraglines are self-operated stripping machinesemployed for removing of overburden materialin opencast mines without assistance from ahaulage machine (Figure 1). Theseearthmovers can be more than 4000 t overallweight, with bucket capacities commonly90–120 m3, and have a capital cost up toUS$100 million (Townson, Murthy, andGurgenci, 2003). The productivity of a draglineis influenced by various considerations arisingfrom operational, environmental, and human-based issues. Irregularities and inhomo-geneities in the environment of operation andthe resultant stress variations are the mainissues that cause unsteady loading of thefront-end components of a dragline. Stressaccumulation and induced damage tomechanical elements results in downtime,delays in the production schedule, andincreased maintenance costs and contractorexpenses. During stripping, the resistanceencountered by the bucket is absorbed andtransmitted to other components of the

dragline such as the drag chain, hoist chain,rigging, and boom. The bucket is the source ofexternal forces during operation. Therefore,investigation of the locations of stress concen-tration on the bucket is of paramount concernfor better clarification of potential deficienciesin the bucket. Finite element analysis (FEA)can be effectively utilized to simulate actualcases of the dragline earthmoving process.

FEA has been extensively utilized in manystudies to develop models of the interactionbetween formation and digging tool. Mouazenand Nemenyi (1999) developed a FEA modelto simulate the formation cutting process insub-layers with various geometries. Fielke(1999) presented a model revealing the effectof cutting edge geometry on the requiredcutting force. Davoudi et al. (2008) generateda model capable of estimating draft forcesduring tillage operation. Frimpong and Demirel(2009) examined the stress distribution alonga dragline boom using FEA together withresults acquired by kinematic and dynamicmodelling of the boom. Brittle and plasticdeformability of the formation subjected to theearthmoving process have also beeninvestigated (Chi and Kushwaha, 1989; Raperand Erbach, 1990; Aluko and Chandler, 2004;Aluko, 2008).

The current study intends to bringinnovation to dragline productivity byexamining stress distributions on the bucketbody.

The development of a 3D solid modeling ofa dragline bucket is described. This is followedby an estimation of the resistive force exertedby the formation encountered in earthmovingactivity. The procedures and assumptionsregarding FEA within the scope of study are

Investigation of stress in an earthmoverbucket using finite element analysis: ageneric model for draglinesby O. Gölbaşı* and N. Demirel*

SynopsisDraglines are massive machines extensively utilized in opencast mines foroverburden stripping. The demanding working environment inducesfractures, wear and tear, and fatigue failures in dragline components andeventuates in extended maintenance, lengthy downtimes, and loss ofproduction. The bucket is the main source of external loads on themachinery, since interactions with ground materials take place in thisregion. This study aims to develop a generic finite element model of thestress on an operating bucket. This entails (i) three-dimensional modellingof a dragline bucket, (ii) analytical estimation of resistive forces in thebucket movement, (iii) three-dimensional simulation of the moving bucketusing finite element analysis (FEA), and (iv) sensitivity analysis toexamine the effect of formation characteristics on stress variation.Simulation results imply that the drag hitch and digging teeth are theelements of the bucket that are most prone to failure. In addition,sensitivity analysis indicates that internal friction angle of the formationis the dominant parameter leading fluctuations in stress values. Changesin stress level are least influenced by formation density.

Keywordsdragline bucket, formation-bucket interaction, stress distribution, finiteelement analysis, sensitivity analysis.

* Department of Mining Engineering, Middle EastTechnical University, Ankara, Turkey.

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. Paper receivedJuly 2013 and revised paper received Mar. 2015.



Investigation of stress in an earthmover bucket using finite element analysis

discussed, an investigation of the released stress distributionand sensitivity analysis is presented, and conclusions aredrawn from the study. The framework of the researchmethodology followed is illustrated in Figure 2.

Solid modelling of dragline bucket A dragline bucket body is composed of a back wall, twosidewalls, a floor, an arch, a bucket lip, and teeth that createspace to gather unconsolidated or soft material duringexcavation. The sidewalls of the bucket are slightly inclined

outward. Borders outlined by sidewalls provide rearwardspace with an upward tapering. The back wall has a convexconfiguration with oblique extension. The anterior sections ofthe sidewalls and the floor are integrated with bucket lip atthe front. A rope and chain assembly moves the bucketvertically, and digging teeth are attached to the bucket lipusing connection links. An example of a dragline bucket with50 m3 capacity modelled in Solidworks (Dassault SystèmesSolidWorks Corporation, 2009) can be seen in Figure 3. Thebucket model has a mouth opening of 4.32 m and six digging

▲


Figure 1 – (a) Dragline in operation and (b) schematic view of dragline components

Figure 2 – Research methodology

teeth 0.35 m in width. The bucket body extends 4.88 m inlength, 3.05 m in height; and the width tapers from front toback.

Prediction of formation resistive forceAn earthmoving process is a succession of consecutiveformation failures due to the interaction between theformation and excavation tool. Estimation of resistive forcesresulting from action-reaction behaviour of the formation isessential to calculate the opposing stresses during thestripping movement of a dragline bucket. There are variousempirical and analytical methods available for analysing theforces that are generated during formation cutting. Someresearchers have observed the performance of variousearthmoving machines to predict the cutting resistance of theformation empirically (Alekseeva et al., 1995; Zelenin,Balovnev, and Kerov, 1986; Nedoredzov, 1992, Hemami,Goulet, and Aubertin, 1994). In addition, there are manyother studies that have investigated the formation-toolinteraction in 3D or 2D perspective using analytical methods.Since empirical definitions are constructed from specific fieldobservations, these methods are not able to offer represen-tative estimations for other sites. Analytical methods,however, can be more objective in defining earthmovingprocesses by a holistic approach. This research study utilizesan analytical approach to calculate the approximate resistiveforces imposed on a dragline bucket in operation.

Analytical techniques can be handled as 2D or 3Daccording to the area of utilization. As cited in Blouin’sreview study (Blouin, Hemami, and Lipsett, 2001), 3Dmodels (McKyes, 1985; Swick and Perumpral, 1988;Boccafogli et al., 1992) incorporate the effect of accumulated

material at the edges of the digging tool during operation. 2Dapproaches, on the other hand, do not consider the side effectof the formation resistance in modelling (Osman, 1964; Gilland Vanden Berg, 1968; McKyes, 1985; Swick andPerumpral, 1988). The shape of the excavation tool can beused in decision-making to designate the dimensional type ofprocess. Excavation tool shapes are generally classified asbucket and blade types. 2D resistance models are convenientfor bucket movement since the sidewalls of the body ensurethe direct passage of cut material to the inside and, unlikescraper blades, accumulation of material is restricted,(Blouin, Hemami, and Lipsett, 2001; McKyes, 1985). Thispaper utilizes McKyes’s 2D model (McKyes, 1985) as givenin Equation [1] to estimate the forces due to weight,cohesion, adhesion, overloading, and inertia to express theresistance of a formation to earthmoving.

T = w(γgd2Nγ + cdNc + CadNca + qdNq + γv2dNa) [1]

whereT is the resultant cutting force w is the cutting width γ is the density of the formationg is the gravitational accelerationd is the tool depthc is the cohesionCa is the adhesionq is the overloadv is the formation cutting velocityNγ is the weight coefficientNc is the cohesion coefficientNca is the adhesion coefficienNq is the overload coefficientNa is the inertia coefficient.



Figure 3 – Dragline bucket views from different perspectives


McKyes’s model covers many parameters that cancontribute to variations in formation resistance. This researchstudy neglects the overload pressure due to additional loadon the formation surface and leading to increased compactionof the formation. Adhesion force is considered to benegligible in overburden stripping since this kind of force isencountered only in presence of frictional interaction betweentwo heterogeneous materials. The metal composition andsmooth surface of digging teeth minimizes such interaction.In addition, the inertial effect of the formation is not includedin the formula since inertial fluctuations come into play onlywhile the formation particles are being accelerated from restto a certain velocity (Abo-Elnor, Hamilton, and Boyle, 2003).However, this study handles the model in terms of cutting theformation with constant velocity by means of a draggingaction. Eventually, the general form of McKyes’s equation isreduced to Equation [2].

T = w(γgd2Nγ + cdNc) [2]

The formula parameters obtain their values from both thecutting geometry and the deforming medium. Geometricalvalues can be acquired from Figure 4, which illustrates theinteraction between the solid model and the medium. Thetotal width of cutting medium (w) is 4295 mm and depth ofthe interaction (d) is 512 mm.

N coefficients for weight (γ) and cohesion (c) are acquiredusing Equation [3] and friction angle charts by Hettiaratchand Reece (1974), where δ and φ denote external and internalfriction angles, respectively.

[3]

Medium parameters required for Equation [2] andEquation [3] are obtained from a tillage research study byMouazen and Nemenyi (1999). The calculated weight andcohesion coefficient and resultant cutting force are presentedin Table I. Effective resistance of the medium against thecutting force is estimated to be about 154 kN.

Finite element mesh and boundary conditionsFinite element analysis (FEA) constitutes a virtualenvironment to measure the reaction of a solid model underexternal and internal loads using nodal displacement of solidelements. Prior to implementing the analysis, pre-processingitems such as material and element type, loading, andboundary conditions should be satisfied to ensure theauthenticity of the model under the prescribed limits. FEAmodelling and all simulation in this research study areexecuted in Abaqus 6.9-2 (Dassault Systèmes SimuliaCorporation, 2010).

Materials are assigned to solid models using character-istics of two metals as given in Table II. The material specifi-cations are for two casting metals with strengths of 510 and410 MPa, exhibiting elastic-perfectly plastic behaviour.Meshing of the solid bodies is carried out using a four-nodelinear tetrahedron continuum element denoted as C3D4.Figure 5 illustrates the resultant meshing body, whichincorporates 199 062 solid elements and 45 318 connectionnodes.

One important issue in FEA pre-processing is thedesignation of loading and boundary conditions in asimulation ensuring the cutting movement of dragline bucket.Dragline buckets are filled by a pull-back motion of thebucket toward the machinery housing over a distance fromtwo to three times the bucket length (Demirel, 2011). Thebucket initially penetrates the formation with the digging

▲


Figure 4 – Area of interaction between the formation and the solidmodel

Table II

Material characteristics of the constituent parts (Matbase, 2010)

Part Density (kg/m3) Young’s modulus (N/m2) Poisson’s ratio Yield stress (N/m2)

Teeth 7800 205 x 109 0.30 510 x 106

Main bucket body 7850 200 x 109 0.29 410 x 106

Table I

Input parameters and cutting force

Parameter Value

Formation cohesion strength, c (kPa) 20.40 Density of formation, γ (t/m3) 1.84Internal friction angle, φ (°) 34.00External friction angle, δ (°) 25.00Weight coefficient, Nγ 1.73Cohesion coefficient, Nc 2.65Resultant cutting force, T (kN) 154.0

teeth using its own weight (Figure 6a) and then proceeds tocut the formation by means of the dragging force transmittedalong the drag rope during whole filling cycle (Figure 6b andFigure 6c). The cutting action of the bucket dominates thefilling cycle and leads to stripping of the formation at avelocity between 0.5–0.7 m/s (Frimpong and Demirel, 2009).Simulation in this investigation uses two external loads: thedragging force applied at the drag hitch element of bucket(Figure 3) and the formation resistance applied on thedigging teeth as a distributed load.

Results and discussion

Stress distribution on the bucketThe developed model simulates the formation cutting actionof the dragline bucket. Stress accumulates between thesurfaces of the bucket and the formation, and the resultantfailure of the formation initiates the earthmoving process.Exposure of the bucket to continuous resistance by themedium can also lead to surface fractures on the bucketbody. A non-homogeneous medium and irregularities in thearea being excavated may initiate stress growth and causemechanical failure. Detection of the zones that are mostprone to failure in these conditions is essential for planningpreventive maintenance. The Von Mises stress distribution inthe formation at 200 mm bucket movement is illustrated inFigure 7. It was observed that the Von Mises stress on themedium can reach up to 100 MPa.

The output of finite element analysis of stress distributionon the bucket is presented in Figure 8. Von Mises stress ismostly accumulated on the front-end elements of the bucketsuch as the digging teeth and drag hitch element. Themaximum stress on the bucket is 3.85 MPa, and the general

stress is between 0.013 and 0.3 MPa. Thematic resultsindicate that the concerning medium sample cannot afford tofail bucket elements. However, any stress fluctuations oroverloading initially induce fractures or fatigue on the red-yellow zone of the solid model as shown in Figure 8.

Sensitivity analysis resultsThe sensitivity of stress value to variations in formationproperties such as density, cohesion, internal friction angle,and external friction angle was examined to determine theformation properties that have the most influence on thestress distribution along the bucket. The effect of eachparameter was measured by changing the value by ±20 percent and determining the resultant changes in resistanceforces exerted by the medium. Best-fit lines of the simulationoutcomes for the modified loading conditions for a represen-tative solid element, coded as 24753, in the tooth body areillustrated in Figure 9. The graphs indicate that fluctuationsin internal friction angle have the greatest effect on the rangeof stress concentration on the element. Density, on the otherhand, has a minimal effect on the stress value variance forthe solid element.

ConclusionsSevere operation conditions on draglines, coupled withpressure for continual production and a high utilization, leadto frequent breakdowns of dragline components and ensuing



Figure 8 – Stress distribution on dragline bucket

Figure 7 – Failure of the medium at the initial contact of the draglinebucket

Figure 6 – Interaction of dragline bucket with formation during the fillingprocess

Figure 5 – Meshing of dragline bucket using C3D4 continuum element


pauses in production. These capital-intensive earthmoversshould be operated with high reliability and availability, aswell as longevity, to sustain effective production scheduling.Deterioration of the mechanical elements of draglines can bereduced by minimizing the factors that have an adverseimpact on the condition of the machinery. The bucket is thesource of external loads on the dragline, which aretransmitted along the chain and rope assemblies. Theresistive force that is generated during the dragging action ofthe bucket in the formation can increase the stress intensityin various parts of the bucket. Detection of areas of stressconcentration in the bucket is vital to identify possibleoverloading states in bucket-filling operation. This studyintegrates solid modelling, FEA, and the analytical resistanceapproach to build up a generic model for the investigation ofstress distribution during the dragging movement of adragline bucket.

The simulation results indicated that the tips of thedigging teeth and drag hitch elements are the most stress-intensive points and therefore most prone to failure. Thehighest stress value obtained on the bucket was 3.85 MPa atthe prevailing boundary conditions in the simulation.Although this stress value is not sufficient to cause failure ofthe entire bucket body, any overloading situation may inducefractures or fatigue. Sensitivity analysis revealed that internalfriction of the medium has the greatest effect on stress distri-bution in the bucket, whereas the density of the medium hasthe least influence.

ReferencesABO-ELNOR, M., HAMILTON, R., and BOYLE, J.T. 2003. 3D dynamic analysis of

formation–tool interaction using the finite element method. Journal ofTerramechanics, vol. 40. pp. 51–62.

ALEKSEEVA, T.V., ARTEM'EV, K.A., BROMBERG, A.A., VOITSEKHOVSKII, R.I., andUL'YANOV, N.A. 1985. Machines for Earthmoving Work: Theory andCalculations. Balkema, Rotterdam.

ALUKO, O.B. and CHANDLER, H.W. 2004. A fracture strength parameter for brittleagricultural formations. Biosystems Engineering, vol. 88, no. 3. pp.369–381.

ALUKO, O.B. 2008. Finite element aided brittle fracture force estimation duringtwo-dimensional formation cutting. International Agrophysics, vol. 22. pp.5–15.

BLOUIN, S., HEMAMI, A., and LIPSETT, M. 2001. Review of resistive force modelsfor earthmoving processes. Journal of Aerospace Engineering, vol. 14, no.3. pp. 102–111.

BOCCAFOGLI, A., BUSATTI, G., GHERARDI, F., MALAGUTI, F., and PAOLUZZI, R. 1992.Experimental evaluation of cutting dynamic models in soil bin facility.Journal of Terramechanics, vol. 29, no. 1. pp. 95–105.

CHI, L. and KUSHWAHA, R. L. 1989. Finite element analysis of forces on a planeformation blade. Canadian Agricultural Engineering, vol. 31, no. 2. pp.135–140.

DASSAULT SYSTÈMES SIMULIA CORPORATION. 2010. Abaqus 6.9-2. Rhode Island,USA.

DEMIREL, N. 2011. Effects of the rock mass parameters on the draglineexcavation performance. Journal of Mining Science, vol. 47. pp. 442–450.

DAVOUDI, S., ALIMARDANI, R., KEYHANI, A., and ATARNEJAD, R. 2008. A twodimensional finite element analysis of a plane tillage tool in formationusing a non-linear elasto-plastic model. American-Eurasian Journal ofAgricultural and Environmental Science, vol. 3, no. 3. pp. 498–505.

FIELKE, J.M. 1999. Finite element modelling of the interaction of the cuttingedge of tillage implements with formation. Journal of AgriculturalEngineering Research, vol. 74. pp. 91–101.

FRIMPONG, S. and DEMIREL, N. 2009. Case study: planar kinematics of draglinefor efficient machine control. Journal of Aerospace Engineering, vol. 22,no. 2. pp. 112–122.

GILL, W.R. and VAN DEN BERG, G.E. 1968. Formation Dynamics in Tillage andTraction. Agricultural Research Service, Washington, USA.

HEMAMI, A., GOULET, S., and AUBERTIN, M. 1994. Resistance of particulate mediato excavation: application to bucket loading. International Journal ofSurface Mining, Reclamation and Environment, vol. 8. pp. 125–129.

HETTIARATCHI, D. and REECE, A. 1974. The calculation of passive formationresistance. Geotechnique, vol. 24, no. 3. pp. 289–310.

MATBASE. 2010. Material Property Database.http://www.matbase.com/material/ferrous-metals/cast-steel/ [Accessed 20June 2010].

MCKYES, E. 1985. Formation Cutting and Tillage. McGill BioresourceEngineering. http://www.mcgill.ca/files/bioeng/BREE512_part1.pdf

MOUAZEN, A.M. and NEMENYI, M. 1999. Finite element analysis of subsoilercutting in non-homogeneous sandy loam soil. Formation and TillageResearch, vol. 51. pp. 1–15.

NEDOREDZOV, I. 1992. Forces prediction of underwater formation cutting byexcavating robots. 9th International Symposium on Automation andConstruction, Tokyo.

OSMAN, M.S. 1964. The mechanics of formation cutting blades. Journal ofAgricultural Engineering Research, vol. 9, no. 4. pp. 313–328.

RAPER, R.L. and ERBACH, D.C. 1990). Prediction of formation stresses using thefinite element method. Transactions of the ASAE, vol. 33, no. 3. pp.725–730.

DASSAULT SYSTÈMES SOLIDWORKS CORPORATION. 2009. Solidworks. © Concord,Massachusetts, USA.

SWICK, W.C. and PERUMPRAL, J.V. 1988. A model for predicting formation-toolinteraction. Journal of Terramechanics, vol. 25, no. 1. pp. 43–56.

TOWNSON, P.G., MURTHY, D.N., and GURGENCI, H. 2003. Optimization of draglineload. Case Studies in Reliability and Maintenance. Blischke, E.W. andMurthy, D.N. (eds). Wiley. pp. 517–544.

ZELENIN, AN., BALOVNEV, V.I., and KEROV, L.P. 1986. Machines for Moving theEarth. Balkema. Rotterdam. ◆

▲


Figure 9 – Effects of formation properties on stress values

IntroductionAdvancements in the mining industry in thelast three decades have primarily focused onimprovements in heavy machinery, supportsystems, and safety equipment. Recently thefocus has shifted towards development ofcommunication systems for better safety andconnectivity. In this context, the wirelesssensor network (WSN) technology, owing toits efficiency, speed, and applicability inemergency conditions, has come out on top(Fiscor, 2011; Liu, 1996; Patri, Nayak, andJayanthu, 2013). The current need is for areliable wireless system in the harshunderground mine environment(Bandyopadhyay et al., 2009), in which radiopropagation models are playing a vital role.

Recent studies have considered theunderground mine as a hybrid case of regularand harsh environments and shown that thesignal propagation models and criticalparameters of wireless channel propagation foran indoor environment are similar to anunderground mine scenario at 900 MHz,indicating that the wireless nodes used in the

indoor environment can be modified for use inmines (Qaraqea et al., 2013; Murphy et al.,2008). Zhang et al. (2001) experimented at900 MHz with two different scenarios, namelythe gateroad and working face of a longwallcoal mine, in order to evaluate the additionallosses due to gateroad curvature and thepresence of mining equipment, andsubsequently modified the wave guidepropagation model. The hybrid tunnelpropagation model developed by Zhang et al.uses both a free space propagation model anda modified waveguide propagation model todescribe the propagation characteristics. Somesimulation tools have also been developed forpath loss calculation and propagationmodelling by taking into account the effects ofbarriers. The simulations were carried out byvarying the frequency with standard tunneldimension, shape, and material properties.Comparison with an actual scenario provedthat the path loss is mostly dependent ontunnel dimension, and signal frequency(Hrovat, Kandus, and Javornik, 2012).

With advances in micro-electro mechanicalsystems (MEMS), transceivers working at2.4GHz are now available at a reasonable price(Wurneke and Pister, 2002). The betterperformance of such transceivers inlocalization within a small range is due tohighly directional antennae and a very highoperational frequency, resulting in less noise.Liu et al. (2009) studied the transmissionperformance of WSN near a mine working faceat 2.4 GHz frequency, incorporating all theelectromagnetic properties in their theoreticalmodel and comparing it with experimentalresults. The effective transmission distancewas studied for IEEE 802.15.4, known as theZigBee protocol (Liu et al., 2009; IEEE Std802.15.4. 2011).

Radio frequency propagation model andfading of wireless signal at 2.4 GHz inan underground coal mineby A. Patri* and D. S. Nimaje*

SynopsisWireless sensor networks and wireless communication systems havebecome indispensable in underground mines. Wireless sensor networks arebeing used for better real-time data acquisition from ground monitoringdevices, gas sensors, and mining equipment, whereas wireless communi-cation systems are needed for locating and communicating with workers.Conventional methods like wireline communication have proved to beineffective in the event of mine hazards such as roof falls, fires etc. Beforeimplementation of any wireless system, the variable path loss indices fordifferent workplaces should be determined. This helps in better signalreception and sensor node localization, and also improves the method bywhich miners carrying the wireless devices are tracked. This paperproposes a novel method for determining the parameters of a suitableradio propagation model, which is illustrated with the results of a practicalexperiment carried out in an underground coal mine in southern India. Thepath loss indices, along with other essential parameters for accuratelocalization, have been determined using the XBee modules and ZigBeeprotocol at 2.4 GHz frequency.

KeywordsWSN, RSSI, path loss index, miner localization, underground coal mining,ZigBee.

* Department of Mining Engineering, NationalInstitute of Technology Rourkela, India.

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. Paper receivedFeb. 2014 and revised paper received Mar. 2015.



Radio frequency propagation model and fading of wireless signal at 2.4 GHz in an underground coal mine

In this paper, the radio frequency (RF) propagation modelhas been developed and the path loss of wireless signal at 2.4GHz was experimentally derived for the GDK 10A incline, alongwall underground mine of Singareni Collieries CompanyLimited (SCCL). Before implementing WSN, the path lossindex and other parameters should be calculated to performbetter localization, base-station placement and optimization,improve receiver design, and combat the fading of signal(Iskander and Yun, 2002). The reception distance wasdetermined by utilizing the path loss of the signal, whichdetermines the energy loss factor. The repeaters should beplaced accordingly and their amplification factors should beset to different values to achieve a high-efficiency wirelesscommunication system for different environments. Theperformance of ZigBee protocol using the XBee modules wasexperimentally studied for the mine.

Radio frequency propagation modelsA wireless propagation model can be defined as amathematical expression or an algorithm for predicting theradio characteristics of a particular type of environment.There are two types of wireless propagation model:deterministic models and empirical models (Iskander andYun, 2002; Rappaport, 2002). The deterministic model doesnot fit into the real environment properly; however, for low-frequency waves, the results produced by the deterministicmodel are approximately equal to the actual result, with avery low rounding error. Since the operating range is muchless, elements present in the surroundings have a significanteffect on propagation in the high-frequency channel whilevariations due to environmental effects are largelyinsignificant in the low-frequency channel. The aforemen-tioned propagation models are again subcategorized intothree types, i.e. free space propagation models, two-rayground models, and lognormal models. These models aredeterministic with the exception of the lognormal model,which is empirical.

Free space propagation modelThe free space propagation model is a simplified model thatassumes line-of-sight communication between thetransmitter-receiver pair and that there is no interveningobstruction. The mathematical representation of the modelcan be written as

[1]

where, Pr and Pt represent the power received and powertransmitted respectively, CT is a constant that depends on thetransceiver, and d is the distance between the transmitter-receiver pair.

Two-ray ground model This model is obtained by modifying the free spacepropagation model after taking into account the effect ofreflection of signals. It is also assumed that both the directand the reflected ray are used for communication. In thismodel the distance between the transmitter-receiver pair ismuch greater than their individual heights, and it can berepresented as

[2]

where, Ct is the constant representing transceiver charac-teristic in the two-ray ground model.

Log-distance model The log-distance model is an analytical and empirical modelwhich can be mathematically represented as

[3]

where, η represents the path loss factor or distance powergradient.

The actual results vary from the results derived using thelog-distance model. Hence, for hostile environments likeunderground mines, models have to be developed by usingshadow-fading phenomena.

At high frequencies, power loss is different for differentlocations owing to obstructions in the path between twocommunicating devices. Figure 1 shows a typical example ofthis phenomenon, where the dotted circle shows the idealboundary of operation for an omnidirectional antenna placedat the centre, and the bold line shows the actual boundary ofoperation with a minimum and maximum range of R1 and R2respectively due to presence of various obstructions. For thispurpose, the empirical model is chosen over the deterministicmodel to predict or calculate power received at a particulardistance from the transmitter (Pahlavan and Levesque,2005).

Moreover, the power loss can be subdivided into twoparts on the basis of fluctuation around the average pathloss, i.e. multi-path fading and shadow fading. In case ofmulti-path fading, the transmitted signal reaches the receiverthrough two or more paths, causing both constructive anddestructive interferences near the receiver which in turn leadsto phase shifting and addition of noise. Therefore in adynamic environment, where both the transmitter andreceiver are stationary, the received signal strength (RSSvalue) varies randomly due to the movement of objects and

▲


Figure 1 – Variation in operation range due to fading of signal radiatedfrom the omnidirectional antenna

small changes in the environment. The long-term average ofRSS values represents the effect of shadow fading of signalthat is caused by the presence of a constant barrier betweenthe transceivers (Pahlavan and Levesque, 2005).

Although time of arrival (TOA), angle of arrival (AOA),and time difference of arrival (TDOA) provide higher accuracyin most cases, they fail in a harsh mining environment (Gentileet al., 2013; Sahoo and Hwang, 2011). Therefore, the receivedsignal strength index (RSSI)-based model for localization hasbeen developed. This low-cost RSSI-based localization providesless communication overhead with lower complexity ofimplementation. The distance or range of the signal can becalculated accordingly by the loss factor of the environmentfrom the RSSI-based equations [4] and [7].

Shadow-fading model and proposed scheme forparameter determinationThe log-distance model can be represented more accuratelyby introducing a Gaussian distribution variable to representthe fading or fluctuation of received signal strength. Themodified model is called the lognormal shadowing model andit is most appropriate for wireless sensor networks since it isall-inclusive in nature and can be easily configured accordingto the target environment (Nafarieh and Ilow, 2008). Themathematical equation for the above relation can be definedas

[4]

where,

[5]

and d0 is the near-earth reference distance. The randomvariable ψ is the zero-mean Gaussian random noise, theprobability distribution function of which is given by

[6]

The value of η depends on the surrounding orpropagation environment as per Equation [4]. The distanced0 is taken to be one metre for simplicity of calculation, and itcan also be represented in the terms of received power orRSSI as

[7]

In Equation [4], there are two unknown terms, η and ψ,which should be determined experimentally. The linearregression analysis for the data-set with distance andreceived power as attributes gives the η value, which can befurther used for that particular place with unknown distanceand known received power to localize a wireless node.

In Equation [4], Var(ψ) = σ2 and E(ψ) = 0. Therefore, itcan be mathematically proven that Var(σψ1) = σ2 and E(σψ1)= 0. This relationship shows that the ψ function has the samedistribution as ψ1, where ψ1 represents the zero-meanGaussian distribution with unit variance. Equation [4] can bemodified as

[8]

Assuming maximum error with 95% confidence interval,the σψ1 value can be replaced by 1.96 σ, which gives

[9]

However, observational analysis shows that the standarddeviation varies as a function of distance, and on the basis ofconsiderable experimental evidence, we claim it to be afourth-degree polynomial function:

[10]

Now the observational error ε can be defined as thedifference of these two terms, i.e. experimental and observa-tional σ.

[11]

In order to avoid negative error and for solving thisexpression, the objective function ∈ can be written as

[12]

To obtain the values of the coefficients of the polynomial,i.e. a, b, c, e, and f, a partial derivative method is adopted,and it can be mathematically represented as the following setof equations:

[13.1]

[13.2]

[13.3]

[13.4]

[13.5]

The above set of equations can be solved in matrix form,to obtain the coefficients




[14]

If the coefficients and path loss index for a particularplace are known, the standard deviation and the power lossdue to fading can be calculated for the new set of data withknown RSSI and unknown distance, for accurate localization.

Mining conditions at GDK 10A mineThe GDK 10A incline of SCCL is situated at Ramagundam inTelengana, India, in the Godavari valley coalfield. Figure 2shows a schematic layout of a longwall mine. The minimumand maximum depths of Seam 1, where the experiments werecarried out, are 175 m and 310 m respectively, and the seamthickness is 6.5 m.

The surface area is flat with undulating terrain having agentle slope towards northeast and south. The coal seam isaccessed via two tunnels, with lengths of 450 m and 500 mat a gradient of 1 in 4.5 and 1 in 5, for haulage and manwayrespectively. The mine floor is mainly grey sandstone and theroof is coal with a 0.30 m clay band.

The length and width of the longwall face are around 150m and 1 km respectively, with an average depth of 350 mfrom surface. Coal cutting is by means of an Andersondouble-ended ranging drum shearer with a diameter of 1.83m and a web width of 0.85 m. Caterpillar independent frontsuspension-based hydraulic powered roof supports areprovided with 101 PMC-R controlled hydraulic chocks.Anderson bridge-type stage loaders are used in the gateroadto transport the coal from the armoured face conveyor (AFC)to the belt conveyor. The 260 m long DBT-manufactured AFCis used in the face, with a pan size of 232 × 844 × 1500 mmand deck plate thickness of 35 mm at an average chain speedof 1 m s-1.

The head and tail gateroads are driven in parallel throughSeam 1. The gateroad wall surface is rough and waterpercolates from the strata and the gateroads. The gateroadbearing the belt conveyor system has an average height andwidth of 3.6 and 4.2 m respectively. The conveyor belt,supported by a steel structure, is at a height of 1.32 m to 1.4m from the floor and carries an average lump size of 200 ×200 × 200 mm. The belt has a width of 0.8 m to 1.2 m, and ismade mainly of rubber. The roof supports are generally wiremesh type with bolts and girders. The material properties,dimensions, and other features of the equipment describedhave a major influence on signal propagation, together withmine dimension, rock properties, slope, and other geo-miningconditions.

Experimental set-up and procedure

Instruments and set-upA pair of XBee series-1 modules, one being used as atransmitter and the other as a receiver, which implement theZigBee protocol, each capable of transmission or reception,were used for wireless communication at 2.4 GHz. Thespecifications of the XBee module are given in Table I. Eachof the XBee modules is configured by setting the preferreddata rate, modulation technique, lapse rate between packets,and other parameters using X-CTU software by mounting themodules on the XBee USB adapter (which has an onboard3.3 V low-drop voltage regulator and light-emitting diode(LED) indicators for RSSI, associate, and power), and thenconnecting to a computer’s universal serial bus (USB) portthrough a FT232 USB-to-serial converter. There are twomodes of operation for the XBee module; in transparent datamode (AT) the signal coming to the Data IN (DIN) pin is sentdirectly to the receivers, while in application programminginterface mode (API) (which was used in this study), the datais sent in the form of packets that include the receiveraddress along with a feedback for the delivered packets,payload information, and various parameter settings toincrease the reliability of the network and to send the signalsafely over the wireless network (Hebel, Bricker, and Harris,2010). The module has a mounted rubber-duck wire antennaor whip antenna, which radiates in a nearly omnidirectionalpattern. As there is very little distortion in radiation pattern,the antenna is considered to radiate equal power in allazimuthal directions (Bandyopadhyay, Chaulia, and Mishra,2010).

▲


Figure 2 – Schematic layout of longwall mining method

Table I

Specifications of XBee module

Parameter Property

Raw data rate 2.4 GHz: 250 kbps (ISM band)Maximum range Indoor: 30 m; outdoor (line of sight): 100 mReceiver sensitivity -92 dBm (1% packet error rate)Channels 16 channelsAddressing Short 8-bit or 64-bit IEEETemperature -40 to +85°CChannel access CSMA-CA (Carrier Sense Multi Access-Collision

Avoidance)

This module also supports universal asynchronousreceiver/transmitter (UART) interface, which is beneficial forclock-setting and connecting it to a microcontroller. TheATMEL Atmega-32 microcontroller (14.7456 MHz crystal)development board was used, which has a compatible UARTserial communication integrated circuit together withelectrically erasable programmable read-only memory(EEPROM), static random access memory (SRAM), and anin-system self-programmable flash memory of 1024, 2k, and32k bytes respectively. It has an inbuilt reverse polarityprotection and the 7805 voltage regulator has a heat sink forcontinuous dissipation to supply 1 A current constantlywithout overheating. The Request to Send (RTS) and Clear toSend (CTS) module pins can be used to provide flow control.CTS flow control provides an indication to the host to stopsending serial data to the module. RTS flow control allowsthe host to signal the module not to send data in the serial-transmit buffer through the UART. Data in the serial-transmitbuffer will not be sent out through the Data OUT (DOUT) pinas long as RTS is de-asserted or set high. The UARTconnections for the transmitter and receiver module areshown in Figure 3. The module operates in a low-voltagerange of 2.8–3.4 V, but for the whole set-up, a pair of 12 V1.3 A.h DC batteries of lead-acid type was used, one for eachnode. This battery can be replaced by a cap-lamp battery usedin underground mines in compliance with Directorate Generalof Mine Safety India (DGMS) standard. A liquid crystaldisplay (LCD) is programmed and connected to the microcon-troller unit at the receiver to display the desired output. Thetransmitter and receiver units are shown in Figure 4.

For use in underground mines, the electronic instrumentmust be intrinsically safe to avoid any fire hazard. SinceZigBee protocol-based wireless modules have been used inunderground mines worldwide, they can be considered asintrinsically safe for most of the underground miningscenarios in India (Bandyopadhyay, Chaulia, and Mishra,2010; Chen, Shen, and Zhou, 2009). Parameters required forthe XBee module to be intrinsically safe are specified in Table II. The ZigBee protocol is based on the carrier sensemultiple access (CSMA) with collision avoidance (CA)channel access to provide energy saving, latency, andnegligible error in the received data packet. Direct sequencespread spectrum (DSSS) modulation is used in the PHY layer,which has high resistance to noise or jamming. The ZigBeestandard supports star, tree, and mesh networks, thuspermitting numerous applications. In sleep mode it uses only0.1 μA which helps in energy saving during idle periods. Itsupports AES-128 encryption that converts a 128-bit plaintext to a 128-bit cipher text. It has a capacity to acquire morethan 256 peer-to-peer connections in a master-slave configu-ration; which is very high compared to other wirelessprotocols used in day-to-day life.

The experiment was divided into two parts, namely anRSSI test and a range test. The RSSI test provides the data fordetermining path loss index and various parameters affectingthe localization and fading of power, and the range test givesthe operation range of the module in different undergroundmine scenarios.

RSSI test The first set of readings was taken at the longwall face withshearer, hydraulic power supports, AFC, stage loader, andother machinery which obstructed the wireless signal. Toavoid fast fading of the signal, the readings were taken in astatic environment free from moving machinery or menbetween the transmitter-receiver pair. A second set ofreadings was taken beside the belt conveyor system, inrunning condition, installed in the gateroad, which wouldhave created some fast fading.

Range test The range test was conducted in three different places – nearthe longwall face, the belt conveyor system, and in theinclined mine car pathway.

Experimental procedure Firstly, an RSSI test was performed and readings were takenby fixing the transmitter node at the beginning of the



Figure 4 – Transmitter and receiver unit

Figure 3 – UART connections for the transmitter and receiver module

Table II

Parameters required for an intrinsically safeinstrument (source: Digi International, n.d.)

XBee Series 1 IEEE 802.1.5.4 Properties Values

Maximum power at antenna connector 2 mWMaximum current at antenna connector 7 mA (AC current at

2.4 GHz)Sum total of all capacitance on PCB 757 pFSum total of all inductance on PCB 60 nHLargest capacitor on PCB 220 pFLargest inductor on PCB 56 nH


longwall face close to the hydraulic powered roof support at aheight of 1.5 m from the floor. The transmitter and receiverset-ups were kept at a distance of 1 m and 2 m from thechocks and the working face respectively. The transmitternode was programmed to send 100 packets with a delayinterval of 500 ms between two subsequent packets, and LCDshowed the average RSSI over these 100 packets. TwentyRSSI readings were taken at each position of the receivernode and the same procedure was repeated up to a distanceof 20 m with a 1 m step size. The packet received rate (PRR)was also calculated and displayed on the LCD at distanceintervals of 1 m, and all the readings were taken in line-of-sight conditions. The second set of readings was taken on thegateroad near the belt conveyor system. The transmitter nodewas fixed at a location exactly 1 m above the floor, 0.5 mfrom the belt conveyor, and the receiver node was kept atvarying distances (1–20 m) from the transmitter node alongthe passage.

The range test for the XBee module was then carried outsequentially in all the three areas by fixing the transmitternode at a particular location and moving the receiver nodeaway until the LCD showed a ‘zero’ value for the RSSI andindicated that the packet sent by the transmitter could not bereceived beyond that particular distance.

Results and analysisThe data collected near the working face and the beltconveyor gateroad is represented in Tables III and IV respec-tively. The standard deviation was calculated for each set ofRSSI values on every location.

The standard deviation (SD) can be calculated as

[15]

where, SDi is same as Yi in Equation [9] for a particulardistance di, Xj represents the different RSSI values recorded ateach distance di, M is the mean RSSI, and n is the totalnumber of observations (i.e. 20). The integer variables i and jboth vary from 1 to 20.

MATLAB version 7.6.0.324 r2008a was used for thelinear regression analysis model. The slope of the fittedgradient line denotes the path loss index for the place of theexperiment, for longwall working face the value was found tobe 2.14. Figure 5 (A) depicts the scatter plot of the receivedsignal for the longwall face corresponding to the logarithmicdistance. The higher value of path loss index indicates thatfading of the signal was due to the presence of moreobstructions than in the normal outdoor scenario. Moreover,it also implies that more repeaters should be placed and theinternode distance should be kept small compared to typicaloutdoor scenario (for which the index is 2). More fading andgradual degradation of power transmitted was due to thepresence of metallic bodies; homogenous obstructions presentin the surroundings and the static nature of the environmentresulted in less standard deviation (more concentrated in theregion of 3.5 to 6) from the mean RSSI values. The values ofPRR show a dependency on both standard deviation andreceived power, with a higher correlation with the former.The signal is marginally affected by the waveguide propertyof the tunnel for the first 3–4 m, after which the effect

increases gradually. A trade-off is observed between distancecovered and the wave guide effect, leading to a fluctuation ofRSSI over a small range. As discussed previously, the curvefitting was done to find a relationship between the standarddeviation and distance to determine the coefficients for thelongwall mining area as shown in Figure 5 (B). The coeffi-cients a, b, c, e, and f of the fourth-degree polynomial arefound to be 2.626 × 10−6, 6.176 × 10−3, -0.2276, 2.403, and-1.721 respectively. R2 and root mean square error (RMSE)were 0.8332 and 0.6958 respectively.

For the belt conveyor gateroad, the path loss index wasfound to be 1.568, using linear regression analysis. Figure 6

▲


Table III

Data collected near the longwall face of GDK 10A

Distance (m) M (dBm) SD (dBm) PRR (%)

1 -51.65 0.48936 1002 -57.65 2.00722 1003 -71.5 4.54799 96.594 -69.8 3.67924 96.765 -73.95 5.78996 96.296 -76.1 4.93004 95.837 -76.85 5.83343 95.78 -78.45 6.88665 95.079 -80.25 6.04261 95.0810 -76.55 6.60522 95.4511 -76.8 5.94491 95.6512 -81.15 4.56828 93.9213 -80.95 3.64872 93.8914 -81.85 4.22119 93.915 -79.35 3.54334 94.216 -80.95 4.20443 93.7717 -82.6 4.87097 92.7118 -81.6 3.93901 93.8519 -84.15 4.51051 90.0520 -86.85 4.88041 86.2

Table IV

Data collected near the belt conveyor gateroad

Distance (m) M (dBm) SD (dBm) PRR (%)

1 -54.2857 3.48056 99.372 -60.0952 1.92106 99.33 -68.5714 7.59402 95.734 -67.0476 7.89087 95.225 -67 7.75887 96.196 -73 4.12311 96.047 -73.6667 6.5904 95.988 -70.6191 5.45414 96.539 -73.1905 6.14261 95.910 -68.2381 5.76052 96.311 -66.1905 4.44491 97.2412 -69.5714 3.35517 96.8313 -69 3.6606 96.8914 -75 5.12119 95.515 -75.3333 4.23478 95.8116 -79.8095 4.7394 9417 -75.5714 3.99464 95.1418 -76.5714 5.59081 94.6319 -74.5455 5.41363 94.9920 -83 5.54076 92.8



Figure 5 – (A) Variation of RSSI with respect to distance near the longwall face, (B) relationship between standard deviation and distance from the longwallface

Figure 6 – (A) Variation of RSSI with respect to distance in the belt conveyor gateroad, (B) relationship between standard deviation and distance for the beltconveyor gateroad

(A)

(B)

(A)

(B)


(A) depicts the scatter plot of RSSI vs. the logarithmicdistance. The lower value of power loss compared to thelongwall face was due to the predominant effect of thewaveguide property of the tunnel. The standard deviations(more concentrated in the region of 4–7.5 m) from the meanRSSI values were high compared to the longwall face area dueto inhomogeneous surroundings like different supportsystems, material, and spacing, machinery, variable coallump size carried by the belt, and other distributiveobstructions. Due to movement of the belt conveyor carryingcoal lumps of various sizes, some fast fading was observed,as indicated by the dispersal of data from the fitted line. Thesignal loss for a particular place was found to be greater thanits consecutive place readings, each taken at 1 m distance,due to presence of girders over the receiver. The presence offewer metallic bodies in the gateroad compared to thelongwall face reduced the fading effects. The signalpropagation was mildly affected by the steel structurebecause the nodes were located higher than the belt conveyorsupport structure. Figure 6 (B) depicts the curve fitting forthe fourth-degree polynomial. The coefficients fordetermining the standard deviation as a function of distancewere found to be -6.685 × 10-4, 0.3418 × 10-1, -0.5813,3.599 and -0.4563 for a, b, c, e, and f respectively. The R2

value of 0.474 and RMSE value of 1.281 indicate thefluctuation of standard deviation due to fast fading.

From the range test, it was found that the XBee moduleprovides satisfactory results up to a range of 40–45 m, 60–65m, and 75–85 m for the longwall face, belt conveyorgateroad, and mine car pathway respectively.

ConclusionThis study reveals that the efficiency of an underground minecommunication system is dependent on the environment.Before implementing any wireless system in undergroundmines, the path loss index and the variance of Gaussiandistribution representing the shadow fading effect should bedetermined. This helps in determining the distance at whichrepeaters should be placed in order to enhance the signal andlocalize the sensor node from its received signal strength.With an increasing number of physical obstructions, the pathloss index increases, resulting in the total loss of signalbeyond a particular range. The XBee module facilitatessatisfactory wireless communication over an adequate rangeof operation with a negligible packet error rate. The PRRdepends upon transmitter distance and dynamic behaviour ofthe surroundings. These intrinsically safe modules areeconomic, energy-efficient, and enhance the mine safetysystem by facilitating tracking of miners and real-time dataacquisition from sensors. The experiment was carried out in ahazard-prone underground coal mine. The experimentalresults may vary for underground mines other than coalmines, due to the variation in the rock mass properties anddimensions of tunnels, passages, galleries, and workingareas, depending on the mining method. In our current work,two nodes were used for experimentation. To ensure theviability of the ZigBee protocol, further studies could becarried out to analyse the network performance using morethan two nodes.

AcknowledgementWe wish to express our sincere gratitude to the authorities of

SCCL for permission and assistance in carrying out theexperiment and collect valuable data at GDK 10A. We thankthe anonymous reviewers for their valuable comments.

ReferencesBANDYOPADHYAY, L.K., CHAULIA, S.K., and MISHRA, P.K. 2010. Wireless

Communication in Underground Mines. Springer Science.

BANDYOPADHYAY, L.K., CHAULIA, S.K., MISHRA, P.K., CHOURE, A., and BAVEJA, B.M.2009. A wireless information and safety system for mines. Journal ofIndustrial and Scientific Research, vol. 68, no. 2. pp. 107–117.

CHEN, G., SHEN, C., and ZHOU, L. 2009. Design and performance analysis ofwireless sensor network location node system for underground mine.Journal of Mining Science and Technology, vol. 19, no. 6. pp. 813–818.

DIGI INTERNATIONAL. Not dated. Intrinsically Safety Certifications.http://www.digi.com/support/kbase/kbaseresultdetl?id=2128 [AccessedJanuary 2014].

FISCOR, S. 2011. Miners quickly adopt new communication systems. Coal Age,January 2011. pp. 2–6.

Gentile, C., Alsindi, N., Raulefs, R., and Teolis, C. 2013. GeolocationTechniques: Principles and Applications. Springer Science.

HEBEL, M., BRICKER, G., and HARRIS, D. 2010. Getting Started with XBee RFModules Version 1.0. (Parallax Inc.).

HROVAT, A., KANDUS, G., and JAVORNIK, T. 2012. Path loss analyses in tunnelsand underground corridors. International Journal of Communication, vol.6, no. 3. pp. 136–144.

IEEE Std 802.15.4. 2011. IEEE Standard for Local and Metropolitan AreaNetworks. Part 15.4: Low-Rate Wireless Personal Area Networks (LR-WPANs).

ISKANDER, M.F. and YUN, Z. 2002. Propagation prediction models for wirelesscommunication systems. IEEE Transactions on Microwave Theory andTechniques, vol. 50, no. 3. pp. 662–673.

LIU, S.Y. 1996. Advances in longwall coal mining techniques. Seminar on HighProductivity and High Efficiency Coal Mining Technology, Beijing.

LIU, X., WANG, M., WEN, J., and ZHAO, Z. 2009. Transmission performance of2.4 GHz wireless sensor nodes when used in a working-face environment.Journal of Mining Science and Technology, vol. 19, no. 2. pp. 185–188.

MURPHY, R.R., SHOURESHI, R., ARNOLD, H.W., ARSLAN, H., BURKE. J., GREENSTEIN,L.J., KILLINGER, D.K., LUNDGREN, C.W., RUSTAKO, A.J., and STOVER, S. 2008.Analysis of Viability and Feasibility of Current and Emerging MiningCommunication and Mine Rescue Technologies. Final Report, Institute forSafety, Security, Rescue Technology, University of South Florida, Tampa,FL, USA.

NAFARIEH, A. and ILOW, J. 2008. A testbed for localizing wireless LAN devicesusing received signal strength. Proceedings of the 6th AnnualCommunication Networks and Services Research Conference, Halifax,Canada, May 2008. pp. 681–687.

PAHLAVAN, K. and LEVESQUE, A.H. 2005. Wireless Information Network. 2ndedn. Wiley-Interscience.

PATRI, A., Nayak, A., and Jayanthu, S.J. 2013. Wireless communication systemsfor underground mines – a critical appraisal. International Journal ofEngineering Trends and Technology, vol. 4, no. 7. pp. 3149–3153.

QARAQEA, K.A., YARKANB, S., GÜZELGÖZC, S., and ARSLAN, H. 2013. Statisticalwireless channel propagation characteristics in underground mines at 900MHz: a comparative analysis with indoor channels. Journal of Ad HocNetworks, vol. 11, no. 4. pp. 1472–1483.

RAPPAPORT, T.S. 2002. Wireless Communications: Principles and Practice. 2ndedn. Prentice Hall.

SAHOO, P.K. and HWANG, I. 2011. Collaborative localization algorithms forwireless sensor networks with reduced localization error. Sensors, vol. 11,no. 10. pp. 9989–10009.

WURNEKE, B.A. and PISTER, K.S.J. 2002. MEMS for distributed wireless sensornetworks. Proceedings of the 9th International Conference on Electronics,Circuits and System, Dubrovnik, Croatia, September 2002. pp. 291–294.

ZHANG, Y.P., ZHENG, G.X., and SHENG, J.H. 2001. Radio propagation at 900 MHzin underground coal mines. IEEE Transactions on Antennas andPropagation, vol. 49, no. 5. pp. 757–762. ◆

▲


IntroductionBlasting is one of the most economical andenergy-efficient methods of rock fragmen-tation, and is widely used in mining, civil,construction, and environmental projectsaround the world. However, there are severaldrawbacks, including (but not limited to)complaints from nearby residents (Kahriman,2001), damage to residential structures (Singhet al., 1997; Gad et al., 2005; Nateghi et al.,2009), damage to adjacent rock masses andslopes (Villaescusa et al., 2004; Yi and Lu,2006; Singh et al., 2005), damage to existinggroundwater conduits, and damage to theecology of the nearby area (Khandelwal andSingh, 2007). The main cause of theseundesirable effects is excessive blast-inducedground vibrations. Thus, predicting theadjacent ground vibrations is essential forsafe, environmentally responsible, andsustainable blasting operations. Groundvibrations can be defined and measured interms of peak particle displacement, velocity,

acceleration, and frequency. The peak particlevelocity (PPV) has been used by manyresearchers as a versatile metric for bothpredicting and controlling the blast-inducedground vibrations. There are three majormethods cited in the literature for PPVprediction, including empirical, theoretical, andartificial intelligence techniques.

Conventionally, there are some widelyused empirical predictors for estimation of theblast-induced ground vibrations. The USBureau of Mines proposed the first groundvibration predictor (Duvall et al., 1959).Subsequently, other empirical predictors wereproposed (Langefors and Kihlstrom, 1963;Ambraseys and Hendron, 1968; Ghosh andDaemen; 1983; Pal Roy, 1993). These methodsconsider two main input parameters –maximum charge used per delay and distancebetween the blast face and the monitoringpoints. Despite the simplicity and fastapplication of these methods, several recentstudies have shown their shortcomings inrendering acceptable predictions (Khandelwaland Singh, 2007). More recently, Chen andHuang (2001) conducted a seismic survey topredict blast-induced vibrations and PPVempirically. Ozer et al. (2008) examined theresults of some 500 blasts in a limestonequarry in Turkey for an experimental analysisof PPV. Ak et al. (2009) performed a series ofground vibration tests in a surface mine inTurkey in order to measure PPV. Aldas (2010)proposed an empirical relationship between theexplosive charge mass and PPV. Deb and Jha(2010) examined the effects of surfaceblasting on adjacent underground workings,using PPV measurements. Mesec et al. (2010)proposed an empirical relationship betweenPPV and distance for a series of vibration testsin some sedimentary rock deposits, comprising

Peak particle velocity prediction usingsupport vector machines: a surfaceblasting case studyby S.R. Dindarloo*

SynopsisAlthough blasting is one of the most widely used methods for rockfragmentation, it has a major disadvantage in that it causes adjacentground vibrations. Excessive ground vibrations can cause a wide range ofproblems, from nearby residents complaining to ecological damage.Prediction of blast-induced ground vibration is essential for evaluatingand controlling the many adverse consequences of surface blasting. Sincethere are several effective variables with highly nonlinear interactions, nocomprehensive model of blast-induced vibrations is available. In thisstudy, the support vector machine (SMV) algorithm was employed forprediction of the peak particle velocity (PPV) induced by blasting at asurface mine. Twelve input variables in three categories of rock mass,blast pattern, and explosives were used for prediction of the PPV atdifferent distances from the blast face. The results of 100 experimentswere used for model-building, and 20 for testing. A high coefficient ofdetermination with low mean absolute percentage error (MAPE) wasachieved, which demonstrates the suitability of the algorithm in this case.The very high accuracy of prediction and fast computation are the twomajor advantages of the method. Although the case study was for a largesurface mining operation, the methodology is applicable to all othersurface blasting projects that involve a similar procedure.

Keywordsblast-induced ground vibration, peak particle velocity, support vectormachine, surface mining.

* Department of Mining and Nuclear Engineering,Missouri University of Science and Technology,Rolla, MO, USA.

© The Southern African Institute of Mining andMetallurgy, 2015. ISSN 2225-6253. Paper receivedDec. 2014 and revised paper received March 2015.



Peak particle velocity prediction using support vector machines

mainly limestone and dolomite. Nateghi (2011) examined theeffects of different rock formations, different detonators, andexplosives on ground vibrations induced by blasting at a damsite.

Generally, empirical methods have two major limitations:lack of generalizability and limited number of input variables.Some researchers have proposed theoretical models based onthe physics of blasting. For instance, Sambuelli (2009)proposed a theoretical model for prediction of PPV on thebasis of some blast design and rock parameters. However,because of the complicated nature of the blasting process andits highly nonlinear interaction with the non-homogeneousand non-isotropic ground, a closed form mathematical modelis almost impossible. Recently, following the rapid growth insoft computing methods, including artificial intelligence,several researchers have tried to benefit from these newlyemerging techniques. In this category, artificial neuralnetworks (ANNs) might be the most widely used method forprediction of the ground vibrations. ANNs are among thetechniques that map input variables into the output(s). Thetechnique is capable of handling extremely nonlinearinteractions between different variables through assigningand adjusting proper weights. However, no functionalrelationship is proposed (‘black-box’ modelling). Khandelwaland Singh (2006) used ANNs for prediction of PPV in a largemine in India. Iphar et al. (2008) employed an adaptiveneural-fuzzy inference system (ANFIS) for prediction of PPVin a mine in Turkey. Dehghani and Ataee-pour (2011)employed ANNs for prediction of PPV in a large open pitcopper mine. Monjezi et al. (2011) used ANNS to predictblast-induced ground vibrations in an underground project.Bakhshandeh et al. (2012) used ANNs to adjust burden,spacing, and total weight of explosive used in order tominimize PPV.

The support vector machine (SVM) is a relatively newcomputational learning method for solving classification andnonlinear function estimation, which is based on statisticallearning theory. The SVM has been adopted rapidly by manyresearchers in different fields of geology, geotechnical, andenvironmental engineering (Brenning 2005; Yu et al., 2006;Samui 2008; Mountrakis et al., 2011; Dindarloo, 2014).Experimental results have revealed the superior performance ofSVMs with respect to other techniques. The reasons behind thesuccessful performance of SVMs, compared to other powerfulapproaches like ANNs, are twofold. Firstly, rather than beingbased on empirical risk minimization (ERM) as ANNs, whichonly minimizes the training errors, a SVM makes use ofstructural risk minimization (SRM), which seeks to minimizean upper bound on the generalization error. Secondly, findinga SVM solution corresponds to dealing with a convex quadraticoptimization problem. Thus, the Karush-Kuhn-Tucker (KKT)statements determine the necessary and sufficient conditionsfor a global optimum (Scholkoff and Smola 2002). For ANNs,however, it is not guaranteed that even a well-selectedoptimization algorithm will achieve the global minimum infinite computation time (Moura et al., 2011).

In this study, the SVM was used for analysis of the blast-induced ground vibration by prediction of PPV. A large ironore mine in Iran was selected as a case study. After obtainingdifferent input variables, a SVM model was constructed andtested.

MethodsDeveloped by Boser, Guyon, and Vapnik (Boser, Guyon, andVapnik, 1992; Vapnik, 1995, 1998), support vector machine(SVM) is a relatively new computational learning method forsolving classification and nonlinear function estimation,which is based on statistical learning theory. SVM is based onVapnik-Chervonenkis theory (VC theory), which recentlyemerged as a general mathematical framework for estimating(learning) dependencies from finite samples. This theorycombines fundamental concepts and principles related tolearning, well-defined formulation, and self-consistentmathematical theory. Moreover, the conceptual framework ofVC theory can be used for improved understanding of variouslearning methods developed in statistics, neural networks,fuzzy systems, signal processing, etc. (Widodo and Yang,2007).

LIBSVM is a library of SVM algorithms (Chang and Lin,2011) that was used along with Rapidminer, a data mining(DM) software package (Hofmann and Klinkenberg, 2013).The theory of SVM regression, used in LIBSVM, is presentedin the following section.

Support vector regressionConsider a set of training points, {(x1, z1), . . . , (xL, zL)},where xi ∈ Rn is a feature vector and zi ∈ RL is the targetoutput. Under given parameters C > 0 and ∈ > 0, the standardform of the support vector regression (SVR) (Equation [1])with constraints (Equations [2]-[4]) are as follows (Changand Lin, 2011):

[1]

subject to

[2]

[3]

[4]The dual problem (Equation [5]) is

[5]

subject to constraints (Equations [6]-[7])

[6]

[7]

where

[8]

After solving Equation [5], the approximate function is:

▲


[9]

The nomenclature is presented in Table I. For moredetailed information about the theory and applications ofSVR, see Burges (1998), Müller et al. (2001), Hsu and Lin(2002), Chapelle et al. (2002), and Smola and Scholkoff(2004).

Case studyGolegohar iron ore mine is located in southern Iran, 50 kmfrom Sirjan, in the southwest of Kerman Province. This ironore complex includes six known orebodies and is one of thelargest producers and exporters of iron concentrate in thecountry. Mining is by open pit methods, and the measuredand indicated reserves of over 1.1 billion tons of ore. TheGolegohar deposits are situated in a metamorphic complex ofprobable Paleozoic age with a northwest-southeast trend,known as the Sanandaj-Sirjan zone, which is parallel to theZagros thrust belt on the southwest and is bounded on thenortheast by the Urmieh-Dukhtar volcanic belt (Moxham andMcKee, 1990). The deposits are considered to be ofsedimentary or volcano-sedimentary origin, laid down indeltaic or near-shore environments that resulted in abruptlateral and vertical changes in the sedimentary facies.Subsequent deep burial, folding, metamorphism, and erosionleft a group of folded or down-faulted magnetite-rich depositsas elongated remnants of an iron formation that originallyhad a broader, perhaps more continuous extent. The mine’smetamorphic rocks consist mostly of gneiss, mica schist,amphibolite, quartz schist, marble, dolomite, and calcite(Karimi Nasab et al., 2011). Figure 1 illustrates one of theoperating pits. The geometry and slope stability factors of themine are summarized in Table II.

Parameter selectionRock mass, blast pattern and explosives, and distance fromthe face are the three major parameters in blast-inducedground vibrations, and hence the measured PPV. Thedominant rock types at Golegohar include amphibolite schist,quartz schist, chlorite schist, haematite, and magnetite.Density (t/m3), Young’s modulus (Gpa), uniaxialcompression strength (Mpa), and tensile strength (Mpa) ofrepresentative samples of all the rock types were measured inthe rock and soil mechanics laboratory at the mine site (Table IIIa). The major discontinuities have a significantinfluence on blast wave propagation in the rock mass. The



Table I

SVR notationsb InterceptC A parameter representing the compromise between

machine capacity and training errorw Weight vectorϕ Mapping functionα Function parameterQ Regression functionβ,β* Slack variablesK Kernel functionl Number of observations

Table II

Geometric parameters of pit No.1, Golegohar.Final wall slopes in ore and waste 45 degreesSlopes in overburden 38 degreesSafety bench height 30 mSafety bench width 10 mSafety bench slope 65 degreesWorking bench height 15 m

Figure 1 – Open pit mining at Golegohar (CNES/Astrium image on Google Earth, 29°05’15.21” N and 55° 19’ 03.24” E. Retrieved 3 April 2015)


spacing, dip, and direction of the two major joint sets arepresented in Table IIIb (Dindarloo et al., 2015). The secondgroup of important parameters is related to the drillingpattern and explosives used. A typical production, buffer, andpre-split pattern are illustrated in Figure 2. The mainexplosive is ANFO, and a blast delay of 15–75 ms betweenrows is used. The descriptive statistics of the patterngeometry, including burden, spacing, hole depth to burdenratio, specific charge, and stemming are presented in Table IV. Thus, the 12 input variables include: density,Young’s modulus, uniaxial compression strength, tensilestrength, joint spacing, burden, spacing, hole depth to burden

ratio, specific charge, stemming, delay per row, and distancebetween the measurement point and the blasting face. Sincethe main charge for all holes was ANFO, the parameter fortype of explosive was omitted, as it was the same for all tests.

Results and discussionsOne hundred and twenty experiments were conducted atdifferent distances, 15 m to 7500 m, from the blasting faces.The PPV was measured using the procedures described byDowding (1992). One hundred data-sets, including the 12input variables and one output (PPV), were used in the SVRmodel. The results of 20 randomly selected experiments were

▲


Table III

Mechanica l and physical properties. (a) Intact rock, (b) discontinuities

Petrology: (a) Intact rocks

Item Amphibolite schist Quartz schist Chlorite schist Haematite Magnetite

Density (t/m3) Mean 2.81 2.69 2.84 4.02 4.41Range 2.76-3.02 2.63-2.84 2.76-2.95 3.65-4.35 4.15-4.62

Young’s modulus (Gpa) Mean 34.8 52.7 37.6 29.7 42.6Range 19.6-47.1 18.6-77.3 15.7-40.3 14.9-41.2 33-55.9

Uniaxial compressive strength (Mpa) Mean 42.8 112.5 105.9 66.8 121.4Range 18.6-77.3 35.2-176.2 33.7-155.1 30.8-114.8 35.2-176.2

Tensile strength (Mpa) Mean 15.4 7.54 13.47 6.95 9.24Range 12.1-17.8 6.99-9.42 8.24-18.42 4.63-10.52 5.5-14.62

(b) DiscontinuitiesMajor joints Spacing (m) Dip (degree) Direction

Set 1 1.1 45 Northeast-southwest

Set 2 0.8 75 North-south

Figure 2 – Blast pattern (red: pre-splitting hole, yellow: ANFO, brown: stemming/crushed rock, white: no stemming/charging). Distances are in metres, andangles in degrees

used for model testing. Figure 3 depicts a scattergram of thepredicted SVR versus the measured PPVs for the 20 testingdata-sets. The coefficient of determination (Equation [10]),root mean squared error (RMSE, Equation [11]), and meanabsolute percentage error (MAPE, Equation [12]) were usedas the statistical metrics for evaluation of the SVR model(Table V). The obtained R2 value of 0.99 shows a very goodcorrelation between the predicted and measured PPVs. Theobtained MAPE value of less than 10% demonstrates thehigh accuracy and applicability of the method in PPVestimation, using the 12 input variables.

[10]

[11]

[12]

whereymeas and ypred are the observed and predicted values,respectivelyymeas and ypred and are mean observed and predictedvalues, respectively.

Sensitivity analysisIn order to analyse the effect of each individual variable on theSVM prediction accuracy, a sensitivity analysis was performed.

The optimized SVM parameters were kept the same for twelvesensitivity analysis runs. In each run, one of the inputvariables was omitted and its effect on prediction accuracy wasexamined. The results showed that omission of distance,specific charge, delay per row, and joints spacing had thehighest negative effects on SVM predictions. Hence the methodis more sensitive to these variables. The results of sensitivityanalysis for other variables are shown in Figure 4.

Comparison with traditional methodsThe partial least-square regression (PLSR) method is mainlyused for modelling linear regression between multipledependent variables and multiple independent variables. Anadvantage of this method over linear and nonlinear multipleregressions is that PLSR combines the basic functions ofregressing models, principal component analysis, andcanonical correlation analysis (Zhang et al., 2009). Inaddition, PLSR avoids the harmful effect of multi-collinearityand regressing when the number of observations is less thanthe number of variables. In the context of linear MR, theleast-squares solution for Equation [13] is given by Equation [14].



Table V

Statistics of SVM in PPV prediction

R2 MAPE (%) RMSE (mm/s)

0.99 8.5 3.45

Figure 4 – Sensitivity analysis

Table IV

Descriptive statistics of the collected data.

No Parameter Symbol Unit Min. Max. Mean DS

1 Burden B metre 3.83 5.88 4.81 0.682 Spacing S metre 4.37 7.11 6.14 0.913 Hole depth – H/B 2.04 4.44 3.40 0.62

burden ratio4 Stemming ST metre 3.86 7.95 5.19 0.795 Powder factor PF kg/t 0.21 0.47 0.32 0.07

Figure 3 – SVM predicted vs. measured PPV (mm/s)


Y = XB + ε [13]

B = (XTX)–1XTY [14]

Often, the problem is that XTX is singular, either becausethe number of variables (columns) in X exceeds the numberof objects (rows), or because of collinearities. PLSRcircumvents this by decomposing X into orthogonal scores(T) and loadings (P) (Mevik and Wehrens, 2007):

X = TP [15]

Furthemore, PLSR regresses Y, not on X, but on the firstα columns of the scores. The goal of PLSR is to incorporateinformation on both X and Y in the definition of the scoresand loadings. The scores and loadings are chosen in such away to describe as much as possible of the covariancebetween X and Y.

The result of the prediction of PPV by the PSLR techniqueis illustrated in Figure 5. Statistics of the predictions, for thesame testing data-set as SVM, are summarized in Table VI.The R2 value in PLSR decreased to 94% (i.e., the PLSR canmodel 94% of the variability in PPV based on the 12independent variables). Furthermore, both the obtainedRMSE and MAPE values in PLSR (see Table VI) were poorerthan the SVM (see Table V).

ConclusionsBlast-induced ground vibration control is a major challengein construction projects that employ blasting. Peak particlevelocity (PPV) is a widely used metric for evaluation of themagnitude and severity of the possible inconvenience topeople and damage to adjacent structures and theenvironment. This study demonstrates that the support vectormachine (SVM) approach is a versatile tool for prediction ofPPV based on the 12 input variables used. The very highaccuracy of prediction and fast computation are the two majoradvantages of the method. Results of the sensitivity analysisdemonstrated the considerable effect of distance, specificcharge, delay per row, and joint spacing on PPV. Thus, inspecific instances where the level of PPV is higher than a pre-specified threshold, appropriate remedies can be applied.Modification of the specific charge and the amount of delayper row are expected to have direct effects on PPV reduction.Although the SVM was used in a large surface mining casestudy, it is applicable to all other surface blasting projectswith a similar procedure.

AcknowledgementsWe would like to thank two anonymous reviewers for theircritical reviews and constructive comments. Golegohar minemanagement and staff are acknowledged for their support.

ReferencesALDAS, G.G.U. 2010. Explosive charge mass and peak particle velocity (PPV)-

frequency relation in mining blast. Journal of Geophysics and Engineering,vol. 7, no. 3. pp. 223–231.

AK, H., IPHAR, M., YAVUZ, M., and KONUK, A. 2009. Evaluation of groundvibration effect of blasting operations in a magnesite mine. Soil Dynamicsand Earthquake Engineering, vol. 29, no. 4. pp. 669–676.

AMBRASEYS, N.R. and HENDRON, A.J. 1968. Dynamic Behaviour of Rock Masses:Rock Mechanics in Engineering Practices. Wiley, London.

BAKHSHANDEH AMNIEH, H., SIAMAKI, A., and SOLTANI, S. 2012. Design of blastingpattern in proportion to the peak particle velocity (PPV): artificial neuralnetworks approach. Safety Science, vol. 50, no. 9. pp. 1913–1916.

BOSER, B., GUYON, I., and VAPNIK, V. 1992. A training algorithm for optimalmargin classifiers. Proceedings of the Fifth Annual Workshop onComputational Learning Theory. ACM, Pittsburgh. pp. 144–152.

BRENNING, A. 2005. Spatial prediction models for landslide hazards: review,comparison and evaluation. Natural Hazards and Earth System Science,vol. 5, no. 6. pp. 853–862.

BURGES, C.J.C. 1998. A tutorial on support vector machines for patternrecognition. Data Mining and Knowledge Discovery, vol. 2, no. 2. pp.121–167.

CHANG, C.-C. and LIN, C.-J. 2011. LIBSVM: a library for support vector machines.ACM Transactions on Intelligent Systems and Technology, vol. 2, no. 3.Article no. 27.

CHAPELLE, O., VAPNIK, V., BOUSQUET, O., and MUKHERJEE, S. 2002. Choosingmultiple parameters for support vector machines. Machine Learning, vol.46, no. 1-3. pp. 131–159.

CHEN, G. and HUANG, S.L. 2001. Analysis of ground vibrations caused by openpit production blasts - a case study. FRAGBLAST, vol. 5, no. 1–2. pp.91–107.

DINDARLOO, S.R. 2014. Support vector machine regression analysis of LHDfailures. International Journal of Mining, Reclamation and Environment.[doi: 10.1080/17480930.2014.973637].

DINDARLOO, S.R., ASKARNEJAD, N-A., and ATAEI, M. 2015. Design of controlledblasting (pre-splitting) in Golegohar iron ore mine, Iran. Transactions ofthe Institution of Mining and Metallurgy, Section A: Mining Technology,vol. 124, no. 1. pp. 64–68.

DEB, D. and JHA, A.K. 2010. Estimation of blast induced peak particle velocity atunderground mine structures originating from neighbouring surface mine.Transactions of the Institution of Mining and Metallurgy, Section A:Mining Technology, vol. 119, no. 1. pp. 14–21.

DEHGHANI, H. and ATAEE-POUR, M. 2011. Development of a model to predictpeak particle velocity in a blasting operation. International Journal of RockMechanics and Mining Sciences and Geomechanics Abstracts, vol. 48, no.1. pp. 51–58.

DOWDING, C.H. 1992. Suggested method for blast vibration monitoring.International Journal of Rock Mechanics and Mining Sciences andGeomechanics Abstracts, vol. 29, no. 2. pp. 145–156.

▲


Table VI

Statistics of PLSR in PPV prediction

R2 MAPE (%) RMSE (mm/s)

0.94 16.7 8.43

Figure 5 – PLSR predicted vs. measured PPV (mm/s)

DUVALL, W.I. and PETKOF, B. 1959. Spherical propagation of explosion generatedstrain pulses in rock. Report of Investigation 5483. US Bureau of Mines.21 pp.

GAD, E.F., WILSON, J.L., MOORE, A.J., and RICHARDS, A.B. 2005. Effects of mineblasting on residential structures. Journal of Performance of ConstructedFacilities, vol. 19, no. 3. pp. 222–228.

GHOSH, A. and DAEMEN, J.K. 1983. A simple new blast vibration predictor.Proceedings of the 24th US Symposium on Rock Mechanics, CollegeStation, Texas, USA, 20–23 June 1983. pp. 151–161.

HOFMANN, M. and KLINKENBERG, R. 2013. RapidMiner: Data Mining Use Casesand Business Analytics Applications. Chapman & Hall/CRC Data Miningand Knowledge Discovery Series. CRC Press, Boca Raton, FL.

HSU, C-W. and LIN, C-J. 2002. A comparison of methods for multiclass supportvector machines. IEEE Transactions on Neural Networks, vol. 13, no. 2.pp. 415–425.

IPHAR, M., YAVUZ, M., and AK, H. 2008. Prediction of ground vibrationsresulting from the blasting operations in an open-pit mine by adaptiveneuro-fuzzy inference system. Environmental Geology, vol. 56, no. 1. pp.97–107.

KAHRIMAN, A. 2001. Prediction of particle velocity caused by blasting for aninfrastructure excavation covering granite bedrock. Mineral ResourcesEngineering, vol. 10, no. 2. pp. 205–218.

KARIMI NASAB, S., HOJAT, A., KAMKAR-ROUHANI, A., AKBARI JAVAR, H., andMAKNOONI, S. 2011. Successful use of geoelectrical surveys in Area 3 of theGol-e-Gohar iron ore mine, Iran. Mine Water and the Environment, vol.30, no. 3. pp. 208–215.

KHANDELWAL, M. and SINGH, T.N. 2006. Prediction of blast induced groundvibrations and frequency in opencast mine: a neural network approach.Journal of Sound and Vibration, vol. 289, no. 4–5. pp. 711–725.

KHANDELWAL, M. and SINGH, T.N. 2007. Evaluation of blast-induced groundvibration predictors. Soil Dynamics and Earthquake Engineering, vol. 27,no. 2. pp. 116–125.

LANGEFORS, U. and KIHLSTROM, B. 1963. The Modern Technique of RockBlasting. Wiley, New York.

MESEC, J., KOVAČ, I., and SOLDO, B. 2010. Estimation of particle velocity based onblast event measurements at different rock units. Soil Dynamics andEarthquake Engineering, vol. 30, no. 10. pp. 1004–1009.

MEVIK, B-H. and WEHRENS, R. 2007. The pls package: principal component andpartial least squares regression in R. Journal of Statistical Software, vol.18, no. 2. pp. 1–23.

MONJEZI, M., GHAFURIKALAJAHI, M., and BAHRAMI, A. 2011. Prediction of blast-induced ground vibration using artificial neural network. Tunnelling andUnderground Space Technology, vol. 26, no. 1. pp. 46–50.

MOUNTRAKIS, G., IM, J., and OGOLE, C. 2011. Support vector machines in remotesensing: a review. ISPRS Journal of Photogrammetry and Remote Sensing,vol. 66, no. 3. pp. 247–259.

MOURA, M.D.C., ZIO, E., LINS, I.D., and DROGUETT, E. 2011. Failure and reliabilityprediction by support vector machines regression of time series data.Reliability Engineering and System Safety, vol. 96, no. 11. pp.1527–1534.

MÜLLER, K-R., MIKA, S., RÄTSCH, G., TSUDA, K., and SCHÖLKOPF, B. 001. Anintroduction to kernel-based learning algorithms. IEEE Transactions onNeural Networks, vol. 12, no. 2. pp. 181–201.

NATEGHI, R. 2011. Prediction of ground vibration level induced by blasting atdifferent rock units. International Journal of Rock Mechanics and MiningSciences and Geomechanics Abstracts, vol. 48, no. 6. pp. 899–908.

NATEGHI, R., KIANY, M., and GHOLIPOURI, O. 2009. Control negative effects ofblasting waves on concrete of the structures by analyzing of parameters ofground vibration. Tunnelling and Underground Space Technology, vol. 24,no. 6. pp. 608–616.

OZER, U., KAHRIMAN, A., AKSOY, M., ADIGUZEL, D., and KARADOGAN, A. 2008. Theanalysis of ground vibrations induced by bench blasting at Akyol quarryand practical blasting charts. Environmental Geology, vol. 54, no. 4. pp.737–743.

PAL ROY, P., 1993. Putting ground vibration predictors into practice. CollieryGuardian, vol. 241. pp. 63–67.

SAMBUELLI, L. 2009. Theoretical derivation of a peak particle velocity-distancelaw for the prediction of vibrations from blasting. Rock Mechanics andRock Engineering, vol. 42, no. 3. pp. 547–556.

SAMUI, P. 2008. Slope stability analysis: a support vector machine approach.Environmental Geology, vol. 56, no 2. pp. 255–267.

SCHOLKOFF, B. and SMOLA, A.J. 2002. Learning with kernels—support vectormachines, regularization, optimization, and beyond. MIT Press, CambridgeMA.

SINGH, P.K., VOGT, W., SINGH, R.B., SINGH, M.M., and SINGH, D.P. 1997. Responseof surface structures to rock blasting. Mineral Resources Engineering, vol.6, no. 4. pp. 185–194.

SINGH, P.K., ROY, M.P., and SINGH, R.K. 2005. Responses of roof and pillars ofunderground coal mines to vibration induced by adjacent open-pitblasting. Environmental Geology, vol. 47, no. 2. pp. 205–214.

SMOLA, A.J. and SCHOLKOFF, B. 2004. A tutorial on support vector regression.Statistics and Computing, vol. 14. pp. 199–222.

VAPNIK, V. 1995. The Nature of Statistical Learning Theory. Springer-Verlag,New York.

VAPNIK, V. 1998. The support vector method of function estimation. NonlinearModeling: Advanced Black-box Techniques. Suykens, J.A.K. andVandewalle, J. (eds). Kluwer Academic Publishers, Boston. pp.55–85.

VILLAESCUSA, E., ONEDERRA, I., and SCOTT, C. 2004. Blast induced damage anddynamic behaviour of hangingwalls in bench stoping. Fragblast, vol. 8,no. 1. pp. 23–40.

WIDODO, A. and YANG, B-S. 2007. Support vector machine in machine conditionmonitoring and fault diagnosis. Mechanical Systems and SignalProcessing, vol. 21, no. 6. pp. 2560–2574.

YI, C-P. and LU, W-B. 2006. Research on influence of blasting vibration ongrouted rockbolt. Yantu Lixue/Rock and Soil Mechanics, vol. 27, no. 8.pp. 1312–1316.

YU, P-S., CHEN, S-T., and CHANG, I-F. 2006. Support vector regression for real-time flood stage forecasting. Journal of Hydrology, vol. 328, no. 3–4. pp.704–716.

ZHANG, M., MU, H., LI, G., and NING, Y. 2009. Forecasting the transport energydemand based on PLSR method in China. Energy, vol. 34, no. 9. pp.1396–1400. ◆



The Jameson Cell was developed in the late 1980s to overcome the design and operating inadequacies of column and conventional flotation cells.

© Copyright 2015GLT2599_05/15

www.glencoretechnology.com

To find out how Jameson Cell technology can improve your operation contact us at [email protected]

“Glencore technology is not just an equipment supplier. Our clients get more than just a flotation cell, they become part of a technology partnership which includes not only coal users but process engineers and designers where over two decades of flotation knowledge and expertise is transferred to your operation.”– Le Huynh,

Jameson Cell Technology

Manager, Glencore Technology.

The first commercial base metals

installation at Mount Isa Mines

in 1989 was followed quickly by

the first coal installation at the

Newlands operation in the coking

coal rich Bowen Basin region of

Central Queensland, Australia in

1990. In the two decades since its

development there are now over

160 Jameson Cell installations in

coal wash plants around the world,

including emerging regions such as

Mozambique and Mongolia.

Simplicity, small footprint,

excellent availability, and very

low maintenance in addition to

consistent process performance

make Jameson Cells the equipment

of choice for coal operations.

From its early development of very

small machines treating 10 tonnes

per hour (tph) of flotation feed, a

single Jameson Cell can now treat

up to 100 tph using a single unit.

With the machine’s consistent fine

bubble generation making it the

highest capacity cell in the market,

coupled with the availability of a

wide number of standard cell sizes,

there is a machine to meet your

design for any size plant.

The Jameson Cell has been the leader in coal flotation for over two decades making it

the obvious choice for your wash plant.

IntroductionSqueezing ground conditions are encounteredin several underground hard-rock mines andcan result in large-scale deformation andground instability. The International TaskForce on Squeezing Rock in Australian andCanadian Mines reported that ‘in a miningenvironment squeezing ground conditionswere identified as closure larger than tencentimeters over the life expectancy of asupported drive’ (Potvin and Hadjigeorgiou,2008). Mine drives are usually in operationfrom 18 months to two years. Squeezingground conditions in mines are associated withconsiderable failure of ground support andrequire significant rehabilitation work. Thismay cause a slowdown in development andcan have severe economic repercussions.

In deep and high-stress mines, squeezingground conditions are driven by the presenceof inherent foliation and the orientation of thedrift walls with respect to the foliation. Failurein bedded rock masses has been studiedthough physical modelling by Lin et al.(1984), and analytical methods by Kazakidis(2002). Potvin and Hadjigeorgiou (2008)reviewed ground support strategies used tocontrol large-scale rock mass deformationunder squeezing conditions in mines. Despitecertain differences in the support philosophies

between Australian and Canadian mines, itwas evident that an effective support systemmakes use of both reinforcement elements andsurface support. A successful systemreinforces the rock mass around theexcavation and mitigates the rate ofdeformation. Experience has shown thatductile surface support is an essential part of asuccessful ground support system.

In a mining environment, squeezingground conditions are defined as those thatexhibit strain higher than 2%. Potvin andHadjigeorgiou (2008) observed that largedeformations are generally associated with thepresence of a prominent structural feature suchas intense foliation, a dominant structuralfeature or a shear zone, and high stress inweak rock. The presence of joint alteration andmineralogy further increases the severity ofsqueezing.

Mercier-Langevin and Hadjigeorgiou(2011) presented a ‘hard rock squeezingindex’ for underground hard-rock mines basedon several mining case studies in Australiaand Canada and calibrated against in-situobservations at the LaRonde mine in Quebec.The authors proposed the use of the index as apreliminary indicator of the squeezingpotential in hard-rock mines with similarground conditions.

Case studies in underground hard-rockminesWhile many hard-rock mines around the worldface problems associated with squeezingground conditions, there are only a few casestudies documented (Struthers et al., 2000;Beck and Sandy, 2003; Potvin and Slade,2007; Sandy et al., 2010; Mercier-Langevin

Large-scale deformation in undergroundhard-rock minesby E. Karampinos*, J. Hadjigeorgiou*, P. Turcotte†, and F. Mercier-Langevin†

SynopsisIn some underground hard-rock mines, squeezing compressive groundconditions are influenced by the presence of rock foliation and high stress.In these cases, the orientation of the foliation with respect to the driftdirection has a considerable impact on the magnitude of the resultingdeformation. Irrespective of the reinforcement and support strategy,keeping drives developed sub-parallel to the rock foliation operational isdifficult, and often requires excessive rehabilitation during the lifetime ofthe excavation. This study uses field observations and convergencemeasurements at the LaRonde and Lapa mines of Agnico Eagle Mines Ltdto provide guidelines of the anticipated squeezing levels at theseoperations.

KeywordsSqueezing ground conditions, large deformations, foliation, angle of driftinterception, design guidelines.

* University of Toronto, Ontario, Canada.† Agnico Eagle Mines, Toronto, Ontario, Canada.© The Southern African Institute of Mining and

Metallurgy, 2015. ISSN 2225-6253. Paper receivedNov. 2014 and revised paper received Apr. 2015.



Large-scale deformation in underground hard-rock mines

and Hadjigeorgiou, 2011; Mercier-Langevin and Wilson,2013; Hadjigeorgiou et al., 2013). Karampinos et al. (2014)presented a methodology for modelling the behaviour offoliated rock masses under high stress conditions using a 3Ddistinct element code. The resulting models addressedexplicitly the effect of foliation and reproduced the observedbuckling mechanism. Vakili et al. (2012) used a 3D distinctelement code as a prelude to a 2D analysis using a finitedifference code.

Table I summarizes the reported level of deformationfrom several hard-rock mines. The LaRonde and Lapa minesreport some of the highest deformations in hard-rock mines.This necessitates excessive rehabilitation of affected drifts.This study focuses on the LaRonde and Lapa mines as theyalso experience a large spectrum of squeezing groundconditions. Both mines are situated in the Abitibi region ofnorthwest Quebec, within 11 km from each other, and areoperated by Agnico Eagle Mines Ltd.

LaRonde exploits a world-class Au-Ag-Cu-Zn massivesulphide lens complex. The ore reserves extend from surfaceto 3110 m and are still open at depth. The mine, which hasbeen in operation since 1988, uses two mining methods –longitudinal retreat with cemented backfill, and transverseopen stoping with cemented and unconsolidated backfill. Thedeepest production horizon is currently at 2930 m,established after the construction of an 832 m internal shaft.The mine is operating in a variety of ground conditions. Atdifferent levels of the mine the observed rock mass behaviourcan be hard and brittle or squeezing. The mine reports thatthe total wall convergence in certain areas can be in excess of1 m, with fracturing extending up to 6 m into the rock mass.

Lapa is a high-grade gold mine and has been in operationsince 2009. Access to the mine is provided by a 1369 m deepshaft and production is by two mining methods – longitudinalretreat with cemented backfill, and locally transverse openstoping with cemented backfill. The mine operates under

challenging squeezing ground conditions. Hadjigeorgiou etal. (2013). Mercier-Langevin and Wilson (2013) provided aninterpretation of the squeezing mechanisms at Lapa and thesupport strategies aimed at controlling large deformations.

Mitigating the degree of squeezing

Ground supportSignificant differences in the ground support strategiesfollowed by Australian and Canadian hard rock mines insqueezing ground conditions were reported by Potvin andHadjigeorgiou (2008). Australian mines often use softreinforcement elements such as split sets, complemented withfibre-reinforced shotcrete. The shotcrete increases thestiffness of the support system and can initially delay therock mass degradation. However, as shotcrete canaccommodate only limited rock mass deformation and cancrack, it is often necessary to install screen over shotcrete.This results in a stiff liner early in the squeezing process,followed by a ductile surface support after the installation ofthe screen. Canadian mines, on the other hand, use a highdensity of bolts with yielding capability (such as Swellex orhybrid bolts) and weldmesh, often accompanied with meshstraps.

In the past, frictions bolts were used at LaRonde withpartial success. The split sets ‘lock up’ between the foliationplanes when buckling occurs and fail at the contact betweenthe bolt and the plate when deformation surpasses a certainlevel. Mercier-Langevin and Turcotte (2007b) reportedlimited success in the use of cemented grouted cable bolts,yielding cable bolts, and modified cone bolts in squeezingconditions, as they either did not yield sufficiently or losttheir ability to yield early in the squeezing process. The minehas been more successful using the hybrid bolt (Mercier-Langevin and Turcotte, 2007b) as part of its ground supportstrategy. The advantages of the hybrid bolt were presented by

▲


Table I

Squeezing ground conditions reported in hard-rock mines

Mine site Strain range % (defined as total wall deformation Magnitude of deformationover the drift width)

Lower bound Upper bound

LaRonde (Hadjigeorgiou et al., 2013) 2.5% 41% 2.1 m (total wall convergence)Lapa (volcanics and ultramafics) 1% >40%(Mercier-Langevin and Wilson, 2013)Wattle Dam (Marlow and Mikula, 2013) 1% 5%Westwood (Armatys, 2012) Up to 8.5% 0.34 m (total wall convergence)Perseverance (Gabreau, 2007) 2.5 m total wall convergence(Potvin and Slade, 2007) Wall convergence > 2 m(Struthers et al., 2000) 3 m total sidewall closure, over 1 m of floor heaveYilgarn Star (Potvin and Slade, 2007) Up to 2 m in the hangingwallBlack Swan (Potvin and Slade, 2007) Up to 1.5 m in one wallMaggie Hayes 1% 2.5%(Mercier-Langevin and Hadjigeorgiou, 2011)Casa Berardi 1% 5%(Mercier-Langevin and Hadjigeorgiou, 2011)Waroonga 1% 5%(Mercier-Langevin and Hadjigeorgiou, 2011)Bousquet 1% 10%(Mercier-Langevin and Hadjigeorgiou, 2011)Doyon altered zone 2.5% 10%(Mercier-Langevin and Hadjigeorgiou, 2011)

Turcotte (2010) and include a setup that prevents the resinfrom escaping into the fractured rock. Furthermore, thehybrid bolt has a high resistance to shear and frictionalresistance as compared to split sets. The in-situ behaviour ofthe hybrid bolt is characterized by a stiff early reaction at lowdisplacements and almost perfectly plastic behaviour whensubjected to high load (approximately 15 t). The currentstandard employed at the LaRonde mine was presented byHadjigeorgiou et al. (2013) and consists of friction sets andscreen in the sidewalls and rebars and screen in the back,complemented by hybrid bolts in the sidewalls and meshedstraps installed 12 m away from the face.

The ground support system employed at Lapa wasinspired by the successful performance of the hybrid bolt atLaRonde (Mercier-Langevin and Wilson, 2013). The supportstrategy recognizes and accounts for the presence of weakschist zones (ultramafic) prone to squeezing. In areas wheresqueezing is anticipated, hybrid bolts are installed instead ofsplit set bolts.

When the sidewalls are subject to excessive deformation,the drives can become narrow and inadequate for the miningequipment. Under these circumstances, the walls are ‘purged’using a scoop to remove excess material. This is a costly andtime-consuming process, and is conducted only whennecessary and under the close supervision of ground controlpersonnel. Following purging of the excavation, additionalsupport such as cable bolts supplemented with screen andstraps is installed to further stabilize the walls. Turcotte(2010) reported a considerable reduction of purging since theintroduction of the hybrid bolt in LaRonde.

Mining under extreme squeezing ground conditions hasdemonstrated that it is not a realistic aim to completely arrestground deformation. This can result in early failure of thesupport and necessitate frequent rehabilitation. Currentpractice aims to control the resulting deformation undersqueezing conditions. Management of drive closure can beimproved through an understanding of the factors that defineand control the squeezing phenomenon.

Mercier-Langevin and Turcotte (2007a) demonstratedthat modifying the mining layout by driving drifts in a morefavourable direction with respect to the foliation resulted inless purging and rehabilitation. Although this necessitatedlonger development drives per level, it significantly reducedrehabilitation and production delays. This practice reducedthe likelihood of major ground instability in the drive.

Influence of drift orientationThe influence of drift orientation on the observed degree ofsqueezing is evident in several places at both Lapa andLaRonde. This was quantified from the angle of interception,defined as the angle between the normal to the foliationplanes and the normal to the sidewall (Figure 1). This isillustrated by three drifts that were driven at 2150 m depth atLaRonde (Figure 2). There is no evidence of squeezing for adrift developed perpendicular to the foliation, only minorsqueezing when driven at 45 degrees, and severe squeezingfor a drift oriented parallel to the foliation.

The angle of interception had a direct impact on theperformance of the ground support systems used at theLaRonde mine. An investigation of the effect of theorientation of the drift on the degree of squeezing wasinitially made by Mercier-Langevin (2005). It was based on

23 field observations from drifts at the LaRonde mine. Theoriginal database reported the observed squeezing level, thedamage to the support, the difference between the orientationof the drift and the foliation, and the influence of stress onthe resulting deformation. For the cases where the drifts weredeveloped sub-parallel to the foliation, regardless of thesupport system employed, it was difficult to keep the drifts inoperation unless they were subjected to regular rehabilitationwork (Mercier Langevin and Hadjigeorgiou, 2011). Thedegree of squeezing varied for drifts driven at differentangles with respect to the foliation.

This orientation phenomenon is supported by mechanisticanalysis. Auto-confinement of foliation planes is greaterwhen the angle between the normal to the free face and thenormal to the foliation increases (Hadjigeorgiou et al., 2013).It has been shown analytically that even a small confiningpressure is sufficient to prevent buckling failure (Kazakidis,2002).

Updated database for investigating the influence ofdrift orientationIn this work, the original database was extended usingquantitative data for drift closure from 57 new case studies atLaRonde and 87 from Lapa. For every case study thefollowing parameters were recorded: the dip and dip directionof foliation; the orientation of the drift; the observed damage;the development date; the stress effect due to mining activity;the support system used; the additional support installed; andthe presence of water. Any intervention such as rehabilitationor purging was also recorded.

The cavity monitoring survey (CMS) instrument CMSV400 (Optech Incorporated, 2010) was used to capture thewall profile. Figure 3 shows an example of multiple CMSreadings in a drift at Lapa. 3D surveys, showing the initialdrift dimensions after the development of every drift, are alsoavailable for both mines. These surveys are made bysurveying one point at the back, the floor, and each sidewallimmediately after the development of a drift.

The distance between the two sidewalls was extractedfrom the CMS, and recorded at heights of 1.5 and 2.5 m fromthe drift floor. The back–to-floor distance was estimated atthe centre of each drift and at 1 m on each side from thecentre. In cases where a CMS profile was not available,measurements were made using a laser measuring device. It


647The Journal of The Southern African Institute of Mining and Metallurgy VOLUME 115 JULY 2015

Figure 1 – Definition of angle of interception (ψ) (after Mercier-Langevinand Hadjigeorgiou, 2011)


was thus possible to determine the total sidewall and totalback–to-floor convergence for each case study.

The greatest convergence was derived from CMS readingsfor 35 case studies at LaRonde and 73 case studies at Lapa.The convergence was determined by comparing the initial 3Dsurvey results during development and the CMS profile after atime interval. Data collection was complicated by the frequentpresence of muck on the side of each drift. Figure 4 shows thevarious methods used to identify the convergence in each casestudy and the reference points along the drift.

Quantifying observed convergencePrevious analysis of the data at the time it was collectedfocused on a qualitative interpretation. As more quantitativedata was collected and more case studies were documented, itbecame possible to provide a preliminary quantitativeinterpretation (Hadjigeorgiou et al., 2013). The workpresented in this paper includes more case studies andfurther information that reports on back-to-floor

convergence, wall-to-wall convergence, and sidewalldeformation. The recorded convergence in the case studiespresented in the past was also updated.

For the purposes of this investigation the total wall-to-wall convergence (δtotal) was estimated from the differencebetween the surveyed width (L) and the lowest sidewalldistance measured. Similarly, the total back-to-floorconvergence was derived from the lowest back-to-floordistance and the height of the drift. The total wall-to-wall andback-to-floor convergences were expressed as percentages ofthe total strain (εtotal):

[1]

It is recognized that operational restrictions can influencethe data collection process. In particular, when the wall-to-wall closure is close to 3.5 m, the drift becomes non-operational for equipment and therefore it is purged.Consequently, a value of 3.5 m was used as the lowest wall-

▲


Figure 2 – Variations in squeezing severity in three locations less than 100 m apart at 2150 m depth, LaRonde mine. (a) Perpendicular – no squeezing; (b) 45degrees – minor or no squeezing; (c) parallel – severe squeezing (after Mercier-Langevin and Hadjigeorgiou, 2011)

Figure 3 – Example of multiple CMS in a longitudinal drift at Lapa, 540 m depth

to-wall distance measured in these cases. A revised classifi-cation scheme for quantifying squeezing in hard-rock minesis proposed:

➤ No or low squeezing (0% < ε < 5%)➤ Moderate squeezing (5% < ε < 10%)➤ Pronounced squeezing or rehabilitated drifts (10% < ε

< 35%)➤ Extreme squeezing (35% < ε).

Published work in civil engineering tunnellingapplications summarized by Potvin and Hadjigeorgiou (2008)reported considerably lower ranges than those proposed here.This is a result of the higher tolerance of rock mass failure ina mining environment. Typical squeezing examples observedin LaRonde and Lapa mines are presented in Figure 5. Theinfluence of the angle of interception (ψ) on the resultingtotal wall-to-wall and back-to-floor strain is shown in Figure6.

For comparison purposes, the convergence was alsoexamined for each wall separately. This was defined as theratio of the highest recorded convergence (δ) for each wall tohalf of the surveyed width (L) for the sidewalls or to half thesurveyed height for the back and the floor. The convergencefor each wall was expressed as percentage strain (ε):

[2]

The influence of the angle of interception (ψ) on the

resulting strain for each wall at the LaRonde and Lapa minesis presented in Figure 7 for the sidewalls and in Figure 8 forthe back and the floor. These diagrams capture only a part ofthe behaviour of a rock mass under squeezing groundconditions. There are further factors that can influence theresulting strain, such as the time of measurement, thefoliation spacing, the stresses, the strength of the rock, andthe condition of the joints. In addition, operationalconstraints do not allow for a drive with a wall-to-walldistance less than 3.5 m.

Figures 6 to 8 include a threshold of the highest expectedstrain for a given angle of interception. It is noted, however,that there is limited data for an angle of interception less than10 degrees in extreme squeezing conditions. The observedtrend is supported by recent numerical modelling work by theauthors. Nevertheless, the field data clearly indicates that anincrease in the angle of interception between the drift and theinherent foliation will invariably reduce the resulting level ofsqueezing.

The south walls demonstrated the highest strain,exceeding 50% in certain case studies. These values are alsohigher than the total sidewall strain recorded. The differencebetween the convergence on each sidewall (north walls andsouth walls) at Lapa was identified by Mercier-Langevin andWilson (2013). Higher strain, resulting in frequent rehabili-tation, was linked with the presence of ultramafics, whereaslower strain, easily managed, was associated with relatively



Figure 4 – Estimation of drift convergence


▲


Figure 6 – Influence of angle of interception (ψ) on resulting total strainat the LaRonde and Lapa mines. (a) Total wall-to-wall strain, (b) totalback-to-floor strain

Figure 7 – Influence of angle of interception (ψ) on resulting sidewallstrain at the LaRonde and Lapa mines. (a) South wall strain, (b) northwall strain

Figure 5 – Examples of drifts subjected to squeezing at the LaRonde and Lapa mines (Hadjigeorgiou et al., 2013)

competent sediments and volcanics. The ultramafics havetypically much smaller foliation spacing and talc-chloritealteration is usually present. The examined south walls atLapa were comprised of ultramafic formations while the northwalls were mostly driven in sediments.

Visual observations suggest that the strain in the northwalls at Lapa is lower than that recorded in many cases. Thestrain may be overestimated as sometimes the walls insediments follow the dip of the foliation (approximately 85°),which was not considered in the 3D survey. Higher strain canalso be a result of errors in the positioning of the CMS or anyshape irregularities on the wall, as the 3D survey considersthe wall as a plane surface.

The reported total sidewall and total back-to-floor strainhas significant practical implications for the functionality ofthe drifts and the need for rehabilitation to maintain them inoperation. The determination of the strain for each individualwall can allow for a more representative consideration of thegeological and mineralogical conditions that can influence thesqueezing level. Consequently, the influence of other factorscontrolling the degree of squeezing, such as the foliationspacing, the alteration, intact rock strength, and the stresscan be explored in greater detail.

Currently there is a lack of a ground support system thatcan fully control pronounced and extreme squeezing groundconditions. Consequently, exploring changes in theorientation of development can be an effective strategy inmining under such conditions. While this has been anopportunity in LaRonde, it is more difficult at Lapa, due tothe lower flexibility allowed by the mining method.

Hoek and Marinos (2000) used the ratio of the uniaxialcompressive strength (σcm) of the rock mass to theoverburden stress (po) to predict the extent of squeezing intunnelling, using a similar method to estimate the wall strain.This approach does not consider the anisotropic behaviour ofthe rock mass and the presence of any dominant structure. Asuccessful classification system for the prediction of the levelof squeezing at LaRonde and Lapa should take into accountthe influence of foliation and the angle of interception (ψ).

This study has demonstrated the need to better defineand capture the transition between the various squeezingzones. This can potentially be attained by combining fieldobservations with numerical studies.

ConclusionsThe LaRonde and Lapa mines exhibit large-scale deformationin a range of ground conditions. The estimation of the totalsidewall and the total back-to-floor strain indicated theproblems encountered in the functionality of the drifts andthe need for rehabilitation work when pronounced andextreme squeezing conditions are faced. An analysis of thereported strain for each drive wall revealed a strongcorrelation between the squeezing level and the geology.Variations in the geology at each side of a drift can result insignificantly more pronounced squeezing conditions. Underthese circumstances it is optimal to implement a differentground support standard for each wall.

Acceptable squeezing levels in a mining environment areconsiderably higher than in civil engineering tunnellingoperations. Although this allows more flexibility, miningoperators have to work under greater economic constraints interms of support. Squeezing ground conditions in miningapplications often involve considerable failure of groundsupport and necessitate significant rehabilitation work.

The LaRonde and Lapa mines follow similar groundsupport strategies for developing excavations in squeezingground conditions. The support systems aim to control theextreme deformation rather than prevent it, which is not arealistic objective in a mining context. This paper haspresented the results of extensive field work in the quantifi-cation of the influence of the angle of interception betweenthe drift and the foliation. The choice of a favourable angle ofinterception can result in a more manageable squeezing leveland increase the performance of an appropriate supportsystem for squeezing ground conditions. The results fromthis study are in agreement with the squeezing indexproposed by Mercier-Langevin and Hadjigeorgiou (2011) andcontribute towards its validation and extension.

AcknowledgementsThe support of Agnico Eagle Mines Ltd, Division LaRondeand Lapa, and the Natural Science and Engineering ResearchCouncil of Canada is gratefully acknowledged.



Figure 8 – Influence of angle of interception (ψ) on resulting back andfloor strain at the LaRonde and Lapa mines. (a) Back strain, (b) floorstrain


ReferencesARMATYS, M. 2012. Modification des classifications géomécaniques pour les massifs

rocheux schisteux. Master's thesis, École Polytechnique de Montréal, Canada.BECK, D.A. and SANDY, M.P. 2003. Mine sequencing for high recovery in Western

Australian mines. Proceedings of the Twelfth International Symposium onMine Planning and Equipment Selection, Kalgoorlie, 23–25 April 2003. [CD-ROM].

GAUDREAU, D. 2007. Current ground support practices at Perseverance mine.Challenges in Deep and High Stress Mining. Potvin, Y., Hadjigeorgiou, J., andStacey, D. (eds). Australian Centre for Geomechanics, Perth. pp. 363–369.

HADJIGEORGIOU, J., KARAMPINOS, E., TURCOTTE, P., and MERCIER-LANGEVIN, F. 2013.Assessment of the influence of drift orientation on observed levels ofsqueezing in hard rock mines. Proceedings of the Seventh InternationalSymposium on Ground Support in Mining and Underground Construction,Perth, Australia, 13–15 May 2013. Brady, B. and Potvin, Y. (eds). AustralianCentre for Geomechanics, Perth. pp. 109–118.

HOEK, E. and MARINOS, P. 2000. Predicting tunnel squeezing problems in weakheterogeneous rock masses. Tunnels and Tunnelling International, vol. 32,no. 11. pp. 45–51.

KARAMPINOS, E., HADJIGEORGIOU, J., TURCOTTE, P., DROLET, M-M., and MERCIER-LANGEVIN, F. 2014. Empirical and numerical investigation on the behaviour offoliated rock masses under high stress conditions. Proceedings of the SeventhInternational Conference on Deep and High Stress Mining, Sudbury, Canada,16-18 September 2014. Australian Centre for Geomechanics, Perth.

KAZAKIDIS, V.N. 2002. Confinement effects and energy balance analyses forbuckling failure under eccentric loading conditions. Rock Mechanics and RockEngineering, vol. 35, no 2. pp. 115–126.

LIN, Y., HONG, Y., ZHENG, S., and ZHANG, Y. 1984. Failure modes of openings in asteeply bedded rock mass. Rock Mechanics and Rock Engineering, vol. 17. pp.113–119.

MARLOW, P. and MIKULA, P.A. 2013. Shotcrete ribs and cemented rock fill groundcontrol methods for stoping in weak squeezing rock at Wattle Dam Gold Mine.Proceedings of the Seventh International Symposium on Ground Support inMining and Underground Construction, Perth, Australia, 13–15 May 2013.Brady, B. and Potvin, Y. (eds). Australian Centre for Geomechanics, Perth. pp.133–148.

MARLOW, P. and MIKULA, P.A. 2013. Shotcrete ribs and cemented rock fill groundcontrol methods for stoping in weak squeezing rock at Wattle Dam Gold Mine.Proceedings of the Seventh International Symposium on Ground Support inMining and Underground Construction, Perth, Australia, 13–15 May 2013.Brady, B. and Potvin, Y. (eds). Australian Centre for Geomechanics, Perth. pp.133–148.

MERCIER-LANGEVIN, F. 2005. Projet de convergence des galeries - Phase 1:Consolidation de l'information disponible à la mine. Internal memo, AgnicoEagle Mines Ltd - LaRonde Division, Cadillac, Canada. 13 pp.

MERCIER-LANGEVIN, F. and HADJIGEORGIOU, J. 2011. Towards a better understanding ofsqueezing potential in hard rock mines. AusIMM, Mining Technology, vol.120, no. 1. pp. 36–44.

MERCIER-LANGEVIN, F. and TURCOTTE, P. 2007a. Expansion at depth at Agnico Eagle’sLaRonde Division – meeting geotechnical challenges without compromisingproduction objectives. Challenges in Deep and High Stress Mining. Potvin, Y.,Hadjigeorgiou, J., and Stacey, D. (eds). Australian Centre for Geomechanics,Perth. pp. 189–195.

MERCIER-LANGEVIN, F. and TURCOTTE, P. 2007b. Evolution of ground support practicesat Agnico Eagle’s LaRonde Division-innovative solutions to high stressyielding ground. Rock Mechanics – Meeting Society’s Challenges andDemands. Proceedings of the 1st Canada–US Rock Mechanics. Symposium.Eberhardt, E., Stead, D., and Morrison, T. (eds). Taylor & Francis, Vancouver.pp. 1497–1504.

MERCIER-LANGEVIN, F. and WILSON, D. 2013. Lapa mine - ground control practices inextreme squeezing ground. Proceedings of the Seventh InternationalSymposium on Ground Support in Mining and Underground Construction,Perth, Australia, 13-15 May 2013. Brady, B. and Potvin, Y. (eds). AustralianCentre for Geomechanics, Perth. pp. 119–132 .

OPTECH INCORPORATED. 2010. CMS V400 Operation Manual, Cavity MonitoringSystem. Ontario, Canada. 126 pp.

POTVIN, Y. and HADJIGEORGIOU, J. 2008. Ground support strategies to control largedeformations in mining excavations. Journal of the Southern African Instituteof Mining and Metallurgy, vol. 108, no. 7. pp. 393–400.

POTVIN, Y. and SLADE, N. 2007. Controlling extreme ground deformation; Learningfrom four Australian case studies. Challenges in Deep and High StressMining. Potvin, Y., Hadjigeorgiou, J., and Stacey, D. (eds). Australian Centrefor Geomechanics, Perth. pp. 355–361.

SANDY, M., SHARROCK, G., and VAKILI, A. 2010. Managing the transition from lowstress to high stress conditions. Second Australasian Ground Control inMining Conference, Sydney, New South Wales, 23-24 November 2010.Hagan, P. and Saydam, S. (eds). Australasian Institute of Mining andMetallurgy, Carlton, Victoria.

STRUTHERS, M.A., TURNER, M.H., MCNABB, K., and JENKINS, P.A. 2000. rock mechanicsdesign and practice for squeezing ground and high stress conditions atPerseverance Mine. Proceedings of MassMin 2000, Brisbane, Australia, 29October – 2 November 2000. Chitombo, G. (ed). Australasian Institute ofMining and Metallurgy, Melbourne. pp. 755–764.

TURCOTTE, P. 2010. Field behaviour of hybrid bolt at LaRonde Mine. Proceedings ofthe Fifth International Seminar on Deep and High Stress Mining (DeepMining 2010), Santiago, Chile, 6–8 October 2010. Van Sint Jan, M. andPotvin, Y. (eds). Australian Centre for Geomechanics, Perth. pp. 309–319.

VAKILI, A., SANDY, M., and ALERCHT, J. 2012. Interpretation of non-linear numericalmodels in geomechanics - a case study in the application of numericalmodelling for raise bored shaft design in a highly stressed and foliated rockmass. MassMin 2012: Proceedings of the 6th International Conference andExhibition on Mass Mining, Sudbury, Canada, June 2012. Canadian Instituteof Mining, Metallurgy and Petroleum, Sudbury, Canada. [CD-ROM]. ◆

▲


More challenging conditions facing thecoal mining industryThere are a number of challenges facing thecoal mining industry. These range fromphysical conditions through to socioeconomicchallenges. This paper will deal mostly withphysical challenges, although the approachdiscussed would apply equally to anychallenge. An analysis of the changes inphysical mining conditions over the past sevento eight years found that the followingelements have had an impact on productivity:

➤ The introduction of the 12 m cutting rulein the first part of the previous decade.This rule limits the distance that acontinuous miner can cut from the lastrow of roof support to 12 m, instead oflonger distances (up to 24 m and more)that were used previously. The rule wasintroduced as a measure to preventexplosions by limiting dust levels, aswell as preventing roof falls

➤ Reduction in seam thickness and miningheight

➤ Shortening of panels and more frequentsection moves, which have an adverseinfluence on production

➤ Greater frequency of geologicalproblems: faults, dolerite intrusions,floor rolls etc. Each problem poses achallenge to both productivity andsafety, and these are more frequent asthe more easily accessed resources onaging mines have already been depleted

➤ Increased support density requirementsto prevent roof falls. Most coal mines arenow on systematic support rules and rib-side support is becoming more prevalent.This is as a result of the impact of agingmines’ strategies (to mine more difficultareas, as described above) as well asrestrictions imposed internally bycompanies and through the Departmentof Mineral Resources (DMR) to combatroof and rib-side related accidents.

Case study: seam thickness and miningheight reductionMCS has recently undertaken an analysis ofthe change in the practical mining height ofapproximately 28 continuous miner (CM)sections in the Witbank and Highveldcoalfields over a period of eight years, from2005 to 2012. The results are shown in Figure 1. As can be seen, the average practicalmining height decreased from above 4 m tojust under 2.9 m over the period analysed, andthe trend seems to be continuing. This haspartly come about due to the depletion of 2Seam reserves with high seam heights, and thesubsequent migration to 4 Seam areas withlower seam heights. The 2 Seam reserves weretargeted first due to their higher yield. Theother major reason is that, where given thechoice, mines elected to mine the higher 4Seam areas in preference to the lower areas,and as the higher seam areas become minedout the average seam height decreases. Thus

Visions for challenging assets in theSouth African coal sector by Z. van Zyl*

SynopsisThe southern African coal industry is facing the reality that coal reservesare becoming deeper and harder to mine than before. Alternative visionsneed to be considered to offset the impact of these challenges on produc-tivity, cost, and profitability. One of the main challenges is to maintaincurrent production levels under more difficult conditions, and improveproductivity where conditions allow. There is a need to move from thereactive event-based management system towards the more adaptable andflexible process-based management system. This paper focuses onunderground coal mining, but the principles are also applicable to othercommodities and mining methods. This paper is based on researchundertaken by the personnel from Mining Consultancy Services Pty Ltd(MCS) on trends in the underground coal mining sector, as well aspractical experience over a period of 16 years in the field of electronicmonitoring of mining machinery and productivity optimization.

Keywordsproductivity, optimization, Prodmate, reporting, handheld device,monitoring, improvement, utilization, key performance indicators, coalmining.

* Mining Consultancy Service (Pty) Ltd.© The Southern African Institute of Mining and

Metallurgy, 2015. ISSN 2225-6253. Paper receivedMar. 2015 and revised paper received June 2015.



Visions for challenging assets in the South African coal sector

more linear metres have to be mined for every ton of coal thatis produced, putting pressure on productivity.

Case study: CM section outputFor underground coal mining, ‘millionaire’ status is anaspirational target, which means that a section produced amillion run-of-mine (ROM) tons in a year. From a benchmarkdatabase of more than 100 CM sections in South Africa weextracted the results to determine the number of ‘millionaire’sections over the past few years. The results, shown in Figure2, clearly illustrate that, during a period when managementpractices and technology continuously evolved with the aimof improving productivity levels, there had been a steadydecline in the number of millionaire sections; largely due tothe factors mentioned above.

Attributes of top performersMining conditions are important, but are not the only driverof productivity. Figure 3 illustrates this by comparing theoutput of 20 CM sections, with similar mining conditions andwith the same equipment, over a period of one year. There isa wide differentiation between the top and bottomperformers.

Through the involvement of MCS with productivityoptimization projects at many of the underground coal minesin South Africa, a model, shown in Figure 4, was developedto explain this phenomenon.

Figure 4 illustrates that all sections are exposed to thesame:

➤ Production events, such as the need to relocate fromone roadway to the next after completing the 12 m cut

➤ Equipment events, such as breakdowns and advancesin technology

➤ Geological events, such as encountering anomalies i.e.faults and dykes

➤ Market events, such as changes in prices and logisticalconstraints

➤ HR events, same staff complement ➤ Legislative events, regulating activities of all sections.

Despite these events, which are common to all sections,there is a difference in output as illustrated in Figure 3. Why?The reasons are that the top performers:

➤ Know about the events, like all average performerswould

➤ Understand the events, like most average performerswould

➤ Understand the production process, like some averageperformers would

➤ Understand the impact of the events on the productionprocess, like few average performers would

➤ Make changes to improve the affected process. Foraverage performers, this rational progression isunlikely.

The steps that connect the events to the change in processthat would mitigate the event’s adverse effect is what we referto as the ‘process-based management ladder’, as shown inFigure 4. The second step in the ladder ‘Provide adequate andreliable information’ is often done in underground coal minesthrough the introduction of electronic machine monitoringsystems on the CMs, which are the primary coal-winningequipment. The results from the monitoring systems are thencompared to other sections with similar systems to benchmarkthe sections in question against industry best practice anddetermine the improvement potential. The results from themachine monitoring systems are expressed in such a way thatthey measure the fundamental mining processes from the CMs.These process-based metrics allow the underlying constituentprocesses of the mining method to be managed and improved.By improving the process that deviates the most from industrybenchmarks, inherent value is unlocked and the productivity ofthe mining method is improved.

▲


Figure 2 – Decline in the number of ‘millionaire’ sections, 2006–2012

Figure 1 – Change in mining height, 2005–2012

Figure 3 – Productivity differences, top and bottom performers

Process-based production management was firstenvisaged and introduced in the late 1980s as a means ofredirecting the focus of production optimization tomanageable practices that would produce sustainableimprovements. This management system encourages thesection personnel to build a strong foundation of the processsteps, instead of focusing on the result.

Through MCS’s experience in more than 30 successfulproductivity optimization projects on coal mines, the authorhas found time and again that the change from event-basedmanagement to process-based management is the key tounlocking latent potential. This can be done only when themine has the ability to measure each of the productionprocesses that occur in a section accurately and reliably. It istherefore no surprise that the application and growth inprocess-based management has gone hand-in-hand withimprovements in the monitoring hardware and software,resulting in progressively more accurate and reliable data.The fact that almost all new continuous miners sold into themarket now come equipped with advanced monitoringcapability, bears testimony to this.

As the demands and pressures of reducing costs andmaintaining performance have recently increased, newadvances and approaches in technology to support process-based management are required, as described further in thispaper.

Process-based production management MCS has come to understand exactly what process-basedproduction management is. In its simplest form it is themanagement of primary productivity drivers. The first steptowards managing constraints effectively entailsunderstanding what the constraints are. Radical advances inthe use of progressive software and hardware to monitor andexamine section data has led to many benefits, such as:

➤ The ability to integrate machine-generated andmanually captured data

➤ Creating a management system that can be used tomanage all aspects of the production operation

➤ Effective time management while fostering simple andless confrontational accountability among employees

➤ Making information transparent so that communicationand effective use of skills and experience is improved.

This is backed up by effective change management so thenew methodologies are less prescriptive and more likely tobecome habit-forming, thereby sustaining the improvements.

Process-based management works. It is an effectivemeans of advancing the value of the asset and output bysustaining the day-to-day improvements gained by revolu-tionizing the operating methods. The effectiveness ofprocess-based management methodology is substantiated bythe results achieved, as illustrated in Figures 5 and 6.

By empowering people with relevant information tomanage the performance of their section, management



Figure 4 – Model illustrating reasons for differences in productivity

Figure 5 – Effectiveness of process-based management, illustrated byincreasing production per shift


encourages workers to embrace improvement throughtechnology and change.

Experience over the past 26 years has indicated that thereare three pillars to effective and sustainable process-basedmanagement and productivity optimization (Figure 7).

➤ Monitoring. Accurate and reliable processmeasurement, using data obtained from productionmachines. The monitoring is based only on the primarycutting machine, and focus on the following productionrate (efficiency) KPIs:• Loading time• Away time• Tram time per metre cutThese are primarily machine-related issues. Time management would include the monitoring of thefirst and last operation of the continuous miner

➤ Training. Training of relevant personnel to analyse,interpret, and understand the measurement system andwhat they can do towards improving each of the KPIsthat are under their control

➤ Improvement process. A practical improvementprocess where KPIs are reviewed; action plans aregenerated, implemented, and tracked; and unques-tionable accountability for the KPIs is held across themanagement structure on the mine.

This holistic approach to productivity optimization hasconsistently delivered the best results. It is based on theinterconnectedness of all the aspects; leaving one out willerode the effectiveness and sustainability of the process.

Owing to the factors described at the beginning of thispaper, it has become increasingly difficult to maintainproductivity at constant and acceptable levels, before aspiringfor improvements. Over the past seven to eight years, produc-tivity improvements were much more likely to be achievedthrough improvements in production rates (efficiency) ratherthan improvements in production time (utilization). This isillustrated in Figures 8, 9, and 10, which show utilization(Figure 8) and efficiency (Figures 9 and 10) KPI trends overa sample of 25–30 bord and pillar CM sections.

Figure 8 shows a decline in the average production timeper shift, but the research personnel found that the produc-tivity of the sections from which the data was taken did notdecline, but rather remained constant. This is due to the

▲


Figure 6 – Effectiveness of process-based management, illustrated bymonthly production increases

Figure 7 – The three pillars of process-based management and produc-tivity optimization

Figure 9 – Improvement in efficiency – decreased loading times

Figure 8 – Decrease in utilization – production time per shift

improvement in efficiency of the sections as indicated inFigure 9, which shows the improvement in the time taken tofill a coal hauler, and Figure 10, which shows improvementsin the efficiency with which the CMs are relocated from onecutting position to the next.

Traditional machine monitoring systems have largelybeen used to improve efficiency using loading and away time,which offsets the losses caused by decreased production time,as illustrated by the CM loading time trend in Figure 9 andtramming efficiency in Figure 10 for the same data sample.The declining trend in utilization (Figure 8) prompted theresearch team to investigate better ways of measuringmachine utilization, as described in the next section.

Mining process measuring systemsTraditional machine monitoring systems have been effectiveat measuring and managing the efficiency of machines, butinadequate when it comes to measuring and managingutilization. This led to the development of a undergroundreporting system, ProdMate®, that allows for monitoring ofkey utilization performance indicators on productionmachines (CMs, roofbolters etc.), downtime (maintenance),planning (HR, inventory, supplies), and procedures. Thereporting system combines manual and electronic data toproduce a range of outputs, which can be utilized by genericenterprise reporting programmes (like SAP) to streamlineproduction and improve utilization. Time-related issuesinclude reporting of utilization of the machine as well as thedowntime. Non-reported time has become an increasedproblem in recent years, and the introduction of handheldreporting devices has reduced this phenomenon significantly.

The system essentially comprises a suite of softwareapplications that are run on an intrinsically safe handheldcomputer or personal digital assistant (PDA) as shown inFigure 11. Information is entered by the user (usually thesection miner) and serves as a replacement for paper-basedreports.

The inputs that relate to machines, materials, and otherdowntimes can then be integrated with the data from theelectronic monitoring systems to provide an integratedreporting solution that encompasses close to 100% of thetotal shift time, as illustrated in Figure 12. Data can beextracted either via Wi-Fi interface or a cradle when dockingthe device to charge. Where a Wi-Fi data transfer system is

used, the information is available in near real-time for use.With the introduction of the unit to coal mines in

Australia and South Africa, users have seen potential beyondthe wide range of applications originally anticipated. This hasled to the expansion of the system’s capabilities (asdemonstrated in Figure 13) to include the following:

➤ Maintenance management system➤ Inventory management system➤ In-time production status updates➤ HR control, time and attendance, and licence control➤ Interactive mine planning and forecasting➤ Task and work order creation and management➤ Production reporting➤ Downtime reporting➤ Material and supplies management➤ Document storage and retrieval➤ Have you done it? (where any ad-hoc or periodic tasks

can be loaded and managed).

ConclusionsThis paper has indicated that the challenges that face the coalmining sector in South Africa are significant and serious. Itshows that process-based management has served as aneffective tool to improve and maintain productivity in the faceof these challenges, but that many of the improvements havebeen in efficiency rather than utilization, i.e. production raterather than production time. It indicates that as thechallenges are likely to become more severe in the future, theapplication of process-based management will remainimportant. With the addition of improved measuring systemson utilization though applications such as the ProdMate®

system, process-based management becomes even moreeffective. ◆



Figure 10 – Improvement in efficiency – decreased CM relocation times

Figure 11 – Personal digital assistant running ProdMate®


▲


Figure 12 – Integrated reporting solution Figure 13 – Capabilities of the expanded ProdMate® system

The Journal of The Southern African Institute of Mining and Metallurgy JULY 2015 ▲ix

20156–7 August 2015 — MINPROC 2015: Southern African MineralBeneficiation and Metallurgy ConferenceVineyard Hotel, Newlands, Cape TownContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156E-mail: [email protected], Website: http://www.saimm.co.za

19–20 August 2015 — The Danie Krige GeostatisticalConference: Geostatistical geovalue —rewards and returns forspatial modellingCrown Plaza, JohannesburgContact: Yolanda RamokgadiTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

25–27 August 2015 — Coal Processing—Unlocking SouthernAfrica’s Coal PotentialGraceland Hotel Casino and Country Club SecundaContact: Ann Robertson, Tel: +27 11 433-0063

26–28 August 2015 — MINESafe 2015—Sustaining ZeroHarm: Technical Conference and Industry dayEmperors Palace Hotel Casino, Convention Resort, JohannesburgContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156E-mail: [email protected], Website: http://www.saimm.co.za

28 September-2 October 2015 — WorldGold Conference 2015Misty Hills Country Hotel and Conference Centre,Cradle of Humankind, Gauteng, South AfricaContact: Camielah JardineTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156, E-mail: [email protected], Website: http://www.saimm.co.za

12–14 October 2015 — Slope Stability 2015:International Symposium on slope stability in open pit miningand civil engineeringIn association with the Surface Blasting School15–16 October 2015Cape Town Convention Centre, Cape TownContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

20 October 2015 — 13th Annual Southern African StudentColloquiumMintek, Randburg, JohannesburgContact: Yolanda RamokgadiTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

21–22 October 2015 — Young Professionals 2015 ConferenceMaking your own way in the minerals industryMintek, Randburg, JohannesburgContact: Camielah JardineTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail:[email protected], Website: http://www.saimm.co.za

28–30 October 2015 — AMI: Nuclear Materials DevelopmentNetwork ConferenceNelson Mandela Metropolitan University, North CampusConference Centre, Port ElizabethContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

8–13 November 2015 — MPES 2015: Twenty ThirdInternational Symposium on Mine Planning & EquipmentSelection Sandton Convention Centre, Johannesburg, South AfricaContact: Raj SinghalE-mail: [email protected] or E-mail: [email protected]: http://www.saimm.co.za

201614–17 March 2016 — Diamonds still Sparkle 2016 Conference BotswanaContact: Yolanda RamokgadiTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

13–14 April 2016 — Mine to Market Conference 2016South AfricaContact: Yolanda RamokgadiTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

17–18 May 2016 — The SAMREC/SAMVAL CompanionVolume ConferenceJohannesburgContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156E-mail: [email protected], Website: http://www.saimm.co.za

May 2016 — PASTE 2016 International Seminar on Pasteand Thickened TailingsKwa-Zulu Natal, South AfricaContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156E-mail: [email protected], Website: http://www.saimm.co.za

9 –10 June 2016 — 1st International Conference on SolidsHandling and ProcessingA Mineral Processing PerspectiveSouth AfricaContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156E-mail: [email protected], Website: http://www.saimm.co.za

1–3 August 2016 — Hydrometallurgy Conference 2016‘Sustainability and the Environment’in collaboration with MinProc and the Western Cape BranchCape TownContact: Raymond van der BergTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156E-mail: [email protected], Website: http://www.saimm.co.za

16–19 August 2016 — The Tenth InternationalHeavy Minerals Conference ‘Expanding the horizon’Sun City, South AfricaContact: Camielah JardineTel: +27 11 834-1273/7, Fax: +27 11 838-5923/833-8156 E-mail: [email protected], Website: http://www.saimm.co.za

INTERNATIONAL ACTIVITIES

▲

x JULY 2015 The Journal of The Southern African Institute of Mining and Metallurgy

Company AffiliatesThe following organizations have been admitted to the Institute as Company Affiliates

3 M South Africa

AECOM SA (Pty) Ltd

AMIRA International Africa (Pty) Ltd

Anglo Operations Proprietary Limited

Anglo Platinum Management Services (Pty) Ltd

Anglogold Ashanti Ltd

Arcus Gibb (Pty) Ltd

Atlas Copco Holdings South Africa (Pty) Limited

Aurecon South Africa (Pty) Ltd

Aveng Moolmans (Pty) Ltd

Axis House Pty Ltd

Barloworld Equipment -Mining

Becker Mining (Pty) Ltd

BedRock Mining Support Pty Ltd

Bell Equipment Limited

Blue Cube Systems (Pty) Ltd

Bluhm Burton Engineering Pty Ltd(BLU003)

CAE Mining (Pty) Limited

Caledonia Mining Corporation

Chamber of Mines

Concor Mining

Concor Technicrete

Department of Water Affairs and Forestry

Deutsche Securities (Pty) Ltd

Digby Wells and Associates

Downer EDI Mining

DRA Mineral Projects (Pty) Ltd

DTP Mining

Duraset

E+PC Engineering and Projects Company Ltd

Elbroc Mining Products (Pty) Ltd

Exxaro Coal (Pty) Ltd

Exxaro Resources Limited

Fraser Alexander Group

Glencore

Goba (Pty) Ltd

Hall Core Drilling (Pty) Ltd

Hatch (Pty) Ltd

Herrenknecht AG

HPE Hydro Power Equipment (Pty) Ltd

Impala Platinum Holdings Limited

IMS Engineering (Pty) Ltd

JENNMAR South Africa

Joy Global Inc.(Africa)

Leco Africa (Pty) Limited

Longyear South Africa (Pty) Ltd

Lonmin Plc

Ludowici Africa (Pty) Ltd

Lull Storm Trading (PTY)Ltd T/A Wekaba Engineering

Magnetech (Pty) Ltd

Magotteaux (Pty) Ltd

MBE Minerals SA Pty Ltd

MDM Technical Africa (Pty) Ltd

Metalock Engineering RSA (Pty)Ltd

Metorex Limited

Metso Minerals (Sweden) AB

Minerals Operations Executive (Pty) Ltd

MineRP Holding (Pty) Ltd

Mintek

MIP Process Technologies

Modular Mining Systems Africa (Pty) Ltd

MSA Group (Pty) Ltd

Multotec (Pty) Ltd

Murray and Roberts Cementation

New Concept Mining (Pty) Limited

Northam Platinum Ltd - Zondereinde

Osborn Engineered Products SA (Pty) Ltd

Outotec (RSA) (Proprietary) Limited

PANalytical (Pty) Ltd

Polysius A Division Of ThyssenkruppIndustrial Sol

Precious Metals Refiners

Rand Refinery Limited

Redpath Mining (South Africa) (Pty) Ltd

Rosond (Pty) Ltd

Royal Bafokeng Platinum

Roymec Technologies (Pty) Ltd

RungePincockMinarco Limited

Salene Mining (Pty) Ltd

Sandvik Mining and Construction Delmas(Pty) Ltd

Sandvik Mining and Construction RSA(Pty) Ltd

SANIRE

Sasol Mining(Pty) Ltd

Scanmin Africa (Pty) Ltd

Sebilo Resources (Pty) Ltd

SENET (Pty) Ltd

Senmin International (Pty) Ltd

Smec South Africa

SMS Siemag

SNC Lavalin (Pty) Ltd

Sound Mining Solution (Pty) Ltd

SRK Consulting SA (Pty) Ltd

Technology Innovation Agency

Time Mining and Processing (Pty) Ltd

Tomra Sorting Solutions Mining (Pty) Ltd

Ukwazi Mining Solutions (Pty) Ltd

Umgeni Water

VBKOM Consulting Engineers

Vietti Slurrytec (Pty) Ltd

Webber Wentzel

Weir Minerals Africa

Worley Parsons RSA (Pty) Ltd

2015◆◆ CONFERENCE

MINPROC 2015: Southern African Mineral Beneficiation andMetallurgy Conference6–7 August 2015, Vineyard Hotel, Newlands, Cape Town

◆ CONFERENCEThe Danie Krige Geostatistical Conference 201519–20 August 2015, Crown Plaza, Johannesburg

◆ CONFERENCEMINESafe 2015—Sustaining Zero Harm: Technical Conferenceand Industry day26–28 August 2015, Emperors Palace Hotel Casino, ConventionResort, Johannesburg

◆ CONFERENCEWorld Gold Conference 201528 September–2 October 2015, Misty Hills Country Hotel and Conference Centre, Cradle of Humankind, Muldersdrift

◆ SYMPOSIUMInternational Symposium on slope stability in open pit mining and civil engineering12–14– October 2015In association with the Surface Blasting School15–16 October 2015, Cape Town Convention Centre, Cape Town

◆ COLLOQUIUM13th Annual Southern African Student Colloquim 201520 October 2015, Mintek, Randburg, Johannesburg

◆ CONFERENCEYoung Professionals 2015 Conference21–22 October 2015, Mintek, Randburg, Johannesburg

◆ CONFERENCEAMI: Nuclear Materials Development Network Conference28–30 October 2015, Nelson Mandela Metropolitan University, North Campus Conference Centre, Port Elizabeth

◆ SYMPOSIUMMPES 2015: Twenty Third International Symposium on MinePlanning & Equipment Selection8–12 November 2015, Sandton Convention Centre, Johannesburg, South Africa

SAIMM DDIARY

Forthcoming SAIMM events...

For further information contact:Conferencing, SAIMM

P O Box 61127, Marshalltown 2107Tel: (011) 834-1273/7

Fax: (011) 833-8156 or (011) 838-5923E-mail: [email protected]

F or the past 120 years, theSouthern African Institute ofMining and Metallurgy has

promoted technical excellence in theminerals industry. We strive tocontinuously stay at the cutting edgeof new developments in the miningand metallurgy industry. The SAIMMacts as the corporate voice for themining and metallurgy industry in theSouth African economy. We activelyencourage contact and networkingbetween members and thestrengthening of ties. The SAIMMoffers a variety of conferences thatare designed to bring you technicalknowledge and information ofinterest for the good of the industry.Here is a glimpse of the events wehave lined up for 2015. Visit ourwebsite for more information.

Website: http://www.saimm.co.za

EXHIBITS/SPONSORSHIP

Companies wishing to sponsor

and/or exhibit at any of these

events should contact the

conference co-ordinator

as soon as possible

© New Concept Mining 2015Patents Pending

Integrated systems of support

+27 11 494 6000www.ncm.co.za

Applying Poka Yokesin the mining industry

saimm 201507 jul

Documents