C O N T E N T SC O N T E N T S

2

4

16

32

46

58

70

82

94

106

116

130

Editorial

Overcoming the Rain Fade Obstacle – UnderstandingSingapore's Rain Dynamics and Feasible CountermeasuresAgainst Rain Fading

Reducing the Risk of Flashovers in the Design of anUnderground Ammunition Storage & Processing Facility

Creating and Valuing Flexibility in Systems Architecting:Transforming Uncertainties into Opportunities UsingReal Options Analysis

Automated Red Teaming: A Proposed Simulation-basedFramework

Spare Parts Management for Large-scale Fleet Scenarios

Continual Systems Development for Command, Controland Intelligence Systems

Software Safety – Back to Basics, Knowing Where to Tap!~ A DSTA Perspective

Hazard Re-classification of 76mm Naval Gun Ammunitionfollowing UN Test Series 6

Business Intelligence in Government Procurement

Introduction to Mine Clearing Technology

Future Energy and Power Challenges

E D I T O R I A L

Tay Kok PhuanEditor, DSTA HorizonsDirector (DSTA College)

2

In this fifth edition of DSTA Horizons, we bring you 11 articles covering a wide range ofprojects and studies undertaken by the defence community. They attest to the diversityof expertise that resides within the defence community and the recognition to respondto the complex and fast-changing milieu today. Some articles in the edition discussalternative ways to achieve greater efficiency and more effective use of resources. Othersseek to share new perspectives on managing challenges and uncertainty. There arecontributions which review and analyse the effectiveness of solutions already implementedand those which are informative reads. Of these, five have been presented at internationaland local conferences. We are in the privileged position to be able to compile theircontents for the purpose of extending them to the defence community.

The first article, Overcoming the Rain Fade Obstacle – Understanding Singapore's RainDynamics and Feasible Countermeasures Against Rain Fading, recounts a research team’sstrive for greater accuracy in the prediction of rain fading and rain statistics in Singapore.Current rain attenuation models rely on data collected in various regions globally and mightnot provide an accurate representation of rain fade in Singapore. The team describes themain rain characteristics that are considered in developing a suitable model based onlocal weather information in the quest to create a reliable satellite communications systemtailored for Singapore’s environment.

Reducing the Risk of Flashovers in the Design of an Underground Ammunition Storage& Processing Facility is the second article which explores the possible causes of flashoverin an underground ammunition storage area, and the engineering efforts adopted toimplement the grounding, bonding and electrical systems within such a facility.

Being prepared to manage uncertainty is an essential skill in the undertaking of a complexproject. The next article, Creating and Valuing Flexibility in Systems Architecting:Transforming Uncertainties into Opportunities using Real Options Analysis, discusses thenotions of ‘Real Options Thinking’, akin to good management intuition and ‘Real OptionsValuation’, akin to stochastic optimisation to promote a common ‘options language' tofacilitate project management especially in uncertainties. This framework increases ourcapacity to better identify, create and secure options, and empower us to actively manageuncertainties to create opportunities and reduce risks.

The fourth article, Automated Red Teaming: A Proposed Simulation-based Framework,discusses simulation-based operational analysis methods to uncover system vulnerabilitiesand to find exploitable gaps in military operational concepts. A proposal to use anautomated red teaming framework, to complement the manual red teaming which canbe human-intensive, was put forth. Both methods were tested on various maritime securityscenarios and the result analysis is documented for a comparison of effectiveness betweenthe two approaches.

The optimal provisioning and allocation of spare parts is vital in the management ofsystems with large fleet size, such as systems in the Army. The article, Spare PartsManagement for Large-scale Fleet Scenarios, discusses an in-house developed simulationmodel known as PIPER that the Army uses to study the maintenance support concept, theimpact of combat damage and workshop manpower staffing.

The traditional software development model is no longer adequate to meet the challengesof the fast-evolving needs of a Third Generation Singapore Armed Forces (SAF). Thearticle, Continual Systems Development for Command, Control and Intelligence Systems,details the risk management framework adopted by DSTA which enables the SAF todevelop a high level of flexibility to respond rapidly to meet changing needs and addressemerging threats in the aspect of Command, Control and Intelligence systems.

The use of software to control and manage C4I systems has become pervasive in recentyears. The seventh article, Software Safety – Back to Basics, Knowing Where to Tap,addresses the importance of fundamental software testing techniques such as white-boxtesting and the use of quantitative metrics as a proxy for software system acceptance tohelp mitigate the risks associated with software safety.

The article, Hazard Re-classification of 76mm Naval Gun Ammunition following UN TestSeries 6, details the motivating factors, test programme and positive results of the re-classification. The results of this undertaking have significantly enhanced emergencyresponse and platform survivability, as well as eased berthing constraints and increasedstorage flexibility and capacity.

Business Intelligence in Government Procurement describes how Business Intelligence, viaa three-pronged approach of Intelligent Procurement, Portfolio Management andPerformance Management, has helped GeBIZ users exploit past procurement experiencesto reduce turnaround time, increase productivity and ensure accountability of publicfunds.

An informative read can be found in Introduction to Mine Clearing Technology, whichprovides an overview of the technologies and methods developed for mine clearingoperations currently used by the military and humanitarian demining organisations. Thearticle highlights that there is no single method to resolve the problem and that acombination of tools would ensure a more successful mine clearing exercise.

The last article, Future Energy and Power Challenges, highlights the various energy andpower challenges faced by the Ministry of Defence and the SAF. The article also discussesthe use of alternative energies, energy management modes and even advanced compositesto overcome challenges and meet the energy and power demands in a cost-effectivemanner.

On a related note, and in keeping with saving the Earth’s resources, we have printedfewer hard copies of DSTA Horizons this year. The online version of DSTA Horizons isposted on the DSTA website at www.dsta.gov.sg for easy access. We seek your help toconvey this message to members of the defence community as we want to continue inour quest to promote learning, writing and sharing of valuable knowledge through DSTAHorizons.

Overcoming the Rain FadeObstacle – UnderstandingSingapore’s Rain Dynamics and Feasible Countermeasures Against Rain Fading

ABSTRACT

Rain fade has always been perceived as the major impediment

to satellite communications operating at higher frequencies,

particularly in tropical zones like Singapore. Researchers have

proposed various rain models to predict the extent of rain

fading in different regions globally. However, there still exists

an element of uncertainty of their applicability to our climate.

As such, it is of paramount importance to establish a suitable

high fidelity rain attenuation model based on local weather

information. By means of a more accurate estimation of rain

fade, and in tandem with the expected benefits of fade

mitigation techniques, these key technologies will enable the

realisation of a reliable satellite communications system.

Dr Lee Yee Hui

Koh Wee Sain

Lee Yuen Sin

Michelle Ho Xiu Mei

6

INTRODUCTION

Commercial ground-space communications or

Satellite Communications (SATCOM) have

traditionally been operating in the C band

(4GHz to 8GHz), the first allocated frequency

band that is used predominantly for the

reception of satellite television programmes

since some 40 years ago (Freer, 1996).

C band is generally characterised by a large

consumer antenna dish size of at least 2.4m

and can offer wide coverage with high

availability due to its resilience in the presence

of heavy rain.

However, the C band occupancy for SATCOM

is increasingly congested due to the rapid

emergence and deployment of terrestrial

services such as WiMax, terrestrial microwave

networks as well as the sharing of limited C

frequency spectrum with radar systems. In

fact, there have already been cases of terrestrial

interference to satellite services in countries

like Australia, Hong Kong and Indonesia

(Hartshorn, 2007).

Concurrently, there is a growing global demand

for mobile, broadband (i.e. high data rate)

applications via SATCOM such as airborne or

maritime Internet broadband connectivity and

land-based on-the-move applications.

This string of events has prompted the need

to explore the use of higher frequency ranges

such as Ku (11GHz to 18GHz) and Ka (26.5GHz

to 40GHz) in meeting the SATCOM worldwide

market requirements. Industry analysis has

shown that Ku band transponders grew by

20% (624 transponders) as compared to the

9% growth rate (260 transponders) for C band

transponders between the years 2000 to 2003

and this trend is expected to carry on (Futron

Corporation, 2003). In Asia, the steady streamof expected launches over the next one yeare.g. Ku entrants such as MEASAT-3a, Optus D3and Intelsat 15 (LyngSat, 2009) is also largely

market-driven.

The main drawback in using these higher Kuand Ka frequencies over C band is thedegradation of the communication channelsdue to severe atmospheric and environmentalfade impact, particularly rainfall. Rain fade orrain attenuation is considered a dominantimpairment for frequencies above 10GHz as itmay limit the availability of the link. Hence,this article will focus on the potential effectsof rain fade in our region and thecountermeasures that can be employed tomitigate them.

RAIN FADE

Rain fade is a common, yet oftenmisunderstood weather phenomenon thatinterrupts wireless communication signals inthe presence of rain. It occurs with all typesof satellite systems e.g. Geostationary EarthOrbit (GEO), Medium Earth Orbit, Low EarthOrbit and Global Positioning System.

Nevertheless, rain fade is not as big a barrieras it has been anticipated to be. A commonapproach in enabling a high percentage ofsystem availability during rain events is toemploy Fade Mitigation Techniques (FMT).Coupled with proper planning and design ofthe SATCOM network, rain fade impact can beminimised.

Causes Of Rain Fading

Due to the higher operating frequency for aKu or Ka band SATCOM system, the signal’swavelength is generally shorter as comparedto the C band. Therefore, it is more susceptibleto signal degradation as the wavelengthapproaches the size of a typical raindrop. Thetwo major causes of rain fade are:

• Absorption – water molecules in a raindroplet absorb portions or all of the signalenergy of the passing radio wave. With shorterwavelength, there will be more interactionbetween the radio wave and water molecules,leading to increased energy losses.

Overcoming the Rain Fade Obstacle –Understanding Singapore’s Rain Dynamics and

Feasible Countermeasures Against Rain Fading

7

• Scattering – this is a physical process, causedby either refraction or diffraction, in which thedirection of the radio wave deviates from itsoriginal path as it passes through a mediumcontaining raindrops. This disperses the energyof the signal from its initial travel direction.

The accumulation of these different reactionsultimately leads to a decrease in the level ofreceived signal, thus resulting in rainattenuation.

Rain Attenuation Model

To date, several rain attenuation models suchas the Global Crane (Crane, 1980) and ITU-RP.618-8 (ITU Recommendation, 2007) have beencreated to compute the rain fade figure forearth-space telecommunication systems. Rainmodels predict attenuation for a givenfrequency from site-related parameters suchas rain intensity statistics, rain height and path-elevation angle. Most of the models availabletoday are developed based on the datacollected from temperate regions (Mandeep,Allnutt, 2007).

An ongoing research collaboration betweenDSTA and Nanyang Technological University(NTU) is looking into the establishment of asuitable high fidelity rain attenuation model,based on weather information fromSingapore’s Meteorological Services Divisionof the National Environment Agency.

The principal objective of a rain attenuationmodel tailored for Singapore’s environment isto achieve an accurate estimation of theattenuation level incurred by the signal dueto rain which is key to realising a reliableSATCOM system.

This rain attenuation model utilises amethodology similar to ITU-R P.618-8, with thederivation of rainfall parameters such as thespecific attenuation and path reduction factorbeing optimised to local data and findings.Together with the site-related parameters, thefour key characteristics that influence the levelof rain attenuation experienced are asexplained in Table 1.

Parameters Description Definition

Specific This is a fundamental quantity Specific attenuation definesAttenuation to model rain attenuation. It is the rain attenuation per unit

a function of rain temperature, distance.velocity of rainfall and mostimportantly, raindrop sizedistribution.

Slant Path Rain fade is not experienced at The slant path length is definedLength a specific collection point but as the length of the satellite to

extended throughout the entire ground path that is affected bypropagation path. rain.

Path Reduction The rain rate along a propagation The path reduction is incorporatedFactor path is not constant, especially for to account for the spatial

large distances as in the case of variability of a rain eventspace-to-ground transmissions. and ensure higher accuracy.

Rainfall Rate Rainfall data is essential for Rain rate data is presented in unitsStatistics determining the degree of rain of mm/hr over an average year asFactor attenuation in a SATCOM system. a function of the cumulative

probability of occurrence.

Table 1. Description and definition of key parameters that affect rain attenuation

8

SINGAPORE’S RAIN ENVIRONMENT

The following sections describe the maincontributing factors and characteristics thatshape the investigation and development ofa high fidelity rain attenuation model forhigher frequencies of K band in Singapore.

Specific Attenuation

Raindrop size distribution (DSD) is a primarycontributing factor to express the specificattenuation and the severity of signalattenuation is a result of different raindropsizes. Preliminary DSD collection results attainedwith a distrometer for varying rain rates inSingapore showed that the critical dropdiameter is between 0.771mm and 5.3mm.A negligible number of drop diameters wereout of this range and hence rendered as havingan insignificant effect on the overall rain rateand specific attenuation.

Slant Path Length

For earth-space propagation, the slant pathlength from the satellite to the ground stationthat transits through the rain cell is dependenton rain height and elevation angle. While ITU-R P.839-3 (ITU Recommendation, 1999) definesthe rain height in Singapore to be 4.5km, theelevation angle varies with different satellitesand earth station locations. Assuming an earth

station sited in Singapore points to ST-1, theslant path length through rain is calculated tobe 4.75km.

Path Reduction Factor

The Meteorological Doppler Weather Radarsystem is developed to intercept returningradio energy, which will aid in identifying theintensity of a rain event, through measuringthe radiated energy that is being backscatteredand reflected to its receiver. This systemprocesses weather information detected bythe Doppler radar to provide a representationof the motion and intensity of rain known asthe reflectivity plot.

As the path reduction factor is introduced tocompensate for the inhomogeneity of a rainevent, an understanding of the rain distributionover the slant path is required. Figure 2 showsa reflectivity plot of both a stratiform andconvective rain event captured on 15 and 19June 1998 respectively.

It can be seen that for a convective event, thereis a larger variation in the rainfall rate overthe propagation slant path as compared to astratiform event. There are typically moreoccurrences of convective rain events in theequatorial region. Thus, the path reductionfactor will not be a static formula but asummation of values due to the fast varyingrain rate.

Figure 1. Slant path length through rain

9

Rainfall Statistics

The location of the earth station is one otherconsideration that will vastly affect the rainattenuation. Due to the lack of comprehensiveactual rain data, prediction models like ITUP-837 (ITU Recommendation, 2003) and CCIRRep. 563-4 (CCIR, 1990) have been developedto determine the rainfall intensity in variousregions. The rainfall rates for Singapore at0.01% exceedance worked out from thedifferent prediction models are depicted inTable 2.

These figures reflect that for 99.99% of thetime, the statistical rainfall rate falls below theindicated value for the respective models. Inother words, the rainfall rate exceeds thatvalue for 0.01% of the time.

Using measured rainfall statistics obtainedfrom the local meteorological stations througha five-year period (2000-2004) over fivedistributed sites, Singapore’s rainfall rate isfound to be 149mm/hr at 0.01% exceedance(Leong, Foo, 2007). This empirical figure revealsthat both CCIR Rep. 563-4 and Crane Globalmodels are currently good reference modelsto be adopted.

Figure 2. Reflectivity plot of a stratiform and convective rain event

Table 2. Predicted rain rate at 0.01% exceedance in Singapore

Location ITU P-837 CCIR Rep. 563-4 Crane Global

Singapore 120mm/hr 145mm/hr 147mm/hr

10

East Asia. In one such study, measurements ofrain attenuation in Indonesia collected over aperiod of two years have indicated that themeasured rain attenuation is relatively closeto the DAH1 and ITU Model (Suryana, 2005).In a separate study done by NTU, the resultsin Singapore pointed to a great discrepancybetween the measured and various models, asshown in Table 3 (Ong, Zhu, Lee, 1997).

These conflicting results potentially suggestthat these rain models might not provide anaccurate prediction of rain attenuation fortropical areas. Continued developing effortsare thus required to build a high fidelity rainattenuation model that is tailored to ourclimate.

Figure 3. Locations of the rain measurement sites

Table 3. Comparison of measured and modelled rain attenuation for Ku band in Singapore

RAIN FADE AT HIGHER FREQUENCY TRANSMISSIONSIN SINGAPORE

One of the most common methodologies forgauging rain fade is the utilisation of rainattenuation models. They predict the long-term rain attenuation statistics from pointrainfall rate given the site-related parameters.Due to the differences in rain characteristicsin both temperate and tropical regions,academics have been relentlessly attemptingto determine the accuracy of these models invarious local environments.

Several studies have been conducted to validatethe different rain attenuation models in South

Rain Attenuation

CCIR 564-3 CCIR 564-4 ITU-R P.618.3 ITU-R P.618.8 Measure Mean

99.77% 6.91 3.85 4.41 5.72 8.9

99.90% 10.27 5.72 6.56 8.69 14.56

AnnualAvailability (%)

11

FADE MITIGATION TECHNIQUES

The occurrence of rain outage does notnecessarily suggest that there will be acomplete tota l d i s rupt ion to thecommunications link. To overcome thedynamics of outages, FMT is undeniably acritical element in the design of SATCOMnetworks as the effects of fading can becombated by exploiting the correct mitigationmeasure(s) to improve the communicationsavailability and reliability of satellite links. Inother words, the incorporation of FMTthroughout the chain from the user terminal,the spaceborne payload to the overall networkdesign, can potentially offer a reduction inoutage time.

Uplink Power Control

The simplest way to compensate for the rainfade effect is to increase the transmission powerat the terminal end. This method is known asuplink power control. However, constanttransmission of high power may result ininterference among users during clear skyconditions. The underlying technologytherefore resides in the dynamism of the systemthat is capable of adjusting the power, inresponse to the fading variations, to therequired signal level essential for higherfrequency operations. Three types of uplinkpower control algorithms can be executed tomaintain carrier power or signal quality duringrainy periods.

• Open loop power control – The satellitegenerates a beacon signal to the receivingground station which is used to ascertainthe level of downlink attenuation. Thepower controller at the ground station thenestimates the uplink fade required byapplying the frequency scaling ratios forcloud, gaseous and tropospheric scintillationattenuation.

• Closed loop power control – Similar to theopen loop power control architecture, theground station utilises a loopbackcommunication s ignal instead. I tstransmitted beacon signal is analysed toestimate and counteract for the rainattenuation i.e. the received signal comprisesboth uplink and downlink rain effectdegradation.

• Feedback loop power control – A centralstation monitors the signal levels of all thereceived carriers and analyses the poweradjustment needed for the affected carriers.This control information and command aresubsequently routed to the transmittingground station for the corrective measure(s)to be effected.

Though uplink power control is affordable andsimple, it provides marginal benefit as it cannotcontinuously amplify its margin. Most amplifiersexhibit a non-linear behaviour and the outputpower will be limited despite an increasein power.

Adaptive Coding andModulation / Variable Codingand Modulation

In recent years, open standards such as theDigital Video Broadcast-Satellite SecondGeneration (DVB-S2) have emerged. One ofits features, Adaptive Coding and Modulation/ Variable Coding and Modulation (ACM / VCM),lies in its dynamic assignment of coding andmodulation as a key enabler for dynamicbandwidth allocation and optimisation.Examples of DVB-S2 compliant satellite modemsthat are equipped with ACM / VCM capabilityare Viasat’s Linkstar DVB-S2 Broadband VerySmall Aperture Terminal System and theiDirect Evolution Series of Modems.

The ACM / VCM technique leverages codingand modulation adjustments to achieveacceptable throughput over fading link

12

channels. In each DVB-S2 frame that istransmitted, the header contains informationon how that frame will be modulated andcoded. An ACM / VCM-enabled demodulatorwill read this information to demodulate anddecode the frame, wherein each header allowsfor the varying of the modulation and codingscheme for every transmitted frame.

The differentiating component between ACMand VCM is the return channel from eachreceiving site to the transmit site which is onlyavailable in ACM. In the case of ACM, thereturn channel supports a dynamic modificationof the coding and modulation rate of eachframe in accordance with the measured channelconditions. The application of this feature willimprove the system capacity, peak data rateand coverage reliability as the signaltransmitted to and from a particular user ismodified to account for the signal qualityvariation. This will exploit good transmissionconditions to obtain higher throughput, whilereducing the demands on the channel whenthe conditions are temporarily bad.

Site Diversity

The downlink transmissions of a satellite covera very large geographical area, with differentareas experiencing different weather or rainintensities. Site diversity takes advantage of

the limited size and spatial extent of a high-rain-intensity cell to yield an improved networkthroughput.

In other words, the concept of this techniqueis to have two or more interconnected earthstations situated at a sufficient distance apartand to evaluate the signal fading at eachindividual remote terminal before choosing asite with better propagation conditions. Thearchitecture of this system encompasses thelinkage of several stations receiving the samesignal and a control centre that will performthe computations and switchings. A diversitysystem that switches between signals obtainedat various receivers can benefit from the localspatial variation and movements of the raincell, thereby enhancing the overall link quality.

The performance of the site diversity technique,carried out in Singapore, was evaluated withKu band signals from INTELSAT. These signalswere monitored at two earth stations, namelyNTU and Bukit Timah (BKT) which are separatedby 12.3km (Isaiah, Ong, Choo, 2000). Thecumulative distribution of attenuationmeasured at NTU and BKT is illustrated inFigure 4.

It can be seen that when the site diversitytechnique is employed, the overall level ofattenuation will decrease. For instance, the

Figure 4. Cumulative distribution of attenuation

13

exceedance percentage is reduced from 0.25%to 0.035% for a typical link margin of 7dBwhich translates to a decrease of 18.8 hours insignal outage in a year.

Frequency Diversity

The strategy is to draw on the lower frequencyband payload during the occurrence of fadingat higher frequencies via a ground controlstation. The affected earth stations will beauthorised by the control station to tap thelower frequency resources when the climaticcondition reaches a certain threshold while aprocessor onboard the satellite ensures theinterconnection among the stations operatingover two different bands.

This method is suitable for satellites operatingin two frequency bands, typically Ka band andC or Ku band, and usually achieves low levelsof outage probability especially during severerain fade. Furthermore, it can serve as acountermeasure to mitigate any interferenceencountered. The downside is that it is not acost-effective solution to be implementedonboard commercial satellites as it makesgreater demands on both the space and groundsegment resources.

Satellite Diversity

Another means of FMT that can be adopted issatellite diversity. This concept allows the earthstation to select one of the available GEOsatellites that will provide the most favourablelink under the existing propagation conditions.Satellite diversity trials involving three earth-space propagation paths between Osaka, Japanand three satellites, namely JCSAT, BS and N-Star, have reflected an attenuation reductionat Ka band of at least 3dB for a time percentageof 0.01% (Maekawa, 2002).

Besides enhancing the link performance andavailability, a vital advantage of satellitediversity is its ability to mitigate the effects ofnetwork operational issues in the event of asatellite failure.

CONCLUSION

It is undeniable that the next generationSATCOM system necessitates the employmentof a methodical approach in its planning anddesigning, in view of the challenge to theavailability of frequency spectrum at lower Cband frequencies as well as the rising demandfor a wide array of mobile and broadbandapplications. Rain fade poses a persistentchallenge which cannot be totally eradicated,especially for SATCOM links above C band.Fortunately, rain fade usually does not lastlong and normal communications return oncea heavy shower has passed.

To better anticipate the impact of rain fade,it is essential to develop an applicable rainattenuation model customised for the localenvironment. This comprehensive modelpresents an opportunity to shed more light onthe rain dynamics experienced in tropicalregions and will form the foundation foryielding a consistent process in the overalldesigning of the SATCOM network.

Further to that, FMTs, either used singularlyor in combination, offer enormous potentialin overcoming the adverse effects of rain fade.Demonstration initiatives and pre-commercialactivities have already been launched andcommenced to examine the use of FMT onmobile DVB-S2/RCS-based broadbandinteractive satellite systems on the railway.With the commercialisation of mitigationtechniques and proper system planning, hopesof a more effective combating of rain fadeeffects in the tropical region may be raised.

REFERENCES

14

CCIR. (1990). Radiometeorological Data. Vol.V,Rep 563-4.

Crane, R.K. (1980). IEEE Transactions onCommunications, Vol. COM-28, No. 9,September 1980; 0090-6778/80/0900-1717.

Freer, J.R. (1996). Computer Communicationsand Networks, 2nd Edition.

Futron Corporation. (2003). Global Analysis ofSatellite Transponder Usage and Coverage.Retr ieved on 25 June 2008 f romhttp://www.satelliteonthenet.co.uk/white/futron3.html

Hartshorn, D. (2007). OpEd: Global Threat toC band Satellite Services. Retrieved on 29August 2008 from http://www.space.com/spacenews/archive07/hartshornoped_0226.html

Isaiah K.T., Ong, J.T., Choo, E.B.L. (2000).Performance of the Site Diversity Technique inSingapore: Preliminary Results.

ITU Recommendation ITU-R P.618-8. (2007).Propagation data and prediction methodsrequired for the design of Earth-spacetelecommunication systems.

ITU Recommendation ITU-R P.837-4. (2003).Characteristics of precipitation for propagationmodelling.

ITU Recommendation ITU-R P.839-3. (1999).Rain Height model for prediction methods.

Leong, S.C., Dr Foo, Y.C. (2007). Singapore RainRate Distributions.

LyngSat. (2009). Satellite Launches Asia.Ret r i eved on 1 Apr i l 2009 f romhttp://www.lyngsat.com/launches/asia.html

Mandeep, J.S., Allnutt, J.E. (2007). RainAttenuation Predictions at Ku band in SouthEast Asia Countries.

Maekawa, Y. (2002). A study on the effects ofsatellite diversity on Ka band attenuation inthree earth-space paths.

Ong, J.T., Zhu, C.N., Lee, Y.K. (1997). Ku bandSatellite Beacon attenuation and Rain RateMeasurement in Singapore - Comparison withITU-R Models.

Suryana, J., Utoro S, Tanaka, K., Igarashi, K.,Iida, M. (2005). Study of Prediction ModelsCompared with the Measurement Results ofRainfall Rate and Ku band Rain Attenuationat Indonesian Tropical Cities.

1DAH (Dissanayake, Allnutt, Haidara Model) isone of the rain models used for rainattenuation predictions.

ENDNOTES

Dr Lee Yee Hui is an Assistant Professor at the School of Electrical and ElectronicEngineering, Nanyang Technological University (NTU). She is currentlyresearching on channel characterisation as well as electromagnetic wavepropogation and the effects of weather on it. Her research interests lie inantenna design, computational electromagnetics and propagation throughelectromagnetic band gap structures. She received her Bachelor and Masterdegrees from NTU, and holds a PhD from University of York, England.

BIOGRAPHY

Koh Wee Sain is a Senior Engineer and Project Manager (Communicationsand Tactical Command and Control) who is responsible for assessing emergingbeyond-line-of-sight wireless technologies for the Singapore Armed Forces(SAF). Wee Sain was part of the pioneer team which set up the SAF Centrefor Military Experimentation (SCME), the one-stop centre for all SAFexperiments. He received his Bachelor degree in Electrical and ElectronicEngineering (Honours) from NTU in 1999.

15

Lee Yuen Sin is a Senior Engineer (Communications and Tactical Commandand Control) with responsibilities in assessing emerging non-terrestrial wirelesscommunications technologies and systems. Previously a seniorconsultant in SCME, Yuen Sin was a key player in designing, conducting andanalysing C4I experiments that explore future operational concepts enabledby cutting-edge technology. She received her Bachelor degree in Electricaland Electronic Engineering from NTU in 2000.

Michelle Ho Xiu Mei is an Engineer (Communications and Tactical Commandand Control). She is actively involved in the conceptualisation and design ofan accurate high fidelity rain model for Singapore, and is also responsiblefor the assessment and implementation of advanced Communications groundsegment technologies and systems. Michelle obtained her Bachelordegree in Electrical and Electronic Engineering (Honours) from NTU in 2008.

Reducing the

Risk of Flashoversin the Design of an UndergroundAmmunition Storage & Processing Facility

ABSTRACT

Flashovers are hazards to ammunition and human safety in an

ammunition storage area. Therefore, it is vital that such areas

are designed with features that prevent flashovers initiated

by electrical faults, external lightning events and electrostatic

discharges. A comprehensive grounding and bonding scheme

is essential to minimise such a hazard. This paper describes the

flashover risks addressed in the design of an underground

ammunition storage area, and the various engineering efforts

adopted to implement the grounding, bonding and electrical

systems within such a facility. The performance and maintenance

of these safety features are addressed.

Chua Hian Koon

Teh Siaw Peng

Ho See Fong

Lim Jiunn Shyan

18

INTRODUCTION

There are many benefits of storing ammunitionunderground. Besides providing goodprotection from adversarial attacks, such afacility could also reduce the land sterilisationarea required to ensure safe land use aroundthe area. Despite these benefits, there are newchallenges in designing a safe environmentfor underground ammunition storage.‘Flashover’ is a general term used to describea disruptive discharge around or over thesurface of a solid or liquid insulator. In thedesign of any ammunition storage andhandling facility, it is important to considerthe possible catastrophic events that can beinitiated by the occurrence of a flashover. Aflashover with a high energy content, and inthe presence of a good mixture of explosivedust or flammable gas, is likely to cause anexplosion. There are three classes of flashoverhazards that need to be addressed:

The first class refers to lightning-inducedflashover hazards. The safety of theammunition stored underground from directlightning strikes or from lightning-inducedvoltages must be considered.

The second class refers to electrical flashoverhazards. In an underground facility, electricalpower is essential for lighting, ventilation andpumping functions. Electrical flashover hazardscan occur in the form of short circuit faultsalong the electrical network, over-voltagesgenerated by the switching of electricalequipment or the aging of or damage in theinsulation of electrical equipment.

The third class refers to flashovers induced bythe accumulation of electrostatic charges.During the handling of sensitive explosivedevices such as electric fuze primers,electrostatic charges accumulated on a humanbody or any charge storage device couldaccidentally discharge through the primercircuit and fire it off.

There is limited knowledge and experience indesigning ammunition storage and processingfacilities in an underground environment. Thereis also a lack of literature that addresses theflashover hazards peculiar to undergroundammunition storage and processing. Muchinsight has been gained from studying themining industries and underground tunnellingprojects. This paper shall address the concernsof flashover hazards in building such a facilityin Singapore and suggest some engineeringmeasures that have been adopted to mitigatethe occurrence of such hazards.

LIGHTNING FLASHOVER HAZARDS

Singapore has one of the highest rates oflightning activity in the world. Lying near theequator, Singapore has weather that is hotand humid almost all year round. Conditionsare favourable for the development oflightning-producing thunderstorm clouds. Anaverage of 171 thunderstorm days (days whenthunder is heard) is recorded annually inSingapore, according to the NationalEnvironment Agency, Singapore (2002).Lightning in Singapore has an average groundflash density of 12.6 strokes per year per squarekilometre. The occurrence will be higher alongthe coastal region, areas with water bodiesand higher ground with vegetation. Thisawesome natural phenomenon, with thecapacity to reach discharges of up to 200kAand create voltage potentials of up to millionsof volts, can wreak havoc if not properlyaddressed.

Conventional above-ground ammunitionstorage facilities can be effectively protectedagainst the effects of lightning by fitting thefacilities with air termination networks, earthtermination systems, appropriate surgeprotection and more importantly, leveragingthe interconnected network of reinforcementbars usually inherent in the construction ofsuch facilities. This concept of protecting the

Reducing the Risk of Flashoversin the Design of an Underground Ammunition Storage & Processing Facility

19

ammunition through potential equalisation,by creating a Faraday cage around theammunition, has been tested to be effective(International Symposium on DefenseConstruction, 2002). By ensuring the placementof ammunition at practical distances from thewalls of the facilities, lightning flashoverhazards could be effectively mitigated.

There is limited knowledge, however, on theeffects of lightning strikes on an undergroundenvironment. An ammunition storage facilitybuilt in granite rocks may appear to be wellprotected, yet the possibility of other lightning-induced flashover risks cannot be discounted.Technical literature from the internationalengineering community has revealed thepossibilities of lightning propagation throughthe earth and the potential for ignitingflammable gases in underground coal mines(Novak and Fisher, 2001). Studies have shownthat a potential difference of up to 15.6kVcould be generated at the base of a coal mine600-feet deep under the surface.

In May 2007, a research team from SandiaNational Laboratories published a news releasesuggesting that the explosion at the SagaoMine, located near Buckhannon, West Virginia,US on January 2006, was likely to be causedby lightning. The research team cited twopossible modes of transmission of lightningenergy deep into the coal mine. The first modeis through a lightning strike on the earth’s

surface and the resulting propagation oflightning current through the overlying stratainto the underground space. The second modeis through a direct lightning strike on metallicpenetrations such as conveyers and power linesat the entrance of the mines and propagatingdeep within.

Therefore, it is important to consider the effectsof lightning in an underground space, especiallyif the underground space is built within high-grade granite rocks. Granite rocks have a highresistivity of between 1000 -m to 3000 -m.The depth of penetration of lightning isproportional to the resistivity of the earth. Thismeans that there is a possibility that significantelectric fields generated by lightning strikeson the surface of high resistive granite rockscould propagate deep into the undergroundspace. If the earth strata were uniform, theelectric fields generated underground shouldbe evenly distributed such that there wouldnot be any build-up of significant potentialdifference. However, in reality, rock strata areusually not uniform, especially at the ceilingof the rock cavity.

In the construction of an underground storagefacility, it is a common practice for steel boltsto be drilled into the rock crevices as a safetymeasure to prevent loose rocks from falling.Additional bolts are also drilled into the rocksto support overhanging structures for theventilation system, as shown in Figure 1. These

Figure 1. Rock bolts and supporting bolts commonly used in underground space

Rock Bolt

SupportingBolts forOverhangingStructures

20

rock bolts and overhanging structurescontribute to create a non-uniform electricfield distribution along the ceiling surface,thus creating potential gradients.

Some forms of ammunition contain flammableliquid which, upon release into the atmospheree.g. during accidental spillage, will easilyvapourise and form flammable vapour clouds.In an underground space, ventilation is limitedand any gaseous emission will potentially becontained in a confined space. If flammablevapour or explosive dust were to accumulateto an appropriate mixture at the top of theunderground space, a lightning-inducedpotential difference might generate flashoversthat could spark off the combustion of thevapour or dust. In the building of anunderground ammunition facility, the questionof whether a lightning strike would generatea potential difference between theunderground rock bolts great enough toinitiate a flashover has to be addressed. Figure2 shows an illustration of this problem.

Assuming that the lightning strikes directlyabove point A, where rA = DE, the inducedpotential difference between two points inthe underground space can be expressed bythe following relationship:

Assuming that the resistivity of the earth stratais uniform and that the current from alightning strike on the surface of earth isdischarged uniformly in a semi-hemisphericalmanner into the earth, the potential differencebetween two points in the ceiling of theunderground space can be estimated by thefollowing relationship:

Figure 2. Illustration of lightning propagation into underground space

There are two possible strategies to prevent abuild-up of potential difference. One strategyis to ensure that the potential gradient in the

Lightning strike on earthsurface, with current IL

21

underground space does not exceed theminimum breakdown strength of air. Thedielectric breakdown strength of air rangesfrom 1MV/m to 5MV/m, depending on thehumidity and the atmospheric pressure. In theworst case scenario, if two adjacent rock boltsare separated by 0.5m, there needs to be apotential build-up of 500kV across the twobolts to initiate a flashover. This scenario canbe avoided by ensuring that there is a minimumseparation distance between adjacent rockbolts and metallic structures. In other words,the further the rock bolts are apart from oneanother, the greater is the potential differencerequired to generate a flashover.

If certain rock bolts and metallic structurescannot be practically separated by theminimum distance, the second strategy is tobond them galvanically together to ensurethat the potential difference across is not highenough to generate a flashover.

The presence of a steel-cased bore hole thatextends from the earth’s surface into theunderground space will increase the magnitudeof lightning-induced potential differencebetween the underground rock bolts (Fisherand Novak, 2001). Computational finiteelement methods can be employed to evaluatethe electric field distribution in theunderground space.

As mentioned earlier, the second mode oflightning propagation into an undergroundspace is through a direct lightning strike onmetallic elements at the entrance of the tunnelor shaft leading into an underground space.These metallic elements could be power cables,communication cables, metal pipes for drainageor fire-fighting applications, metallic ventilationducts or metallic supporting structures. A directlightning strike on these elements will causea high discharge current to flow in them. Asthe lightning current flows along the path ofthe metallic element, it will generate pointsof potential difference between the elementand its surrounding environment.

A transmission line model as illustrated inFigure 3 can be used to study the voltage andcurrent transients that will propagate alongthe stretch of the metallic element leadinginto the underground space. The metallicelement such as a stretch of structural steeloverhanging support is modelled as a series ofunit inductances: Ls and resistance Rs. Eachunit section is earthed with a resistance Re andhas a capacitance to ground Ce. Circuit transientanalysis software such as EMTAP or Pspice canbe used to calculate the voltage transient onthe metallic element at the end of theunderground tunnel, for the worst cases oflightning strike.

The model can be used to determine thenumber of grounding points that is necessaryto bring the conducted voltage transient atthe end of the circuit to a safe and tolerablelevel. This will ensure that the potential risewill not be high enough to generate aflashover.

A flashover will be generated when there is apotential difference between two points builtup beyond the electrical breakdown strengthof air. To mitigate this mode of lightning effects,the metallic elements must be intentionallygrounded at multiple sections to provide goodcurrent discharge paths as it stretches alongthe length of the tunnel from the entrance tothe storage space. Equi-potential bondingshould be applied to adjacent metallic elementson the wall and floor sections along the tunnelon ensure that the ground potential risebetween these elements is minimised. Thesection on ‘Grounding and Equi-potentialBonding’ shall elaborate more on this aspect.For power-carrying conductors andcommunications conductors which cannot begrounded practically, surge suppression devicesare installed at these service entry points toprevent the lightning-induced currents frompropagating down these conductors.

Another possible mode of lightningpropagation that warrants more study is howlightning will propagate through cracks in the

22

rocks. It is common for rock formations to havecracks either naturally occurring or artificiallymade during blasting operations to create theunderground space. Groundwater often flowsthrough these cracks and may form aconductive discharge path for lightning currentsto reach the underground space.

ELECTRICAL FLASHOVER HAZARDS

Electrical flashovers can be generated in manyways. The following are some known ways:

a. Damage to power cables along the tunneldue to rock fall

b. Arc initiation during the switching of circuitbreakers

c. Flashover across power-carrying bus-barsdue to insulation failures

d. Flashover across medium voltage transformerwindings due to insulation failures or surfaceconduction across conductive dust

e. Flashover across electrical terminals due tohuman error, rodents or vermin

f. Loose connections in electrical wiring causingoverheating and minor arcing betweenconnections – this may extend over a periodof time to cause air ionisation in an enclosurewhich could lead to a flashover to ground

g. Accidental energising of bus-bars whenmaintenance personnel are working onthem, or when tools are left in the switchgear compartments

h. Accidental energising of bus-bars when thegrounding switch of circuit breakers has notbeen opened

i. Overheating of motor control panelsresulting in insulation failures that mayprogress to flashover

The incidences in the above list can take placein any installation, and the effects will becatastrophic if the flashover ignites flammablegas or explosive dust in an ammunition facility.The design of the protection systems and themethod of installation for the electricalinfrastructure have to be such that they preventflashover hazards.

Figure 3. Circuit model of metallic element along the tunnel

23

Fundamental in the design of the electricalnetwork of an underground ammunitionfacility is the type of grounding system toadopt. The choice of grounding system isimportant as it affects not just the overallelectrical safety but also the lightning safetyof the facility.

A five-wire TN-S electrical grounding systemcan be used. It is suited to the distributionsystem in Singapore as it provides a protectiveearth conductor in each outgoing circuit,allowing the safe return of fault currents inthe case of any phase-to-earth faults. This thusprovides protection against direct and non-direct contacts. For such a TN-S electrical system,the low voltage network is solidly groundedat one point at the secondary winding of thetransformer. At the low voltage level,overcurrent protection and ground faultprotection devices should be installed toprovide two levels of circuit protection fromelectrical faults. The protective earth conductorprovides a continuous low impedance path forthe earth current to flow through, thuspermitting the positive action of the groundfault protection devices. However, this willinevitably also generate higher fault currents.

It is sometimes necessary for ammunitionfacilities to be sited deep underground toprovide the necessary protection. In suchsituations, electrical networks may stretchseveral kilometres. It is thus not uncommonfor medium voltage distribution of electricityto be used to overcome the problem of voltagedrop. It is important to define zones withinthe underground space, such that mediumvoltage electrical substations, distributionrooms, equipment and control panels arelocated at safe distances away from the zoneswhere ammunition is physically stored orprocessed. At the medium voltage level, pilotwire differential-relay protection should beincorporated to provide discrimination andtrip circuit breakers during real cable faults.

Active ventilation, fire suppression andcompartmentalisation systems should bespecifically designed such that an electrical fireor flashover hazard at the equipment zonesdoes not ignite any flammable compound orpropagate into the ammunition storage orprocessing areas. Electrical switchgears or circuitbreakers can be located in positive-pressurisedplant rooms and housed in metalcompartments that prevent the ingress of dust.

Figure 4. Illustration of a TN-S electrical earthing system

24

Gas-insulated switchgears are designed suchthat the bus-bars and the switching contactsare enclosed either in a vacuum or in SF6 gaspressurised tubes. They should be consideredfor application as they confine flashover arcsat the points of switching contacts within thepressurised tubes, thus providing additionalsafety compared to air circuit breakers.

ELECTROSTATICFLASHOVER HAZARDS

Lessons in the history of ammunition handlinghave shown us the risks of electrostaticflashover or discharge on components ofammunition systems. During the 1950s, the USNaval Ordnance Laboratory investigated theaccidental firing of a Mk112 type electric primerand concluded that some electric fuze typeprimers are extremely susceptible to initiationby electrostatic discharges due to the chargesaccumulated during the handling procedures(Kabik and Ayres, 1951). As modernammunition systems are increasingly integratedwith miniaturised electronic firing and controlfeatures, they may not be much safer undersimilar exposure to electrostatic discharge.

The environment in the underground space isusually warm and humid, with the relativehumidity level in some areas reaching as highas 99%. This benefits electrostatic discharge(ESD) control as static charges are easilydischarged from a body through the moist air.However, it is still important in the design ofthe underground facility that ESD controlmeasures are adopted. It is uneconomical toenforce ESD control measures throughout theentire span of the facility, thus specific zoneswould need to be identified for the handling,unpacking and processing of ammunition. ESDcontrol measures can thus be limited to thesezones.

The explosive safety code of the UKDepartment of Defence suggests a list ofcomprehensive ESD control measures for atypical above-ground facility and can besuitably applied in an undergroundammunition processing environment.

Safety features should be installed to facilitatethe discharge or ‘bleeding’ of body staticcharges. In zones for ammunition storage, air-conditioning equipment installed shouldmaintain the relative humidity level above60% so as to prevent the build-up ofelectrostatic charges. Personnel static dischargebars should be located at accessible locationse.g. the entrance of storage areas andammunition processing areas.

In zones of ammunition processing, anti-staticor conductive flooring can be installed as ameans of personnel discharge. Anti-staticflooring helps to discharge any charged bodyslowly to earth via a resistive layer. The use ofanti-static flooring for zones processingammunition with ignition energy between1mJ and 156mJ and the use of conductiveflooring for ammunition with ignition energyless than 1mJ are recommended (Joint ServicePublication, 2006). The UK ESTC coderecommends that anti-static flooring have asurface-to-earth resistance between 50k to2M , whereas the conductive flooring is tohave a surface-to-earth resistance of less than50k .

GROUNDING AND EQUI-POTENTIAL BONDING

Due to the poor conductivity of the rocks inan underground rock cavern, it will be difficultto obtain low resistance grounding throughoutthe whole facility. The choice of the groundingsystem adopted will affect the electrical andlightning safety of the underground space.The mining industry in the US is required byregulations to install safety grounds in minesthat are electrically isolated from the groundsof the mine substations. This is to ensure thatthe safety grounds are maintained at absolutepotential and not energised to a dangerouslevel during a lightning strike on incomingpower lines or during ground faults. One ofthe reasons is due to the fact that most minesextend far away from the substations. It isimpractical to ensure that the mine walls andfloors are well bonded to the power system

25

ground when there are no large metallicstructures, steel beams or embedded reinforcedconcrete that span the length of the minetunnels. However, electrical mining equipmentin these areas have ground conductors thatextend to the grounds of the substations.Mineworkers are thus prone to shock hazardsas they come between two elements tied todifferent isolated grounding systems.Maintaining two separate ground systemsensures that the potentials of both theequipment ground and mines’ floor are keptindependently at a safe absolute potential.

However, such a separation of groundingsystems does have its problems as discussed inCooley and Hill (1986 & 1988). An undergroundammunition facility, unlike a mine, is a long-term fixed installation. Multiple large metallicstructures are usually installed along the lengthof the tunnel to provide for ventilation, smokecontrol and other fire-fighting features. Inpractice, it is difficult to ensure that the twogrounds are not coupled, especially when theselarge metallic structures extend along thelength of an underground tunnel. Having twoseparate grounding systems in close proximitybut not interconnected will inevitably resultin ground to ground potential differences. Thismay lead to a shock hazard, which could belethal in the event of an electrical fault or alightning strike.

The type of method used to ground anunderground electrical system should becarefully considered. There are four possibleoptions: (1) ungrounded, (2) solidly grounded,(3) low resistance grounded or (4) highresistance grounded. Each type of groundingsystem offers its advantages and disadvantages.

An ungrounded electrical system will result insevere transient over-voltages as high as sixtimes the nominal system voltages. There isalso difficulty in locating and removing groundfaults in such a system as there is no directreturn for ground fault currents.

Solidly grounded electrical systems potentiallygenerate high fault currents, typically in the

order of 1,000A for low-voltage systems.Although such high fault currents have high-energy content that is sufficient to generatearc-flash hazards, they will also trip over-currentprotection devices more easily. Solidly groundedsystems do not generate high over-voltagesand allow the quick location of faults. Theyare more suited for a TN-S five-wire electricalsystem as they are intended for fault currentsto flow through a low earth loop impedanceto trip the ground fault protection devices.Solidly grounded systems ensure that thetransformer secondary windings are Wye-grounded to the substation ground and willcomplement the low impedance groundingfor lightning protection.

It has been a popular practice since the 1970sto adopt high resistance grounding in minesand petrol-chemical industries to reduce themagnitude of short circuit currents. Highresistance grounding reduces the magnitudeof fault currents to low levels (typically 5A to25A), thus preventing the destructive effectsof ground fault currents and reducing the riskof arc flash hazards. Since fault currents arelimited to such low levels, process plants cancontinue to operate without interruption.Therefore, high resistance grounded systemsoffer service continuity. A high resistancegrounding system is applied to three-phasethree-wire loads where phase-to-neutral loadsare not served, and thus it is not practical tobe used for a TN-S five-wire electrical network.Another concern is the conflict with lightningsafety-grounding practices. Lightning safetycodes prefer a low impedance grounding andbonding system to ensure that the potentialrise between two points is kept to a minimum.In the case of lightning-induced equipmentpotential rise, there may be a risk that highpotential differences may be built up acrossthe equipment chassis and phase conductors.A higher insulation level will hence berequired on equipment connected to theelectrical system (as commented by Dr AbdulMousa, 2008).

26

consists of metallic grid networks embeddedunder the concrete floors of the electricalsubstation as well as zones where ammunitionis stored or processed. The grounding gridsbelow electrical substations prevent hazardousground potential rise in the event of a phase-to-ground fault. In zones where ammunitionis to be processed or stored, the groundinggrids prevent ground potential rise whenlightning strikes propagate into theunderground space. Along the stretch of thetunnel, bare copper conductors are embeddedunder the road surface. At various intervals ofthe tunnel, bonding conductors provide equi-potential bonding between the buried copperconductors and the overhanging metallicservices. For storage areas, a Faraday Cageconcept is adopted. Metallic services are equi-potential bonded to entry points of the storageareas to further ensure that any lightning-induced currents flowing through would notcause ground potential rise at the storageareas. Figure 6 shows an illustration of suchan installation. Ammunition safety is further

It is thus more practical to install a commongrounding system and achieve as low aresistance to the ground as possible.

An extensive grounding system should bedesigned for an ammunition facility in therocks in order to prevent both lightning andelectrical flashover hazards due to unevenground potential rise. Figure 5 gives anillustration of an underground commongrounding system. The grounding system

Figure 5. Illustration of grounding system for an underground ammunition facility

Figure 6. Equi-potential bonding of metallic services atentry points before concrete walls are cast

27

Figure 7. Illustration of a ground electrode installation in high-resistive rocks

enhanced by storing ammunition in metalcontainers that will act as a second layer of anequi-potential Faraday Cage.

There are various ground enhancementmaterials that can be used to achieve lowimpedance grounding amidst high resistivityrocks. These include coke, bentonite clay andconductive concrete. Expected high rate ofcorrosion and groundwater wash-off must beconsidered in the choice of the groundelectrode installation adopted. In the tropics,the weathering effects on rocks make theselection of aluminum unsuitable as anelectrode material. Coke powder or chemicalenhancement materials are also unsuitable inunderground environments, as groundwatercan wash off such materials over time, andthey may eventually exhibit unstable resistivity(Switzer, 1998).

The impedance of a ground electrodeinstallation to remote earth can be imaginedto be similar to conductive onion-like layers ofequal thickness (Paschal, 2000). If the

conductivity of the materials immediatelysurrounding the metal electrode is improved,the overall earth impedance of the groundelectrode would be significantly reduced. Thisis more effective than extending the length ordiameter of the electrode.

Low impedance grounding can thus beachieved in the high-resistive rocks by drillingdeep bore holes at various sections of thetunnel. Each of these bore holes should havea copper electrode planted within and filledwith conductive concrete materials. The copperelectrode is further bonded at the top to thetotal grounding grid network. Figure 7 showsan illustration of such a ground electrodeinstallation. Bonding several groundingelectrodes in parallel to the grounding gridnetwork can further reduce the impedance toremote earth.

The presence of groundwater that washes offminerals and salts from the topsoil willcontribute to the weathering effects on rocks.A higher rate of metal corrosion may be

28

experienced in underground space. It is thusimportant to conduct regular checks on thecondition of the grounding and bondingconductors as part of the maintenance plan.

ZONING

As with the masterplanning of any majorfacility, the zoning of functional areas withinthe ammunition facility and the subsequenthazardous area classification are importantparts of the design consideration. Areaclassification provides a methodology foranalysing and classifying an environment whereflammable vapour, mist or dust may occur withthe expected operations of the area. Thiszoning template will help designers in theproper selection and installation of apparatusto be used safely in the environment.

JSP 482 and ATEX (‘Atmosphères Explosibles’)Directive (94/9/EC) on safety requirements forelectrical installations and standards ofprotection for equipment give a good guidein defining zones of hazardous areas. JSP 482recommends classification of areas based onthe type of explosive substance present andthe frequency of the presence of such explosivegases, vapour or dust. It also further definesthe type of protection for different EquipmentZones. Areas deemed to have a potential forthe release of explosive gas or vapour will beclassified as Category A. Category B facilitiesrefer to areas where there is a risk of explosivedust present in the atmosphere. Category Ccomprises all explosive buildings in whichexplosives do not give rise to flammable vapouror explosive dust at normal storagetemperature.

Further zone classification is assigned toCategory A and B areas in recognition of thediffering degrees with which explosiveconcentrations of gases, vapours or dust mayarise in terms of both frequency of occurrenceand probable duration of existence on each

occasion. Zone 0 is assigned to places where

explosives gases and/or vapour will be present

continuously or for long periods. Electrical

equipment should not be installed unless it is

absolutely essential and must comply with

ATEX Equipment Category 1. Locations where

explosives gas and/or vapour atmosphere will

be likely to occur in normal operation will be

classified as Zone 1 and the electrical equipment

installed shall be certified to meet ATEX

Equipment Category 2 standard. Zone 2 will

cover areas where a flammable atmosphere is

not likely to occur in normal operation but will

exist for only a short time if it does occur.

Electrical equipment will be certified to meet

ATEX Equipment Category 3 standard.

JSP 482 provides further reference to specific

types of protection appropriate for different

zones (e.g. intrinsic safety types for Category

A Zone 0 and flameproof types for Category

A Zone 1) as well as maximum surface

temperature requirements for the electrical

equipment in explosive storage. Special

consideration should also be given to allow

for the possibility of change in use of the area

which may require a different classification

and involve a different set of site compatibility

issues related to operational and maintenance

activities.

As discussed, areas within the facility can be

divided into Equipment Zones according to

the nature of the explosives that are stored or

handled as well as the processes to be

undertaken. Electrical installations and

equipment are then afforded the same

Equipment Category as the areas in which they

are installed or used. Equipment that meets

the high safety and explosion-proof standards

for Category A and B zones is often more costly.

The use of this zoning concept helps to ensure

that the appropriate equipment is used for

the different zones instead of the entire facility.

This will effectively reduce hazards from

electrical flashover and also ensure the

economical usage of the equipment.

29

REFERENCES

I. Kabik and J.N. Ayres. (1951). The initiationof Electric Fuze Primers by ElectrostaticDischarge, US Naval Ordnance Laboratory,Report 1762, pp 1-11.

John Paschal (September 2000). Not all GroundRoads are Created Equal. Published by ElectricalConstruction & Maintenance (EC&M) magazine.R e t r i e v e d o n J u n e 2 0 0 8 f r o mh t t p : / / e c m w e b . c o m / i m a g e s / a r c h i v e /0900groundrods.pdf

Keith Switzer, Erico Inc. (October 1998).Achieving an Acceptable Ground in Poor Soil.Published by Electrical Construction &Maintenance (EC&M) magazine. Retrieved onJune 2008 from http://ecmweb/grounding/electric_achieving_acceptable_ground/

Lightning Activity in Singapore, NationalEnvironment Agency Singapore. Retrieved onJune 2008 from http://app.nea.gov.sg/cms/htdocs/article.asp?pid=1203

ENDNOTES

Comment made by Dr Abdul Mousa, Ph.D., P.Eng., Fellow IEEE, Lightning protectionconsultant, Vancouver, Canada, Co-moderatorof Lightning Safety & Power Quality IssuesInterest Group.

CONCLUSION

Lightning, electrical and electrostatic flashoverhazards could occur in ammunition storagefacilities and those built underground are noexception. Safety measures such as potentialequalisation and sound grounding could beimplemented to mitigate these risks. Thechallenge is always in the application of thesemeasures to the actual facility, especially whenit is underground and houses complex servicesand systems. Prior planning to classify thefacility into different functional zones wouldprovide a focused approach to meet therequirements of each area. Finally, good designand implementation would need to beconsistently complemented with soundoperation and maintenance practices to keepthe hazards at bay and ensure that theammunition facility remains safe for storage,operations and other related activities.

Marvin E. Morris, Cliff A., Chua H.K. and TehS.P. (2002). Using the Faraday Cage Conceptfor Substantially Improved Lightning Protectionof Critical Facilities: Measurement of theLightning Response of Explosives StorageStructure 7919 Using the Transfer ImpedanceMeasurement Instrumentation System.International Symposium on DefenseConstruction, 2002.

Novak T. and Fisher T. (2001). LightningPropagation through the Earth and its potentialfor Methane Ignition in Abandoned Areas ofUnderground Coal Mines. IEEE Transactions onIndustry Applications, Vol 37, No.6,November/December 2001, pp 1555-1562.

Sandia National Laboratories (2007). Sandiaresearch indicates that lightning was the likelycause of Sago Mine explosion. Retrieved onJune 2008 from http://www.sandia.gov/news/resources/releases/2007/sago.html

UK, Joint Service Publication (JSP) 482. (2006).Safety Standards for Electrical Installations andEquipment in Explosive Facilities. Vol 1. Edition2, Change 3, Chapter 8.

Wils L. Cooley, Herman W. Hill, Jr. (1986).Coupling of Mine Grounds to Surface Grounds.IEEE Transactions on Industry Applications, Vol22. No. 2, March/April 1986, pp 360-364.

Wils L. Cooley, Herman W. Hill, Jr. (1988). MinePower System Grounding Research. IEEETransactions on Industry Applications, Vol 24.No. 5, September/October 1988, pp 846-852.

This paper was first presented at the 2008 USDepartment of Defense Explosives Safety BoardSeminar on 12 - 14 August 2008 in Palm Springs,CA and has been adapted for publication inDSTA Horizons.

BIOGRAPHY

30

Chua Hian Koon is Assistant Director (Mechanical and Electrical – ProtectiveInfrastructure) and oversees the strategic development and implementationof mechanical and electrical engineering and operational masterplans fordefence infrastructures. He is the Deputy Chairman of the Technical ReviewCommittee for Code of Practice for Lightning Protection, Singapore. HianKoon was part of the team that was awarded the Defence Technology Prize(DTP) 2006 Team Award for their work in protective technology research. APublic Service Commission scholar, Hian Koon received Dip D'Ingenieur inElectrical Engineering (First Class Honours and Master) from the Ecole NationaleSuperieure D'Ingenieurs Electriciens de Grenoble in 1988, and also holds ajoint Master in Engineering Science from the Ecole Centrale de Lyon.

Ho See Fong is a Senior Engineer (Mechanical and Electrical – ProtectiveInfrastructure). He is involved in the development and project managementof electrical networks in MINDEF facilities and advises on the lightning safety.He has also been involved in projects exploring new technologies in Energyand Power using fuel cells. He received his Bachelor degree in Electrical andElectronic Engineering from the Imperial College, London, United Kingdom,before furthering his training to attain a Master of Science in Electric PowerEngineering from the Royal Institute of Technology, Sweden.

Teh Siaw Peng is a Programme Manager (Mechanical and Electrical – ProtectiveInfrastructure) and oversees the engineering masterplanning, design andimplementation of essential electrical systems for the Ministry of Defence(MINDEF) and the Singapore Armed Forces. He is a member of the TechnicalReview Committee for Code of Practice for Lightning Protection as well asa registered Professional Engineer (Electrical) with the Professional Engineers'Board. Siaw Peng received his Bachelor degree in Electrical Engineering fromNanyang Technological University.

31

Lim Jiunn Shyan is a Senior Engineer (Mechanical and Electrical – ProtectiveInfrastructure). He is involved in the planning, design and implementationof protective electrical systems and building management systems in MINDEFfacilities. Jiunn Shyan also contributes to the study and implementation oflightning protection systems for MINDEF facilities, as well as the developmentof lightning protection for some of the equipment onboard air and landplatforms. He was part of the team that was awarded the DTP 2007 TeamAward for their work on the Underground Ammunition Facility. He receivedhis Bachelor degree in Electrical and Computer Engineering (Honours) fromthe National University of Singapore in 2005.

Creating and Valuing Flexibility in SystemsArchitecting: Transforming Uncertaintiesinto Opportunities Using

Real Options Analysis

ABSTRACT

Flexibility is one of the strategies to ensure the value delivery

of a System of Systems (SoS) during its life cycle in light of the

changing external environment. This article discusses the notion

of ‘Real Options Thinking’ to create flexibility and ‘Real Options

Valuation’ to value flexibility when designing an SoS to manage

significant uncertainties. We propose a ‘Real Options Analysis’

framework consisting of the dual parts of ‘Real Options

Thinking’ (akin to good management intuition) and ‘Real

Options Valuation’ (akin to stochastic optimisation) to promote

a common ‘options language', enrich vocabulary, sharpen our

thinking and guide quantitative analysis when managing

technical projects. Adopting ‘Real Options Thinking’ and ‘Real

Options Valuation’ will increase our capacity to identify, create

and secure both technical and managerial options through

deliberate choice, in a cost-effective and timely manner. It will

also increase our tolerance of uncertainties and empower us

to actively manage uncertainties to create opportunities and

reduce risks.

Angela Ho Wei Ling

Ng Chu Ngah

34

Creating and Valuing Flexibility inSystems Architecting: Transforming Uncertaintiesinto Opportunities Using

Real Options Analysis

“Uncertainties are a source of risks andopportunities” (de Neufville, 2004).

Traditionally, for large-scale complex systems,uncertainties are treated as risks and aretherefore undesirable. The approach towardsuncertainties is consequently to manage themthrough risk mitigation. However, ‘Real OptionsThinking’ introduces a paradigm shift in lookingat uncertainties as sources of opportunities aswell. This paradigm shift in thinking isessentially the act of capitalising on theopportunities embedded in uncertainties whilelimiting the extent of risks involved. Thisviewpoint is critical as most current engineeringpractices are conservative and may not exploitthe potential upsides of uncertainty. We wouldbe able to perform better if we can develop adynamic strategy in anticipation ofuncertainties arising from external factors.Through an SoS architecture that can evolveand adapt to a changing environment, we willincrease our potential and ability to capitaliseon emerging opportunities throughuncertainties. Real Options Analysis is thus anActive Uncertainty Management technique asopposed to Passive Uncertainty Managementtechniques like risk management.

FLEXIBILITY IN SOS ARCHITECTING AS A STRATEGY TO ENSURE SOS VALUE DELIVERY

Systems architecting is an approach to designand build effective and efficient SoS. The SoSarchitecture "provides the structure or skeletonof the system, as well as the principles, rulesand guidelines governing the system design,creation and evolution. It also provides thebroad framework, system level constraints aswell as the relationship for the substructuresand modules of the system. It determines theoption available for future development." (Tan,Yeoh, Pang & Sim, 2006). In order to preservevalue over its life cycle, an SoS has to be ableto handle dynamic complexity and a changingoperational environment.

INTRODUCTION –UNCERTAINTIES ASBOTH RISKS AND OPPORTUNITIES

In the context of Singapore defence, our reallife complex systems are invariably Systems ofSystems (SoS) which always operate in uncertainand dynamic environments. Typically, whilesystems architects design an SoS, programmemanagers develop an SoS that will operate inan uncertain and dynamic environment. Thedrivers of uncertainties can be both externaland internal. External factors such as emergentthreats, new operational concepts, disruptivetechnologies and shifts in supplier industrystructure are largely beyond our control. Onthe other hand, internal factors come fromuncertainties in the programme delivery andare generally well managed by current projectmanagement practices. The approaches andmethods to deal with uncertainties arisingfrom external factors are not well understoodtoday. In addition, there is an increasing needto design and deliver SoS in fast-changingenvironments. Thus, it is essential to anticipatethese known unknowns and, as far as possible,prepare for this class of uncertainties.

History shows that we are poor forecasters ofexact trends. Over-confident forecasts like“640kb ought to be enough for anyone” fromBill Gates underestimated the uncertainty indemand and technology. Designing an SoS tospecific trends may turn out to be costly as asingle ‘unknown unknown’ can throw off ourpredictions and ability to react in the future.We would be better off predicting a range ofscenarios and developing the feasible solutionspace containing all flexible on-demandresponses that we can choose to adapt as newinformation becomes available with time,rather than designing our SoS to forecastedtrends.

35

Flexibility is one of the key strategies to copewith environmental changes and dynamiccomplexity, and Real Options Analysis is oneway to tangibly create and value flexibility inSoS. A working definition of ‘flexibility’ is asfollows:

An SoS is flexible if we have the freedom tomake our choices during life cycle operationto cope with the large external changes.Flexibility can be created and valued in systemsarchitecting and designed using the RealOptions Analysis framework.

Breakthrough engineering solutions have acommon thread of incorporating flexibledesigns to cater to wide-ranging scenarios.These flexible designs are also known as‘options’ and people have been practising ‘realoptions thinking’ even though the languageof options may not be as ubiquitous. A goodexample is the flagship UndergroundAmmunition Facility (UAF) project, managedby DSTA, for high-density ammunition storage.DSTA has invested in the research anddevelopment (R&D) on protective infrastructureand related technologies for years to createoptions for MINDEF to address various defenceneeds. When the new challenge of creatingnew ammunition storage for the SAF arose inthe face of tremendous pressure to free upland for national development, DSTA was readyto exercise the option and embark on thebuilding of our first UAF. Apart from derivingdirect value from the options created throughR&D, this effort also generated emergentbenefits by allowing DSTA specialists to setnew safety standards, including obtaining theNorth Atlantic Treaty Organisation’s acceptanceand becoming a leader in the field.

Flexibility is the life cycle propertythat allows an SoS to endure sets ofchanges with ease. It is an active andlargely external approach to managingchange (adapted from Moses, 2003).

DEFINING REAL OPTIONS

The science of options analysis began withfinancial options. A financial option gives onethe right but not the obligation to buy an assetlater at a pre-determined price. This meansone can be better off during good times andhave a fallback position during bad times.Should the value of the asset increase, one canprofit from it. Should the value of the assetdrop, one’s losses are limited to the optionpremium. Black and Scholes (1973) developedthe theory of pricing financial optionswhile/and considering uncertainty, decisionspaces and time in the equation. Option pricinghas led to flexible financial structures, createda market of options transactions and reducedthe volatility of the commodities.

A real option is the extension of the idea tovalue the flexibility of management decisionson a real or tangible asset. Asset owners needto know the price of their productive asset todetermine its economic value. Since an assetcan be put into creative use in multiple waysand the management has the flexibility tocontrol how much resources to invest orwithhold from it in the future, depending onthe future demand – the value of the assetdepends on their own possible courses ofaction. This viewpoint changed the perspectiveof asset valuation from historical cost valuationto prospective contingent valuation. The valueis contingent on the future possible actions ofthe management. Today, real options analysisis being accepted as a conceptual and analyticaltool to support strategic decision making underuncertainty by extending existing techniquesof Net Present Value valuation. It is a bridgingtool for strategic and financial decision makers.

It is important to distinguish an ’option’ fromour common understanding of it as an‘alternative’ or a ‘choice’. An ‘option’ is notanother ‘alternative’ or ’choice’. Instead, itrefers to our right, and not the obligation, to

36

take an action (i.e. exercise it) at some pointin time with an upfront cost. An option is asecured choice that makes it available ondemand.

A simple example of a real option inengineering design is the option to ‘expand’(the action) in the design of the bridge acrossthe Tagus River in Lisbon, Portugal. In the1970s, the first bridge was designed forvehicular traffic when there was no demandfor commuter railway system. However, thegovernment insisted that the bridge structurebe strong enough to carry rail traffic. Twentyyears later, there were changes in technologiesand in rail demand (pre-determined time toexercise options) and the government was ableto exercise its option (option to expand) andextend rail service on the bridge. There wasan upfront cost for this option – the bridgehad to be reinforced when it was built (pre-determined cost). However, when theenvironment changed in the 1990s, Portugalwas able to take advantage of this opportunityto exercise the option that it had built into thetechnical design of the bridge 20 years earlier(Gesner & Jardim, 1998).

At face value, we may view such an exampleas simplistic. We caution readers to checkagainst having any hindsight bias. We shouldreflect on the uncertainties and constraintsthat the decision makers would have faced inthe context of their time. Additionally, thedecision equation will be more complicatedwhen there are several sources of uncertaintyand only a few options for judgement.

A real option is a technical or managementchoice that we secure today at a

pre-determined cost for a pre-determinedtime to have the right to exercise when

needed without any obligation(de Neufville, 2001).

REAL OPTIONS TYPES

A flexible SoS must have both technical andmanagerial options built into the architecturethat can be exercised when needed. Both ofthese options are classified as ‘Real Options’in our engineering SoS and are different fromthe Acquisition Options in procurement thatwe are more familiar with.

Technical Options are options that are createdin the design of the technical SoS itself.Technical Options are called Real Options inSoS (de Neufville, 2002) as they are embeddedin the technical design of the SoS architecture.Identifying technical options within the SoSwould require a good understanding of theSoS and modular architecture. A classic technicaloption is the design of a dual operating mode.For example, straight stretches of roads canbe designed and built so that we can use themas runways for aircraft. However, this meanswe have to design the roads with somelimitations for transport use while ensuringheavier load requirements and removablebarriers.

Managerial Options are options that arecreated to manage the process of SoSdevelopment and operation. They treattechnology as a ‘black box’ and are essentiallyfinancial options taken on technical projectsand SoS. Managerial options are the RealOptions on SoS and are enabled by thetechnical options, contractual obligations ofsuppliers as well as financial and resourcecontrol over the process of SoS developmentand operations. Classic managerial optionsinclude options to defer a decision, alter theoperating scale of the SoS or abandon somesub-initiatives. Back to the road-runwayexample, the management has the option touse roads as aircraft launch pads if enabled bybuilt-in technical options.

Real Options allow the SoS to adapt tochanging scenarios over time. This flexibilityincreases our ability to capitalise on upside

37

opportunities and to limit our exposure todownside risks. Flexibility that is embedded inthe technical design and management of SoSis important to improve the operational andtechnical effectiveness. The greater theuncertainty, the higher the value of each realoption.

Both technical and managerial options comeat a cost known as the ‘option premium’. Theoption premium can be seen in two ways: asthe maximum price we should pay to have themanagerial and technical flexibility, or as theprice of our uncertainty. Certainly, we will notpay an option price that is higher than the costof the uncertainty. When we have a bundle ofoptions available across several review pointsin future, we can use a quantitative way todecide how much flexibility to incorporate intothe SoS design and develop the roadmap ofdecisions.

Real options in SoS are thus an additionaltechnique that can be used to value theflexibility of technical choices on SoS design.It is impossible to execute many of themanagement choices to expand capability orswitch operating modes unless the initial design

had the benefit of forethought. Securingtechnical options is costly and is subject tomuch inquiry because they will not be perceivedas required, unless we measure the cost againstthe probable scenarios.

LANGUAGE OF REAL OPTIONS THINKING

Technical and managerial options can bebroadly classified by whether they reducedownside exposure to risks, allow status quoor leverage upside opportunities, dependingon how uncertainties evolve. If uncertaintywas deemed to have a negative impact,exercising real options like the option todownsize or mothball would reduce exposureto risk. On the other hand, if the uncertaintyturned out favourably, exercising real optionsthat increase the exposure to theseopportunities would reap a greater return.Table 1 groups some of the possible options(Trigeorgis, 2001) according to this classification.In the event where more information mightchange the course of action, the option toremain status quo might have to be deliberatelycreated.

Table 1. Types of Real Options

Function Options Description

Option to To alter operating scale and reduce capacity by downsize removing features and resources

Option to To temporarily remove from active service and put mothball into protective storage

Option to To stop and cut losses, typically when projects terminate become unprofitable

Option to defer To postpone starting or initiating an investment

Option to continue To maintain status quo

Option to switch To move to an alternative mode of operationor design

Option to expand To alter operating scale and expand capacity by adding features and resources

Option to grow To invest in an option so it may open new options

Option to restart To restart a temporarily closed operation

Reducerisk

exposure

Status quo

Leverageopportunities

38

REAL OPTIONS EXAMPLES

Real Options Thinking can be applied in

different scenarios of uncertainty. In the

example of the bridge in Portugal, the

uncertainty lies in the demand profile. The

Portuguese government decided to create a

real option which would become valuable if

the demand for a commuter railway systemindeed arose. This real option gave them thereadiness to switch quickly, but there wasnaturally an option premium involved whichincluded the cost of the extra load. Table 2provides other case examples where realoptions can be applied and briefly illustratesthe benefit and cost impact of theseconsiderations.

Table 2. Examples of Real Options

No Uncertainty Without With Example Benefit Cost (Option Optionoption option premium) Type

1 Demand Single Dual Roads as Readiness Operator Technical:profile may equipment use of runways to switch training, Switchchange mode equipment quickly cost of

acquiringand operatingboth modes

2 Demand level Right Spare Extra Handle Cost of under Technical:is volatile sizing equipment network small -utilisation Expand

capacity bandwidth demand and cost ofchanges spare

capacity

3 Likelihood of No Redundant Critical Resilience Original and Technical:failure of redundancy equipment equipment and recovery backup cost Switchoperating as backup for 24/7equipment readiness

4 Technical Go (high R&D to Defence R&D Readiness to R&D cost Technical:feasibility; stakes) or investigate to explore develop into GrowDemand No-go feasibility new capability project if

(forgo any and gain feasible, orpotential) subject else exit with

knowledge no furthercost

5 Unknown user Waterfall Spiral Prototype Better risk Management Management: requirements development development development management cost is higher Expand,

and technical life cycle life cycle and customer Downsize,uncertainty satisfaction Terminate,

Continue

6 Future Buy now Lease with Rent capacity Stop or Higher total Management:demand option to with option continue at cost Defer

purchase later to purchase favourableat lock-in price later price

7 Availability of Purchase Purchase with Buy x Fix the price Higher total Management:new suppliers reduced option to platforms today cost than Continue,

set now or buy more with option upfront buy Expandall now at lock-in price to buy y

more later iffavourable

8 Different Customised Standardise Multi-platform Mass Design cost Management:requirements platforms base platform missiles, basic customisation for multiple Switchfrom different and modules interfacesusers software

package

39

CHARACTERISTICS AND APPLICATION OF REAL OPTIONS

Real Options has its pros (+) and cons (-).

+ Real Options enable the staging of decisionsin roadmaps. We create flexibility by buildingin decision review points in the future and bydefining the conditions under which theyshould be exercised. For example, an iterativedevelopment life cycle is superior to a waterfalllife cycle when we explore new technicalconcepts. The deferred review points give themanagement the option to scale, hold orterminate the project based on the informationfeedback as uncertainty unfolds. It is theflexibility to stage the decisions that potentiallycreates more value than a waterfall approach.Similarly, a dual-purpose road is quickly usedas a launch pad when an emergency needarises. Instead of formal timed reviews, theoption to switch back and forth always existsduring the life cycle of the project.

+ Real Options increase in value as uncertaintyincreases. The greater the uncertainty in aparticular undertaking, the greater the needto have options, and the higher thecorresponding value of the options we create.For example, the additional investment costof extra capacity in a complex network willbecome viable once the uncertainty in thedemand crosses a certain threshold.

- Real Options increase initial costs. Optionsadd to the baseline configuration of the SoSand increase costs. Options are identified atthe start of the project and hence, a higherinitial upfront cost will need to be factored into acquire the option at the ‘option premium’.For example, suitable civilian transportationresources are requisitioned for military purposeswhen needed. However, there are certain

administrative and periodic trial costs to ensurethat the suppliers and equipment are readyfor the activation. This option premium is theprice we pay for the flexibility to manage theuncertainty.

In view of the essential characteristics of RealOptions, we should realise that:

• Options are derived from the SoS. An optionmay be a logical or physical addition to thebase SoS. This concept is a useful guide whenwe start to identify options and expand ouroption or solution space. For instance, amissile that must have multi-platform launchingcapability or customised software built upona baseline module needs careful designof interfaces.

• Options are choices that are secured today.Unless we invest resources to secure thealternative that we may need later, it remainsan unrealised possibility. This notion is prettyclear for technical options as they have to beincorporated in the initial design. However,this notion can be quite subtle for managementoptions: a procurement option is ‘purchased’today as an option premium so that we canchoose to exercise it later at a pre-agreed price.

• Option investments have to be balanced.Options are secured at a price and we mustinvest in an economical portfolio that balancestheir benefits against their costs. This notionbecomes apparent during front-enddevelopment when there are several sourcesof uncertainty with several types of optionresponses.

REAL OPTIONS ANALYSIS METHODOLOGY

This section walks through the major steps inrealising flexibility in SoS. The Seven-StepMethodology for Systems Architecting shown

40

in Figure 1 is an iterative and recursive processthat moves from establishing systemsarchitecting objectives to reviewing of thearchitecture. While flexibility is one of therequirements in painting the big picture, RealOptions Thinking and Real Options Valuationwill enhance the generation of solution spacei.e. Build the Big Picture (Step 3) andIdentify Capability Gaps (Step 4).

Figure 2 is our proposed Real Options AnalysisMethodology which outlines the key steps increating and valuing the real options. Thejoined Red circles are the scenarios atconsecutive time stages (x-axis). The uncertainty

of the future naturally increases over time. Thesmall floating circles are the potential responses(i.e. decisions, secured choices or real options)that management and the technical team cantake for each scenario. Both scenarios andresponses are dependent on preceding choicesand remind us of the legacy effect we maycreate.

1. Determine Evolving Requirements – Thefocus of Step 1 of Real Options Analysis is todetermine evolving requirements that canaddress uncertainties in the future. This requiresa hard look at multi-time stages (e.g. in Figure2, we illustrated it with three time stages on

Figure 1. DSTA Seven-Step Methodology for Systems Architecting(Tan, Yeoh, Pang & Sim, 2006)

Figure 2. Real Options Analysis Methodology for Active Uncertainty Management

41

the x-axis) and the probable scenarios at eachtime stage. We consider the technologycapability today vis-à-vis projected uncertainties(i.e. known unknowns) coming from possiblethreat evolutions, key technology trends,changes in our operational concepts, operatingenvironment and stakeholders.

2. Conceptualise Architecture – The art ofsystems architecting is a subject on its own. Itis likely that there are several competingarchitectures proposed to meet the evolvingrequirements. Apart from functional needs,the systems architect has to trade off thevarious strategies for coping with dynamiccomplexity. This develops insights into thephysical and non-physical aspects of the SoS,sources of changes and the dynamic behaviourof the SoS. The architectural descriptions shouldallow for a qualitative understanding and aquantitative inquiry for the valuation ofoptions.

3. Real Options Thinking to expand solutionspace – Real Options present the most flexibilityand value in areas of greatest uncertainties.After projecting where the biggestuncertainties lie (i.e. Step 1), we should explorethe option/solution space and identify wherereal options can be created in the SoS. It isimportant to identify and clarify bothmanagerial (i.e. Real Options on SoS) andtechnical options (i.e., Real Options in SoS).

4. Real Options Valuation – After identifyingand creating the real option space, the realoptions are modelled by varying the designsand consequences. The steps are outlined inFigure 3. Specifically, there are two models tobe developed – the Ops/SoS model and theOptions Valuation model.

a. Ops/SoS model – The Ops/SoS model is anOperational Analysis model that eitherproduces a suitable response for a giventhreat scenario and/or generates the threatscenario that an SoS can handle given aspecific set of constraints. It requires anoperational and systems understanding ofthe specific engagement. At this level, a low

resolution analytical or simulation modelthat transforms the key design inputs tomeasurable outputs is sufficient. We will usethe Ops/SoS model to populate the responsedomain for the set of threat scenarios,and then proceed to compute the transitioncosts between sequential responses.

b. Options Valuation model – Given the set ofthreat scenarios, responses and transitioncosts, we proceed to define a Measureof Effectiveness (MOE) for evaluatingdifferent roadmaps and possible constraintsthat preclude certain roadmaps. Anoptimisation model that considers theuncertainty and time will facilitate anycombinatorial search. A point to highlightis that the solution is a recommendation ofa suitable roadmap for the entire set ofthreat scenarios, and not just a response toa particular threat scenario.

c. Roadmap – In practice, a first look at theroadmap is likely to trigger further inquiryand adjustments of both choices and inputdata. Once refined and stable, the roadmapis a guide for action. The roadmap chosenis the ‘optimal’ roadmap with the highestMOE score that satisfies the given constraints.It is important to note again that theroadmap is not a single path but a set ofpossible paths.

d. Tradespace – The tradespace plots theevolution of the optimal roadmap againstits cost and effectiveness. We can compareother roadmaps that are heuristically chosento understand the differences.

e. Option value – The roadmap is derived froma probability tree. The MOE of a roadmap isan average value based on the responsesfor the given scenario. To have a goodunderstanding of the variability of the MOE,we plot its probability distribution. We canthen compare it with any other heuristicallyderived roadmaps. The difference willcorrespond to the value of the options thatwe have built in.

42

5. Invest in Real Options – The managementis presented with the results, starting with theroadmap for the responses to the set ofpostulated scenarios. They will have to decidewhich real options to invest in, in order tosecure the right to respond in the future. In

practice, the roadmap will trigger additionalquestions on new options to be built today sothat we can secure the future we want. Itwould likely be necessary to go back to earlierstages to reiterate the process.

Figure 3. Real Options Valuation Process

43

recommend a roadmap that triggers furtherthinking. Together, we call this the Real OptionsAnalysis framework. Adopting Real OptionsThinking and Real Options Valuation willincrease our capacity to identify, create andsecure both technical and managerial optionsthrough deliberate choices in a cost-effectiveand timely manner.

We hope to promote a common ‘optionslanguage', enrich vocabulary, sharpen thinkingand guide quantitative analysis whenmanaging technical projects. Having optionswill always be useful especially if they spreadover space and time and both the managementand the project management team can exercisethem under appropriate circumstances.

Flexibility is a critical consideration that willenable us to deal with uncertainty throughActive Uncertainty Management. It is activebecause we would be engaged in exploringthe real options that can be designed into theSoS. As real options are designed to beexercised in the future, we would also needto think about when these options should beexercised. It is a dynamic strategic planningexercise with the aim of securing the future,whichever way it turns out to be. It is also astrategic decision making tool for SoS architects,the management and project managementteams.

6. Review Real Options – During operations,we will review if the real options investedearlier need to be exercised or deferred. If realoptions were built into the SoS, themanagement can and does have the right toexercise the options when the situation callsfor it. Exercising the correct real options canonly be done after re-evaluating their currentbenefits, and reassessing their impact as moreinformation is revealed with time. Theperceived value of each real option will becomehigher or lower over time as more informationis revealed and uncertainty decreases. We haveto keep reassessing the benefits and impact ofeach option at major decision points in orderto decide which options should be exercised.

CONCLUSION

We will continue to design and build large-scale and complex SoS to face a future that isuncertain. By leveraging the input from pastprogramme experiences and subject matterexperts, we can broadly map out possible futurescenarios. Uncertainty creates both risks andopportunities and our SoS has to remain flexibleso that we can exercise the choices we wantin response to future changes. We need tobuild real options into our technicalengineering SoS that will enable such securedchoices on demand. Real options are bothtechnical and managerial in nature. Whiletechnical options are pre-built into the SoSupfront, managerial options give us themandate to execute the real optionswith time, but without obligation.

A fresh perspective on flexibility is enhancedby Real Options Thinking and Real OptionsValuation. Real Options Thinking is akin togood management intuition and gives focusby listing the canonical management choicesand facilitating the identification and creationof real options. Real Options Valuation is akinto stochastic optimisation which helps us tocompute the maximum economic price thatwe should pay for the uncertainty and

REFERENCES

44

Black, F. & Scholes M. (1973). The Pricing ofOptions and Corporate Liabilities. The Journalof Political Economy, Vol. 81, No. 3 (May - June,1973), pp 637-654.

de Neufville, R. (2001). Real Options: DealingWith Uncertainty In Systems Planning AndDesign. 5th International Conference on"Technology Policy and Innovation", paper forpresentation, June 2001.

de Neufville, R. (2002). Class notes for“Engineering Systems Analysis for Design”ESD.71. MIT Engineering School-Wide Electivetaken in Fall 2004. Cambridge, MA.

de Neufvil le, R. (2004). UncertaintyManagement for Engineering Systems Planningand Design. Engineering Systems Monograph,MIT. Retrieved on 10 February 2007 fromh t t p : / / e s d . m i t . e d u / s y m p o s i u m / p d f s /monograph/uncertainty.pdf

Gesner, G. and Jardim, J. (1998). Bridge withina Bridge. Civil Engineering, October, pp 44-47.

Moses, J. (2003). The anatomy of large scalesystems. ESD Internal Symposium. Cambridge,MA.

Tan Y.H., Yeoh L.W., Pang C.K., Sim K.W. (2006).Sy s tems Arch i tec t ing fo r 3G SAFTransformation, DSTA Horizons 2006,pp 36-49.

Trigeorgis, L. (2001). Real Options: An Overview.Chapter 7 of Real Options and InvestmentUnder Uncertainty: Classical Readings andRecent Contributions. Ed. by: Schwartz, E.S. &Trigeorgis, L. MIT Press, pp 103-134.

ACKNOWLEDGEMENTS

We would like to thank Tan Yang How, TeoSiow Hiang, Lim Hang Sheng, Ang Choon Keatand Lee Keen Sing for their ideas andsuggestions during our discussions. We wouldalso like to acknowledge Palvannan RKannapiran, an ex-colleague who had helpedto develop the ideas in this paper. Lastly, wealso acknowledge and reference ProfessorRichard de Neufville of Massachusetts Instituteof Technology for his seminal work in RealOptions Analysis in large-scale system design.

BIOGRAPHY

Ng Chu Ngah is an Analyst (Operational Analysis and Simulation, DMSA). Sheis currently involved in the structuring and modelling of complex systemsand processes through the use of operations research, simulation techniquesand combat models so as to provide an objective basis for decision making.After receiving the DSTA Scholarship in 2003, she attained her Diplômed'Ingénieur from Ecole Nationale des Ponts et Chaussées, as well as herBachelor (First Class Honours) and Master degrees in Industrial and SystemsEngineering from the National University of Singapore.

45

Angela Ho Wei Ling is an Engineer with a portfolio spanning across EnterpriseIT and DSTA Masterplanning & Systems Architecting (DMSA). In EnterpriseIT, she has a key role in facilitating the Business Process Transformation andEnterprise Architecture for large-scale, complex Enterprise IT projects of theMinistry of Defence and the Singapore Armed Forces. In DMSA, she is a keydriver in developing Systems Architecting Methodologies. A recipient of theDSTA Overseas Undergraduate Scholarship, she graduated with a Bachelordegree in Computer Science (Honours) with a focus on Human-ComputerInteraction from Carnegie Mellon University. She also earned a Master ofScience in the Technology Policy Programme, with a focus on Aeronauticsand Human Factors, from the Massachusetts Institute of Technology.

Automated Red Teaming:

A Proposed Simulation-based Framework

ABSTRACT

Manual Red Teaming (MRT) is a traditional, human-intensive

analysis effort to uncover system vulnerabilities or to find

exploitable gaps in military operational concepts. The overall

goal is to reduce surprises, improve and ensure the robustness

of the Blue operational concepts. In this paper, an Automated

Red Teaming (ART) framework is proposed. It is a simulation-

based concept that uses Evolutionary Algorithm, Parallel

Computing and Modelling & Simulation techniques to

complement MRT and enable red teaming to be conducted in

a more automated fashion. The ART framework was

experimented in two maritime security scenarios, and the

results were analysed to demonstrate the capability of the ART

framework vis-à-vis MRT i.e. automated versus manual red

teaming effort. The evaluation showed that, in general, results

obtained from ART were better than those from MRT, some

of which were non-intuitive and surprising solutions.

Victor Tay Su-Han

Choo Chwee Seng

Chua Ching Lian

48

INTRODUCTION

Red teaming is a technique commonly used inthe military Operational Analysis communityto uncover system vulnerabilities or to findexploitable gaps in operational concepts, withthe overall goal of reducing surprises,improving and ensuring the robustness of theBlue operational concepts (Upton andMcDonald 2003; Upton et al 2004). It is currentlya manually intensive process that typicallybrings together experts relevant to the systemand scenario under consideration to identifysystem weaknesses. However, the vulnerabilityassessments made are usually ‘bound’ by theknowledge of these subject matter experts.

Advanced technologies such as ParallelComputing and Evolutionary Algorithms (EA)can be leveraged to automate the red teamingprocess. Parallel Computing allows millions ofsimulation runs to be generated andinvestigated in an automated fashion. EAs use

an iterative process, inspired by biological

mechanisms of evolution, to optimise a certain

fitness value that can be considered to be the

objective function. The growth of the

population in an EA is determined through a

set of operators such as recombination,

mutation and selection.

EA was identified as the search algorithm for

the ART framework. The goal of the search is

to fix the Blue parameters and search for Red

parameters that will result in the ‘defeat’ of

Blue, within the least amount of time.

Information obtained during this process can

then be used to either enhance or assist the

manual efforts.

ARCHITECTURE DESIGNOF THE ART FRAMEWORK

The architecture of the ART framework consists

of several key components (see Figure 1). The

architecture was designed to be modular and

flexible enough to incorporate new simulationmodels (non man-in-the-loop type) and the

Automated Red Teaming:A Proposed Simulation-based Framework

Figure 1. Architecture design of the ART Framework

49

following EAs (shown in Figure 1, clockwisefrom top left):

The ART parameter setting interface allowsthe initial selection of the parameters thatare to be varied.

The simulation model dependent modulesadd a layer of data flow to and from theART framework and simulation models.Data flowing into the simulation modelswill be the parameters to be executed andthe data outflow will be the results fromthe simulation runs. This data is translatedwith specially created wrappers from thesimulation format to the ART frameworkdata structures.

The EA module houses the EA library fromwhich the user can choose. The EA modulecurrently contains the Strength ParetoEvolution Algorithm Version 2, anEvolutionary Multi-Objective optimisationalgorithm. The library is also expected toexpand with the addition of otheralgorithms. The role of the EA module is toprepare the parameters for the individualsimulation, analyse the results and distill thedesired red teaming objectives. The role ofthe execution is handled by the Condorcontroller.

Condor is a special ised workloadmanagement system which provides jobqueueing, scheduling policy and resourcemanagement for distributed computing(Thain et al 2005). The Condor controllerwill submit the run of each individualsimulation to the Condor cluster. The currentCondor cluster comprises 48 compute nodes,which can run in parallel 48 simulations. Itwill monitor the completion of the individualruns and flag to the ART controller forfurther processing.

The ART output module will providefeedback on the whole process, updatingthe user on the selected parameters and therun results.

Finally, the ART controller is the heart of theframework providing coordination of thewhole process.

Detailed explanations of the various modules,including the choice of EA, are given in Chooet al (2007).

MAP AWARE NON-UNIFORM AUTOMATA

Map Aware Non-uniform Automata (MANA)is one of the models that have beenincorporated into the ART framework. It is anAgent-Based Simulation developed by theOperations Analysis group at the DefenceTechnology Agency in New Zealand (Lauren2002). MANA has been used in a number ofstudies involving land combat, civil violencemanagement, and maritime surveillance.

The strengths of MANA include the user-friendliness of the interface, the relative easein creating a scenario from scratch and fastexecution from being run on pre-compiledexecutables. Additionally, MANA supports theexperimentation with intangibles which addcomplexities to the models. These intangiblesinclude behaviour like proxies for aggression,leadership and determination. The tool as awhole allows the analyst to investigate warfareas a complex adaptive system and to observeemergent behaviour (McIntosh et al 2006).

MARITIME SECURITY

With shipping at the heart of the globaleconomy, maritime security is required toensure freedom of the seas and to facilitatefreedom of navigation and commerce. Twokey aspects of maritime security are theprotection of a Key Installation (KIN) and theanchorage against threats from terrorists andcriminals.

The effectiveness of ART was explored in twomaritime scenarios: the first was on therobustness of plans for the protection of KINs

50

against the intrusion of fast boats (Sim et al2007); and the second was related to anchorageprotection (Wong et al 2007).

PROTECTION OF KEY INSTALLATIONS

Scenario

In this scenario, the Blue force was conductingcoastal patrols to guard against threats on theKINs, which were Coastal Surveillance Radars(CSR) equipped with minimum level of self-protection. Red force would attempt topenetrate the Blue defence and inflict damageusing various approaches. Any damage to thecoastline could be seen as a severe psychologicalblow to the Blue defence force. It was assumedthat the Area of Operation (AO) was far awayfrom the main shipping traffic, and henceneutral shipping was not considered. Similarly,the effects of weather and sea state were notconsidered. The scenario (see Figure 2) wasmodelled in MANA.

0.5nm and maintaining this distance for oneminute. The dynamics of the close watercombat was not modelled. In addition, thePVs would be activated to investigatedetections made by the CSRs so as to achievetarget identification and neutralisation. TheBlue force was assumed to have perfectcommunication. A summary of the key inputsused for the Blue KINs and PVs is given inTable 1 and Table 2 respectively.

Red Force

Five Red boats were modelled as small fishingboats with a maximum speed of 25 knots andloaded with explosives. These boats had a shortvisual detection and identification range of1nm. The five Red small boats would actindependently without communicatingwith one another. By limiting the Red’sdetection/identification and communicationabilities, small autonomous low technologyunits were being modelled. These are the kindsof threats maritime security units are likely toface during peacetime. Table 3 summarises thekey inputs used for the small boats.

Blue Force

The Blue force consisted of three KINs andthree Patrol Vessels (PV). The operationalcharacteristics of the KINs and PVs are distilledfor modelling in MANA. Each KIN wasprotected by General Purpose Machine Guns.Each PV conducted normal patrol at 15 knotsand gave chase at a maximum speed of 25knots. The PVs were assumed to be capable ofneutralising the Red boats by closing in within

Figure 2. Scenario for protection of key installations

Table 1. Key inputs for Blue KIN

Table 2. Key inputs for Blue PV

Table 3. Key inputs for Red boats

CSR Detection Range (nm) 5

Weapon Range (km) 2

Weapon Single Shot 0 .1Probability of Hit

PV Speed [Patrol] (knots) 15

PV Speed [Chase] (knots) 25

PV Detection Range (nm) 3

PV Identification (ID) Range (nm) 1

Maximum Speed (knots) 25

Detection/ID Range (nm) 1

51

Measures Of Effectiveness

There were two Measures Of Effectiveness(MOE):

Mean Red mission success – defined as thenumber of successful Red attacks on theKINs/coastline. Red mission was consideredsuccessful when at least one boat managedto penetrate the Blue defence.

Mean Red attrition – defined as the averagenumber of Red boats that were destroyed.

Manual Red Teaming

Manual Red Teaming was adopted first to workon the Red attack plan against a fixed Bluepatrol plan through MRT. Two plans weredeveloped based on the dual objectives ofmaximising the mean Red mission success andminimising the mean Red attrition.

The first plan was based on ‘Flanking’ wherethe Red small boats penetrated two flanks andin doing so, stretched the Blue resources i.e.the Blue PVs (see Figure 3). Using this tactic,Red force was able to achieve a mean missionsuccess of 100% and a mean attrition of 0.85.

The second plan was based on ‘Saturation’. Inthis tactic, three Red small boats penetratedthe centre and tried to saturate the Blue PVs,leaving some gaps for the remaining two smallboats to sneak through the sides (see Figure4). The mean mission success and mean attritionfor Red were 100% and 3.05 respectively.

Automated Red Teaming

The ART process was used next. Besides theevolution of the Red penetration plan, otheraspects of Red behaviour such as aggression,cohesiveness and determination were assessedto see if the plan could be improved.Aggression towards Blue PV, unit cohesivenessand determination in moving towards theobjective (KIN/coast) were chosen to representthe behaviour of the Red force.

The intangible parameters were presented andeffected in MANA through a range of values.A negative value for aggression meant thatthe Red boats feared the Blue PV; a positivevalue would imply the opposite; and zerowould mean indifference. Similarly, a negativevalue for cohesiveness would imply the Redboats tended to spread out; positive meantpreferring to cluster; and zero meant beingindifferent to each other’s presence. Fordetermination, a positive value implied apre-disposition towards the final objective;zero was indifference; and negative meantavoidance in reaching the final objective.

It was interesting to note that the ARTframework produced a decoy tactic that wassurprising, as shown in Figure 5. In this tactic,one of the Red small boats (the one in thecentre acting as a decoy) was deployed to lurethe Blue PV on the left towards the right sideto create an opening for the other two smallboats to charge towards their objectives. Thesweeping movement of the decoy causedenough distraction to result in the left side ofthe map being exposed for the two small boats

Figure 3. Flanking tactics developed through MRT

Figure 4. Saturation tactics developed through MRT

52

to succeed in their mission. The mean Redmission success and the mean Red attritionachieved were 100% and 1.89 respectively.Thus, the effectiveness of the ART-generatedplan was somewhere between the tactics of‘Saturation’ and ’Flanking’.

It was also observed that in general, the Redsmall boats were very focused on chargingtowards the KINs (highly determined) whiletrying to avoid the Blue PVs (negative valuesfor Red aggression). However, the Red smallboats were more cohesive in the decoy tactics.The findings of the ART runs, as compared tothe MRT effort, are summarised in Table 4.

ANCHORAGEPROTECTION

Scenario

Another scenario was developed to continuethe evaluation of the ART framework, this timein anchorage protection. In this scenario, theBlue force conducted patrols to guard againstthreats on an anchorage, with 10 commercialships anchored in the protected area. Red forcewould attempt to penetrate the Blue defenceand inflict damage on the anchored vessels,using various approaches. Any damage to thecommercial shipping would deal a severepsychological blow to the Blue defence force.As in the previous scenario, the AO wasassumed to be away from the main shippingtraffic. The anchorage covered an area of 30nmby 10nm. The AO was designed to be 100nmby 50nm so as to allow the Red force greaterdepth in their movement. The scenario wasmodelled in MANA as shown in Figure 6.

Figure 5. Decoy tactics developed through ART

Figure 6. Scenario for anchorage protection

Table 4. Summary of results for theprotection of the KINs

Flanking Saturation DecoyTactics Tactics Tactics(MRT) (MRT) (ART)

Red -60 -60 -83Aggression

Red -100 -100 8Cohesiveness

Red 60 60 53Determination

Red 100% 100% 100%Mission Success

Red 0.85 3.05 1.89Attrition

53

Blue Force

The Blue force consisted of three PVs, eachconducting normal patrol at eight knots andpursuing threats at a maximum speed of 16knots. The PVs were assumed to be capable ofneutralising the Red boats by closing in within2nm. Once again, the dynamics of the closewater combat was not considered, and the PVswere assumed to have perfect communication.Table 5 provides the key inputs for the PVs.

Red Force

There were five small boats with a maximumspeed of 16 knots and loaded with explosives.Each boat had a detection and identificationrange of 2nm, and acted independently. Table6 shows the key inputs for the small boats.

Measures Of Effectiveness

Three MOEs were identified:

Mean Red mission success – defined as thenumber of successful Red attacks on NeutralCommercial Shipping. Red mission wasconsidered successful when at least one boatmanaged to penetrate the Blue defence.

Mean Red attrition – defined as the averagenumber of Red boats that were destroyed.

Mean Neutral shipping destroyed

Manual Red Teaming

Similarly, MRT was adopted first to developthe Red attack plan against a fixed Blue patrolplan. The restriction was that three Red smallboats had to start from the northern edge ofthe AO, and the other two small boats fromthe southern edge. The starting point was alsolimited to a 100nm by 10nm area off thestarting edge so as to avoid placing the startpoints too close to the anchorage.

In the MRT exercise, the numerical advantageof the Red force was fully utilised and asimultaneous attack on the anchorage area tosaturate the Blue PVs was launched (see Figure7). This was developed based on maximisingthe mean Red mission success, minimising meanRed attrition and maximising mean Neutralshipping destroyed.

The results obtained were 100% for mean Redmission success, 1.96 for mean Red attritionand 3.05 for Neutral shipping destroyed.

Automated Red Teaming

ART was applied next. The same intangibleparameters were chosen to be evolved:aggression towards Blue PV, unit cohesiveness,and determination in moving towards theobjective (anchorage) representing thebehaviour of the Red force.

Similar to the plan developed by MRT, ARTgenerated a simultaneous red attack tactic

Table 5. Key inputs for Blue PV

Table 6. Key inputs for Red Boats

Figure 7. Saturation tactics developed through MRT

PV Speed [Patrol] (knots) 8

PV Speed [Chase] (knots) 16

PV Detection Range (nm) 6

PV Identification (ID) Range (nm) 2

Maximum Speed (knots) 16

Detection/ID Range (nm) 2

54

towards the centre of the anchorage area withre-attack flexibilities. This would cater to caseswhere the anchored vessels were dispersednearer to the anchorage edges. This is arefinement over the MRT tactic where theobjective was to reach the anchorage. TheART results pushed it further with the Redboats traversing the anchorage looking fordispersed vessels. The ART has given the insightthat it is not enough to just stop the Red boatsfrom reaching the anchorage but it is alsoimportant to prevent the boats which hadpenetrated from manoeuvring within theanchorage. Insights like these can help theplanners to refine the security tactics. TheART tactic for the Red team is shown inFigure 8.

The ART-generated tactic was able to performbetter than the MRT plan, achieving a meanRed mission success of 100% with a lower meanRed attrition of 0.48, and a higher meanNeutral shipping destroyed of 4.52. It was alsoobserved that the Red small boats had only amild fear of the Blue PVs (-4 as compared to-60 for the MRT case), and they were morecohesive (-16 as compared to -100). The findingsof the ART runs, as compared to theMRT effort, are summarised in Table 7.

CONCLUSION

The evaluation of the ART framework vis-à-visMRT showed that ART could complementmanual efforts. In particular, ART could uncoverunique and surprising solutions (as in the‘Decoy’ tactic for the KIN scenario) or providerefinements to the manual plans (as in the‘Saturation’ tactic in the anchorage protectionscenario), which might otherwise not surfaceor be considered during the MRT process.

However, it is important to stress that theobjective of ART is not to replace MRT. Rather,it is to complement the MRT effort. There isstill a need to involve human analysts to makesense of the ART results, at least for the currentstate of ART.

FUTURE WORK

The ART framework is still in its early stage ofdevelopment. Besides the plan to incorporatenew EAs and models into the framework,further tests are required to benchmark itsperformance and assess its robustness.

Figure 8. Saturation tactics developed through ART

Table 7. Summary of results foranchorage protection

Saturation DecoyTactics Tactics(MRT) (ART)

Red Aggression -60 -4

Red Cohesiveness -100 -16

Red Determination 60 45

Red Mission Success 100% 100%

Red Attrition 1.96 0.48

Neutral Attrition 3.05 4.52

55

REFERENCES

There are also some potential spin-offs fromthis project. It is natural to extend ART to fulfillthe concept of Blue Teaming vis-à-vis RedTeaming i.e. Automated Co-Evolution (ACE)where both sides evolve and adapt againstchanging tactics. The initial research anddevelopment work on ACE has beencompleted. Another potential application is inthe calibration of models e.g. determining thevalues to be assigned to the parameters toachieve certain desired outcomes.

Choo, C. S., C. L. Chua, and V. Tay. (2007).Automated Red Teaming: A ProposedFramework for Military Application. GECCO2007, Vol 2, pp 1936-1942.

Johnson, S. K., M. J. McDonald, Upton, S. C.(2004). Breaking Blue: Automated Red TeamingUsing Evolvable Simulations. GECCO 2004.

Lauren, M. K. and R. T. Stephen. (2002). MapAware Non-uniform Automata (MANA) a NewZealand Approach to Scenario Modelling.Journal of Battlefield Technology, Vol 5,No. 1.

Livny, M., Tannenbaum, T. and Thain, D. (2005).Distributed Computing in Practice: The CondorExperience, Concurrency and Computation:Practice and Experience, Vol 17, No. 2-4, 2005,pp 323-356.

M. J. McDonald and Upton, S.C. (2003).Automated Red Teaming Using EvolutionaryAlgorithms. WG31 – Computing Advances inMilitary OR.

McIntosh, G., Galligan, D. P., Anderson, M. A.,and M. K. Lauren. (2006). Recent Developmentsin MANA Agent-based Model. Scythe, Issue 1,pp 38-39.

Sim, W. C., Choo, C. S., Ng, E. C., Martinez-Tiburcio, F., Toledo-Ramirez, E. R. and K. Lin.(2007). Applying Automated Red Teaming ina Maritime Scenario. Scythe, Issue 2, pp 26-29.

Wong, A., Sim, W. C., Chua, C. L., Lim, Y. L.,Chin, S. C., Teo, C., Lampe, T., Hingston, P. andB. Abbott. (2007). Applying Automated RedTeaming in a Maritime Security Scenario.Scythe, Issue 3, pp 3-5.

56

Victor Tay Su-Han is a Principal Engineer and currently holds the appointmentas Acting Assistant Director (Simulation) in the Networked Systems ProgrammeCentre. In this appointment, Victor is responsible for the acquisitionmanagement of simulation programmes for the Singapore Armed Forces(SAF). He is also an Adjunct Senior Fellow in the Temasek Defence SystemsInstitute in the National University of Singapore (NUS). Victor recently wasa member of the National Research Foundation Interactive Digital Media(NRF IDM) Operations Working Group, where he contributed to its initialfledgling efforts in national IDM research and development. He received hisMaster of Science in Interactive Simulation from the University of CentralFlorida in 1999, under the then Defence Technology Training Award.

BIOGRAPHY

Choo Chwee Seng is a Principal Member of Technical Staff in OperationsResearch Laboratory, DSO National Laboratories (DSO). He leads and managesprojects that involve Combat Modelling, Simulation and Analysis, ExperimentalDesigns, Evolutionary Computation and Data Farming. He received his Bachelordegree in Physics from NUS in 1992, and his Master of Science in OperationsResearch from Stanford University in 1997.

Chua Ching Lian is currently a Senior Member of Technical Staff in OperationsResearch Laboratory, DSO. He works on projects that involve Combat Modelling,Simulation and Analysis, Experimental Designs, Evolutionary Computationand Data Farming. His research interests lie in operations research, simulationsand algorithms. He received his Bachelor degree in Mechanical and ProductionEngineering with specialisation in Mechatronics from Nanyang TechnologicalUniversity in 2002, and his Master of Science in High Performance Computationfor Engineered Systems from Singapore-Massachusetts Institute of TechnologyAlliance, NUS in 2003.

57

Spare Parts Management

for Large-scale Fleet Scenarios

ABSTRACT

In managing spare parts for large-scale fleet scenarios, one

needs to optimally allocate spares across multi-echelons of

maintenance agencies. This paper discusses an in-house

developed simulation model (PIPER) for the Army to solve

problems such as the evaluation of maintenance support

concept, the impact of combat damage during wartime and

workshop loading. PIPER is validated successfully against

commercially available tools with good agreement between

the models. It uses the combined technique of analytical

marginal analysis while heuristics are employed for optimising

spares and maintenance resources. The article also discusses a

demand forecasting model that addresses the different

equipment reliability when they are operated in different

environments. This model is based on a multi-population

mixture of Weibull failure distributions.

See Chuen Teck

60

INTRODUCTION

It is well known that spare parts managementfor large-scale fleet scenarios is a complexproblem. In particular, one needs to optimallyallocate spare parts across multi-echelons ofmaintenance agencies. The Pipeline Simulator(PIPER) is a Monte Carlo simulation modeldeveloped by DSTA to manage spare parts forthe Army. The model solves problems such asthe evaluation of maintenance supportconcept, the impact of combat damage andworkshop loading. This model provides fullaccess to the source code for customisationand integration with other models orManagement Information Systems. It is alsodesigned to be scalable whereby models andnew functionalities are created via the additionof ‘building blocks’. It analyses multiple combatunits, quantifies the effect of sharing sparesand men, handles war scenario (time varyingutilisation rate, combat damage and attrition,base open and close) and explicitly modelsrepair manpower, heavy transporter andperiodic re-supply. A combinatorial techniqueof analytical marginal analysis and heuristicsis employed for optimising spares andmaintenance resources in PIPER.

The first step to good management of spare

parts is to forecast demands realistically from

real life data. However, in many operational

fleets, differences in field operating conditions,

usage intensity and product inception time

often result in varying age and conditions of

the equipment. While the case of homogenous

equipment failure characteristic has been

popular, it is inappropriate in many military

systems where the deployment of the fleet is

dispersed over a multiplicity of operating

environments. To improve the accuracy of PIPER

in optimising spares and maintenance

resources, a demand forecasting system for

spare parts management in large-scale fleet

scenarios with substantial variability in

equipment reliability is being developed. Our

model is based on a multi-population mixture

of Weibull failure distributions.

MODEL

The PIPER model is built using Extend

(developed by Imagine That Inc.), a simulation

tool widely used by academics and the

simulation industry. The PIPER model shown

in Figure 1 consists of four echelons of repair

Figure 1. Four levels ofrepair echelon

Spare Parts Management

for Large-scale Fleet Scenarios

61

agencies. Systems can be deployed at any ofthe four repair echelons. System repair is carriedout at the second, third and fourth repairechelons. All four echelons hold a suite ofmaintenance resources i.e. test equipment andtechnicians to service the repair jobs. Themaintenance resources follow a user-definedoperation schedule (the base open and closetiming). The quantity of maintenance resourcesis allowed to change over time.

The transport time and milk-run frequencyamong the various repair agencies could bedefined to take special values and to override

the default parameters. This may be used torepresent certain Line Replacement Units(LRUs) transported by special mechanisms e.g.helicopter lift, pseudo stores or repair echelons.The milk-run frequency may be variable overthe timeline and a frequency of zero milk-runcan be used to represent a temporarystoppage of supply i.e. enemy action or truckgetting ‘lost’.

The model is synthesised from building blockspresent in the PIPER libraries as shown inFigures 2 and 3.

Figure 2. Unit level

Figure 3. Second repair echelon

62

Validation against other commercial tools suchas SPAR (a tool developed by ClockworkSolutions) shows good agreement betweenthe two models. Shown in Figure 4 is the resultof validation from two Army case studies. Itshould be highlighted that the validationfocused on the simulation aspect of the model.Validation on the optimisation algorithm wasnot addressed in this portion.

OPTIMISATION HEURISTICS

The simulation model is not known to be thebest technique for optimisation problems. Dueto the error in running a finite number of

replications and computational speeds, answers

obtained via simulation models cannot be

shown to be mathematically optimal. PIPER

hence utilises the well-known METRIC model

(Sherbrooke, 1992) for optimising spare

demands that arise from reliability failures.

METRIC is an analytical model widely used in

the industry for spares optimisation and have

shown to provide near-optimal answers. The

algorithm consists of two main steps. In the

first step, the best location to stock the item is

derived via an enumeration of the solution set.

The best location is the location that maximises

Operational Availability (Ao). The non-

dominated set of Ao versus stock quantity (also

known as the efficient frontier) is generated.

Figure 4. Validation results from two Army case studies

Figure 5. Optimisation algorithm

63

This is carried out for all the LRU types.

Combining the various LRU types via marginalanalysis (or greedy heuristics) generates thecost-effectiveness curve. A detailed write-upon the algorithm can be found in Sherbrooke,1992.

The method of marginal analysis optimisesspares well but is inadequate for maintenanceresources optimisation. This is because a suiteof technicians is needed to service a failureand the technicians are shared across LRUs orsystem types performing LRU removal,replacement as well as repair. The maintenanceresources optimisation problem is henceinseparable. It has a non-linear objectivefunction and integer decision variables whichmake the problem hard. The following is theheuristic adopted to derive the minimumquantity of maintenance resources:

• Run simulation with no technicians. Whenmore than one technician is available to servicea job, the job will be assigned to the technicianwho has worked the most. This serves thepurpose of minimising the quantity oftechnicians. Create technicians only if there is

a need, and track the usage of each techniciancreated. The number of technicians created isthe quantity that would give zero bottlenecksto the system which is tantamount to ‘Infinitespares’ and may be huge.

• Backtrack the solution iteratively i.e. runsimulation with x% of the total man-hoursneeded. For example, in iteration 1, set thetechnician quantity to obtain 30% of the totalman-hours that is needed to derive thebottleneck to zero. The technician quantitycan be increased from 30% to 40% in iteration2 and so on.

• Select an operation point on theeffectiveness curve. Improve the solution byremoving technicians with low usage rate i.e.idle technicians. Re-run to confirm that the Aois still above the required level.

Figure 6. Graph showing a sample result of spares optimisation

64

SPARE PARTS DEMAND FORECASTING

A large-scale fleet management organisationsuch as the military often operates equipmentat many different sites. However, due toenvironmental differences, equipmentdeployed at different operating sites overprolonged periods of time have differentobserved failure rates. In addition, as mostlarge system houses typically phase in newequipment over many stages, the entire fleetof equipment tends to exhibit different levelsof reliability. While it has been a commonpractice to use a single distribution similar tothe study done by Peltz, Colabella, Williams &

Boren (2004), this method is clearly not

sufficient here. Murthy, Xie & Jiang (2004)

described a comprehensive suite of Weibull

models. Out of the seven models mentioned

therein, Type III (a) model which is a multi-

population mixture of related Weibull models

looks more appealing because of its proximity

to real life scenarios. The authors mentioned

that other classes of Weibull models such as

Type III (d) Sectional Weibull are a theoretical

postulation, and they are unable to offer any

physical explanation of the underlying process.

Although many methods for multi-population

Weibull have been proposed, there is little

research done at making a direct comparison

between the methods.

Figure 7. Results from an Army case study on maintenance resources optimisation

Eliminatebad

solution

65

The reliability function for a mixture of twoWeibull distributions is described below.

The idea of a mixture of two Weibulldistributions was described by Jiang &Kececioglu (1992) and Jiang & Murthy (1995).In this paper, we are following Jiang &Kececioglu’s method of parameter estimation,which is summarised as follows:

rates in the population, cannot be less than

one. When is more than six, it can then bemodelled using a mixture of three-parameter

Weibull distributions. In addition, all the values within the same mixture of Weibulldistributions are constrained to be the same.This is because we assumed the nature of the

failure mechanism to be similar. If the valuesare different, there will be a possibility thatthe time-to-failure percentile of the strongerpopulation is smaller than the weakerpopulation, which is contrary to intuition. The

assumption of the same values is a commonassumption made in accelerated life tests whichmost military equipment undergo. Hence, it isnot unreasonable to as sume thatenvironmental stress factors affect values

and not .

The above parameters estimation algorithmcan be generalised for a mixture of n Weibulldistributions, with the first four steps beingsimilar. The rest of the steps need to bemodified as follows:

• Locate the points,T1, T2,..., Tn-1, where thesecond derivative of the fitted curve is zero.Obtain p1 from yT1 = 1n(–1n(p1)), p2 fromyT2= 1n(-1n(p2)),... , and pn-1 from yTn-1

= 1n(–1n(pn-1)).

• Determine by taking the average of thetwo values of slopes of the tangent lineswhich are drawn at each end of the curve.

• Determine i by determining the y-axis atthe values of expressions (1.1) and (1.2), for

1 and i respectively, horizontally;intersecting the curve and dropping down,the i can be read from the exponential ofthe x-axis values, where 1 i n.

(1)

1. Rearrange the failure time in ascendingorder. Let ti ≥ 0, 1≤ i ≤ n.

2. Compute xi and yi, 1≤ i ≤ n, where

3. Plot yi vs. xi.

4. Fit a smooth curve to approximate theplotted data.

5. Locate the point T, where the secondderivative of the fitted curve is zero.

6. Obtain p from yT = 1n(-1n p).

7. Determine 1 and 2 from the gradient ofthe tangent lines of the left and rightasymptote of the curve.

8. Determine 1 and 2 by determining they-axis at 0.632p and p + 0.632(1-p)horizontally; intersecting the curve anddropping down, the 1 and 2 can be readfrom the exponential of the x-axis values.

For most practical purposes, we note that the

value of ranged from one to five. As spareparts have the property of increasing failure

and

empirical distribution.where

66

The reliability of a mixture of n Weibull distributions is as follows:

Divide throughout,by

Since , we have

When

When

(2)

When

and

where

where

if

(3)

67

Although it is possible to estimate theparameters of a mixture of n Weibulldistributions, we focus our discussion on thecase of two populations for simplicity. Weaddress the case of complete data i.e. nocensoring.

The crux of the estimation problem lies in theapproximation of the curve (step four) becausethe point of inflexion is used to estimate thepopulation mixture, pi. Hence, the curve fittingis an important step. Our main idea is tominimise the absolute deviations of theresiduals, which will result in the linearprogramme below.

Problem CF:

subject to

order of failure timein ascending order

is the coefficient of the polynomial

= ith data point

We penalised the positive deviations (slack)and negative deviations (surplus) equally andconstrained the estimated curve, g(x), to benon-decreasing for every data point. Equation(4) is an equivalent formulation of Problem CF.

(4)

subject to

is an empirical density function and isnon-decreasing, which means yi is non-decreasing and therefore, g(x) needs to benon-decreasing. This is equivalent to fitting a50% percentile line (the median line) inquantile regression by setting t in Equation(2.1) to be 0.5. The L1 line enjoys two mainadvantages: Firstly, L1 regression contains somerobust properties and is superior to L2 (leastsquare line) in y-deviations, as mentioned inRousseeuw & Leroy (1987). Secondly, L1

regression is a linear programme whereas L2

regression is a quadratic programme, whichmakes L1 more attractive.

For a two-population distribution, we fitted afourth degree polynomial curve to the plottedpoints. Though any high order polynomialgenerally suffices, the main advantage of afourth degree polynomial is that the secondderivative is quadratic which makes the solutionof the inflexion point easy.

CONCLUSION

Due to the increasing need to maintain highcombat capabilities while keeping the budgetlow, it is essential to perform realisticforecasting on the demands of spare partsbefore using PIPER to improve its accuracy inspares and resources allocation. In ourdevelopmental work, we did a comprehensivecomparison of the popular methods in theliterature. Our test cases have shown that amixture of multi-population Weibulldistributions can be used to represent thedemands of spare parts for large-scale fleetscenarios with substantial variability inequipment reliability.

Ui>= 0, Vi>= 0

REFERENCES

68

A. Agostini. (2001). Critical Survey of Multi-Echelon Repairable Inventory. Msc Thesis, 2001.

A. Dubi. (2000). Monte Carlo Applications inSystems Engineering. John Wiley & Sons, NewYork, 2000.

Jiang, R. and Murthy, D. N. P. (1995). ModellingFailure-Data by Mixture of 2 WeibullDistributions: A Graphical Approach, IEEETransactions on Reliability, Vol. 44, No. 3, pp477-488.

Jiang, S. and Kececioglu, D. (1992). GraphicalRepresentation of Two Mixed-WeibullDistributions, IEEE Transactions on Reliability,Vol. 41, No. 2, pp 241-247.

Kao, J. H. K. (1959). A Graphical Estimation ofMixed Weibull Parameters in Life-Testing ofElectron Tubes, Technometrics, Vol. 1, No. 4,pp 389-407.

Murthy, D. N. P., Xie, M. and Jiang, R. (2004).Weibull Models, U.S.A.: Wiley.

Peltz, E., Colabella, L., Williams, B. and Boren,P. M. (2004). The Effects of Equipment Age onMission-Critical Failure Rates: A Study of M1Tanks. Santa Monica: RAND.

Rousseeuw, P. J. and Leroy, A. M. (1987). RobustRegression & Outlier Detection, U.S.A.: Wiley.

Sherbrooke C.C. (1992). Optimal InventoryModelling of System: Multi-Echelon Techniques.John Wiley & Sons, New York, 1992.

This paper was first published in the Institutionof Engineers, Singapore (IES) CommemorativeBooklet produced for the IES SystemsEngineering Seminar on 19 March 2008 andhas been adapted for publication in DSTAHorizons.

See Chuen Teck is a Principal Analyst (Systems Engineering) and is in chargeof developing Decision Support Solutioning. He has been practising OperationsResearch for over 10 years. His research interest includes probability andstatistics and he has published papers in the European Journal of OperationalResearch, Statistics and Probability Letters as well as the Journal of Inequalitiesin Pure and Applied Mathematics. Chuen Teck obtained his Master of Sciencein Industrial and Systems Engineering from the National University of Singaporeand was awarded the Operations Research Society of Singapore Book Prizein 1998. He is currently pursuing his PhD in Industrial and Systems Engineering.

BIOGRAPHY

69

Continual SystemsDevelopment for

Command, Control and Intelligence Systems

ABSTRACT

The traditional software development model is no longer

adequate to meet the challenges of the fast-evolving needs

of Command, Control and Intelligence (C2I) systems. To enable

the transformation of the Third Generation Singapore Armed

Forces (SAF), a continual systems development approach has

been adopted by DSTA to develop C2I capabilities for the SAF.

The key advantage of this approach lies in its flexibility to

respond rapidly to meet changing needs and address emerging

threats. The complexities of network-centric operations entail

the fielding of systems quickly for operational trials and to

continually discover capability gaps so as to evolve systems

capabilities over its life cycle. This paper addresses the

framework adopted by DSTA to develop the C2I system

of systems.

Dr Yeoh Lean Weng

Teo Tiat Leng

Lim Horng Leong

72

INTRODUCTION

The proliferation of network-centric operationsconcept has introduced layers of complexityin designing and developing Command, Controland Intelligence (C2I) systems to supportmilitary operations. After 9/11, thecharacteristics of asymmetrical threats haveintroduced new chal lenges in ourunderstanding of the capabilities needed. Weneed to go through a process of learning andexploration to discover the new systemsrequirements and to translate these into thearchitecture of C2I systems. Given the evolvingrequirements and the emerging capabilitydemands from known and hidden threats,there is a pressing need for technologists toshorten development cycles and put the systemsinto the hands of commanders and warfightersas quickly as possible. The results gatheredthrough field trials and experiments enablethe stakeholders to understand the needsbetter, explore new operational concepts andidentify capability gaps. DSTA’s ContinualSystems Development approach leads to shorterand more frequent systems releases, resultingin better capabilities for the Singapore ArmedForces (SAF).

The traditional waterfall model is inadequateas many unknown requirements cannot bespecified at the early developmental stages.During the late 1990s, Boehm’s spiral model(Boehm, 1988) inspired numerous C2I systemsdevelopers to rethink and attempt to adopt itas a risk management process to deal with thecontinually changing requirements. Hightechnical risks were mitigated throughinitiating multiple spirals to understand therisks better and synthesise solutions to addressthe risks. The various spiralling efforts must beorchestrated to generate the synergy to deliversystems capabilities to the commanders toconduct Integrated Warfare.

DEVELOPMENT CYCLES FOR C2I

System Life Cycle. In the 1980s whenmainframes and Unix workstations dominatedinformation technology, the life cycle of C2Isystems was between 10 to 20 years. For theC2I systems that were installed onboard militaryplatforms, the life span could extend beyond30 years. During those years, the C2I systemswere less complex and software upgrades wereinfrequent. A typical C2I system would beintegrated into a suite of sensors and performtracking and data fusion to present thesituation to the commanders.

VADM Cebrowski conceptualised Network-Centric Warfare in 1998 (Cebrowski, 1998) toexplore the effectiveness of networkingcommand and operating nodes to achieve thebenefits of speed of command and self-synchronisation. In 2002, the SAF began itsjourney to transform into the Third GenerationSAF that can respond to the challenges of the21st century. In tandem with this, IntegratedKnowledge-based Command and Control(IKC2) was conceptualised. Leveraging theknowledge-centric paradigm, C2I systemswould be network-enabled and organisedaround knowledge for effective command andcontrol (Lee et al., 2003). The shift from theplatform-centric to a knowledge-centricparadigm generated chal lenges forstakeholders to fully specify the completecapabilities of IKC2. Since the advent of IKC2,C2I systems have evolved through severaliterations over their life cycles, constantlyrenewing themselves to maintain relevanceand meet the challenges of future threats. Ageneration of renewal could happen betweenfive to eight years, during which manyenhancements would be incorporated todeliver new capabilities. With the challengesof impending asymmetrical threats, a shorterdevelopment timeframe has become a necessityfor developers to build and deploy systemsexpeditiously to explore and validate the SAFoperational concepts and identify capabilitygaps. Through the knowledge gained from

Continual SystemsDevelopment for Command, Control and Intelligence Systems

73

operational trials, developers can proceed tointegrate or construct new capabilities to plugthe gaps. The Continual Systems Developmentapproach has resulted in a blurring ofdelineation between a prototype and anoperational system. This developmentalapproach has become the choice for developingand delivering C2I capabilities to the SAF.

Continual Systems Development. Thedevelopment model has to exhibit threecharacteristics to qualify as Continual SystemsDevelopment. Firstly, it has to possess the agilityto handle changing requirements to maintainthe relevance of the systems over time.Secondly, the time taken from conceptualisingthe system to developing the capabilities hasto be short to enable rapid fielding of thesystem. This is not unlike commercialcompetitive pressures of time-to-market.Thirdly, it needs to be evolving to embracetechnological opportunities. When therequirements are well understood, systemsfunctionalities and capabilities can beincrementally integrated into the baseline C2Isystems through a properly staged schedulefor releases. Every incremental release adds tothe widening range of operational capabilitiesof the system. Developers must have the agilityto field the system in a shorter cycle and torespond effectively to evolving needs. Asrequirements are packaged into differentreleases, developers can work with stakeholdersto prioritise the schedule to introduce thesenew functional capabilities.

When C2I systems exhibit medium to high risks,a flexible model is needed to manage theserisks. Risk areas could be due to emergent andevolving operational concepts, leading andbleeding-edge technologies and the resultantarchitectural risks. Boehm’s spiral model is arisk-driven process that guides multi-stakeholders to engineer software-intensivesystems (Boehm, 2001). The cyclic approachleads developers to incrementally implementthe system while decreasing the risks. Whilethe well-known spiral figure showing the radialand angular growth at each progression seems

to suggest that there is a single thread ofdevelopment, Boehm has highlighted thatparallel spiral cycles could be spun off for eachsoftware component. The parallel spirals createcomplexity and project managers need tomanage these parallel spirals in congruence.

To better manage the risks holistically,engineering master plans are developed to layout the approach, risk mitigation strategy andestimate the number and frequency of thespirals. Usually, three to five spirals are neededfor the system to stabilise. In a single spiral,there will be several mini spirals executed inparallel or in series. Each mini spiral can lastbetween three to six months while some canbe as short as two weeks. Unlike the agiledevelopment model that follows a stricttimebox control, the mini spirals create theflexibility, time and space for the stakeholdersto manage the risks effectively and deliverseveral releases for rapid operational trials.Developers are not burdened with full-scaledocumentation but would produce onlysufficient artefacts to capture the designconsiderations, risks, and decisions made whichcan be referenced to guide future spirals.

A single spiral is typically planned to becompleted within a year so that the C2I systemcan be fielded for at least one majoroperational exercise. Through the exercises,capability gaps and system deficiencies wouldbe identified and addressed in future spirals.While the spiral model may adequately managethe risks of developing a single C2I system,System-of-Systems (SoS) architecture risksinherently straddle various systems at theenterprise level and pose a different set ofchallenges. While developing C2I SoS, multiplespirals for each individual system will beconcurrently executed. The collective effort toconstruct these component systems andintegrate them into the SoS to achieve theintended result has to be carefully plannedand orchestrated. Otherwise, the SoS couldbecome dislocated and would fail todemonstrate its intended Integrated Warfarecapabilities.

74

Managing Concurrent Spirals. Beforeembarking on a massive development of theC2I SoS, deliberate and comprehensive planningis carried out. An enterprise architecture hasto be constructed so that it can provide thestrategic framework to deal with theintegration issues pertinent to the developmentof the C2I SoS (Yeoh et al., 2007). To mitigatethe architectural risks, an experimentalapproach is taken to construct a baseline C2IEnterprise Architecture for experimentation.An experimental System A is then developedto verify the completeness and correctness ofthe C2I Enterprise Architecture. The C2IEnterprise Architecture continues to evolveinto version 1 for the basis of constructing theremaining individual systems. Version 1 of thearchitecture is verified through developingSystem B for experimentation. Figure 1 showsthe concurrent spirals development for theC2I SoS.

The concurrent spirals have to be managed ina concerted effort to mitigate architectural,technical, process and scheduling risks. Multiplesynchronisation points have to be establishedduring the planning phase so that the systemscan converge throughout the developmenta n d i m p l e m e n t a t i o n p h a s e s . T h eexperimentation-to-operationalisationapproach separates the risks into manageable

pieces during the concurrent spirals. Afterverifying the C2I Enterprise Architecture, theindividual systems are then staged fordeve lopment and managed underconfiguration control. As shown in Figure 1,each version is an incremental release ofcapabilities into the C2I SoS. The incrementalrelease also reduces integration risks andenables faster time-to-fielding for operationaltrials. The C2I Enterprise Architecture itselfwill also evolve with version changes.

Operations and Support (O&S). Prior to thecontinual development, the development teamwill hand over the system to a dedicatedsupport team to provide systems maintenanceduring the O&S phase. The transitioning toO&S entails training of the support team bythe development team. The support teamwould need to re-learn the design from thedevelopment team and would invariably beless competent than the development teamin fixing the bugs. Therefore, such anarrangement might affect the system andoperational readiness.

In the Continual Systems Developmentframework, the same development teamcontinues to support the system throughoutthe exercises and daily operations. The teamwill fix software bugs and develop additional

Figure 1. Concurrent Spirals Development for C2I System-of-Systems

75

functions to support the operations. There is

no learning curve and we have benefited from

the leanness that can be achieved as the

knowledge and experience is retained within

the team. The veteran developers are around

to groom the juniors and pass on the

knowledge through on-the-job coaching and

supervision. The team continues to support

the system and then dovetail to the next spiral

development or major upgrade.

ENABLING CONTINUALDEVELOPMENT FOR C2I SYSTEMS

To ensure proper governance, DSTA has

established the Enterprise Architecture

Framework to guide the innovation and

experimentation of new operational concepts

(Yeoh et al., 2007).

Enterprise Technical Architecture (ETA). The

early generations of C2I systems for the SAF

were developed with a low degree of

connectivity among the three Services, namely

the Army, Air Force and Navy. With the

advancement in information and networking

technologies, C2I systems were rapidly

networked across the three Services through

sharing and integrating common services.

Over time, an unwieldy mesh would be created

if there were no clear framework for propergovernance. As such, DSTA established theEnterprise Architecture Framework to ensurethat the C2I systems are developed incompliance with this framework so that thesystems are interoperable by design. Theframework offers better scalability through alayered and integrated design to create a well-integrated and interoperable developmentenvironment.

The ETA is one of the components in theEnterprise Architecture Framework (Yeoh etal., 2007), and is adapted from the Service-Wide Technical Architecture established by theInfocomm Development Authority ofSingapore. It aims to establish the principles,standards and development guidelines in thedesign, development and acquisition ofInformation Technology (IT) systems that rangefrom ubiquitous corporate IT systems tospecialised systems such as C2I systems. Thereare eight principles to guide the developers toconstruct the enterprise architecture. Table 1shows the eight architecture developmentprinciples.

The ETA is further organised into a nine-domainarchitecture to provide the guidance andstandards for the developer to construct anenterprise system. Figure 2 shows the nine-domain architecture.

Figure 2. Nine domains of the Enterprise Technical Architecture

76

Table 1. Eight principles for developing architecture

Architecture Development Principles

S/N Principle Definition

1. Information is a DSTA should be linked as a single virtual networkcritical asset. It must in order to provide all personnel with on-demand accessbe effectively to authoritative, relevant and sufficient informationcollected, managed, to perform their duties effectively.exploited, sharedand protected.

2. Adequate security Security must adequately protect information from unauthorised access and protect systems from attacks,both internal and external.

3. Reduced integration Interoperability and ease of integration (both intra andcomplexity through inter agency) should be achieved through adherence to openstandards standards with wide industry acceptance and implementing

simplified and well-defined interfaces. In the case where noopen standards exist, the organisation should adopt a product standard with wide industry acceptance. In addition, standard user interfaces and access protocols should be usedfor all systems, where standards are available.

4. Reuse through Systems should evolve to employ and share reusable component-based components and infrastructure services across DSTA.model

5. Highly granular Systems should be engineered to be 'highly granular' and 'loosely coupled'. This can be achieved through N-Tierlogical partitioning and implementing event-driven systems and message-based interfaces.

6. Architecture and Architecture design and development should incorporateinfrastructure degrees of robustness, scalability, and adaptiveness torobustness, support continuity, growth, and evolution of the business scalability, respectively. Performance requirements of a system shouldadaptiveness and be considered in totality which may lead to design trade-offsperformance of components within each domain.

7. Cost-effectiveness Minimising total cost of ownership should be a goal ofand operational architecture development. Both initial development costsefficiency and ongoing operational costs like system administration

and maintenance must be considered in totality. Operational efficiency of the architecture should be considered.

8. Minimise Create a small number of consistent configurations for configuration deployment across the enterprise.

support

77

The Application domain describes theframework for the design of applications forinteroperability, maintenance of a high levelof distributed systems integration, reuse ofcomponents and rapid deployment ofapplications to enable a high responsivenessto changing operational requirements.

The Collaboration and Workflow domaindefines the environment for the automationof the distribution of ideas, notices anddocuments throughout workgroups and theentire organisation. The nature of thecollaboration and workflow among the usersand machines could be based on the processesneeded in the workgroup or the conversationaltype of the interactions.

The Internet and Intranet domain definesthe technologies, standards and guidelines forseamless and platform-independentcommunications among the internal andexternal nodes.

The Data Management domain defines themechanics for managing, securing andmaintaining the integrity of every data entity.It also describes the structure of authoritativedatabases and provides the standards to accessdecision support data.

The Distributed Environment Managementdomain defines the hardware and softwarecomponents of the environment that will becontrolled through configuration management.The broad disciplines of the domain also includefault detection and isolation, testing,performance measurement, problem reportingand software upgrades and control.

The Middleware domain defines anintegration environment between workstationsand servers in order to improve the overallusability of the distributed infrastructure. Itcreates a uniform mechanism for applicationintegration independent of network andplatform technologies.

The Platform domain defines the technicalcomputing components of the infrastructurefor the client and server hardware to interactwith the operating systems. It also describesthe storage, backup and high availabilitycomponents that constitute the hardwareinfrastructure.

The Network domain defines thecommunication infrastructure for thedistributed computing environment. It describesthe logical elements such as topology, physicalhardware components and protocols for thenetworking infrastructure.

The Security domain defines thetechnologies, standards and guidelines toensure the availability, integrity andconfidentiality of data. The elements includeidentification, authentication, authorisation,access control, administration and audit.

Common Repository. The common repositoryis an asset that DSTA has created to supportthe Continual Systems Development. Therepository preserves the intellectual capital ofC2I systems business application and technicalcomponent services from which developerscould draw upon to rapidly assemble anddeploy C2I systems for the SAF. As the repositoryapplications and services are thoroughly testedfor operational deployment, the assembledC2I systems would achieve a high degree ofassured quality for operational trials.

To continually evolve and expand therepository, a rigorous process was adopted tobuild common applications and services on topof the existing services. When capability gapsare identified, C2I systems developers willdevelop the new applications and services ontop of the baseline C2I systems. The C2I systemsare then deployed for operational exercises toverify the implementation and validate thatthe gaps are satisfactorily covered. After thesuccessful completion of experiments, the newapplications and services will be enhanced to

78

incorporate the non-functional requirementssuch as exception handling and reliability. Thevalidated services are then integrated into thenext baseline C2I systems and added into thecommon repository for future development.Figure 3 shows the process of continuallybuilding up the common repository.

Using Modelling and Simulation (M&S). M&Shas been exploited to assist the stakeholdersto discover new requirements. As such, C2Idevelopers will team up with M&S analysts tohelp users to define operational scenarios anddevelop simulation models to allow users towalk through their scenarios. Through theprocess of modelling and simulation,stakeholders can develop a betterunderstanding of the operational issues,emerging threats and capability gaps. Thesystem effectiveness could be analysed toestablish the factors affecting the overall systemperformance. Leveraging the newly acquiredknowledge, C2I developers can design differentarchitectures. Developers can further employM&S to evaluate the relative performance ofvarious architectures to select the optimal

architecture to meet the desired performanceof the C2I SoS.

Using an M&S approach to discover newrequirements and develop architectures in earlyspirals is critical to the continuity of developingC2I systems in future spirals. The right systemlevel and functional level service componentsare identified for development in the earlyspirals. These service components can beassembled to form new capabilities andvalidated using M&S. Another benefit ofemploying M&S to study the needs is to levelup all stakeholders’ knowledge on theoperational and technical issues. This helps tobring the operational and technicalcommunities to a common footing to buildbetter systems.

Emulator-based C2 Development (EC2D). TheEC2D is another approach to address thedifficulties of designing first-of-its-kind systems(Yeoh et al., 2007). In the early spirals, emulatorscan be constructed and assembled to provideemulated inputs from other interacting systemcomponents to the C2I systems. These system

Figure 3. Process for maintenance of the Common Repository

79

components can be the missile system or thesurveillance radar system. The C2I systems canbe easily assembled using the componentsfrom the common repository and thenintegrated with the emulators to form anemulator-based C2I system environment tofacilitate the exploration of operationalconcepts and evolving requirements. The EC2Dprovides a low cost and low complexity optionin the early development phase wheninteracting system components are beingspecified. As and when the interacting systemcomponents are available, they will thenreplace the emulators in the follow-on spirals.

Resource Management System (RMS). Humanresources are paramount to the growth andsuccess of the organisation. This is especiallytrue for a technology organisation like DSTA.To maintain leanness, utilisation of limitedresources has to be properly planned andallocated to maximise the desired outcomes.The RMS was created to facilitate the planningand allocation of manpower. The developercompetencies are kept up to date in the RMSand the system is able to recommend the rightmatch between the resource requirements ofthe developers and the project. Matching thedevelopers to the appropriate tasks enables ahigher probability of success in meeting thesystem’s requirements. In addition, individualdeveloper workload is also captured in thesystem to avoid overloading the developer.Maintaining a healthy balanced workloadallows the developers to spend time thinkingcreatively and delivering innovative solutionsto the SAF.

Developing Competencies. While developersare actively engaged in developing C2I systems,the organisation needs to create space andtime for the developers to upgrade their skillsand develop new competencies. The effortincludes identifying emerging competenciesneeded for future challenges and committingresources, time and effort to develop the C2Iprofessionals. DSTA established the

Organisation Capability Development (OCD)entity in 2006 to embark on this endeavour.OCD works with the C2I developers to chartthe individual annual learning plan. The planhighlights the training needs of each developerand determines the training methods, scheduleand resources to build his/her capacity andcompetency. The investment in the developersand their training ensures the sustenance ofContinual Systems Development.

CONCLUSION

Continual Systems Development for C2I systemsis a risk management approach to developingwell-architected and integrated systems. Thisapproach has enabled the rapid developmentand deployment of C2I systems for the SAF,moving in tandem with its evolving andexpanding scope of desired capabilities totackle the challenges of the 21st century.

REFERENCES

80

Adams, G., Daniel, D. (2000). ManagingConcurrent Development – A SystemsEngineering Approach, AutotestconProceedings, IEEE.

Aoyama, M. (1993). Concurrent-DevelopmentProcess Model, Software, IEEE, Vol. 10, Issue 4,July, pp 46-55.

Boehm, B. (1998). A Spiral Model of SoftwareDevelopment and Enhancement. Computer,Vol. 21, No. 5, May, pp 61-72.

Boehm, B., and Hansen, W. (2001). The SpiralModel as a Tool for Evolutionary Acquisition.CrossTalk.

Cebrowski, Vadm A. and Garstka, J. (1998).Network-Centric Warfare: Its Origin and Future.Proceedings of the Naval Institute, January.

Infocomm Development Authority of Singapore(2006). Service-Wide Technical Architecture.R e t r i e v e d o n A u g u s t 2 0 0 7 f r o mhttp://intranet.igov.gov.sg/GovernanceandManagement/PoliciesAndStandards/SWTA/

Krygiel, A. J. (1999). Behind the Wizard’sCurtain. CCRP Publication Series.

Lee, J., Ong, M., Singh, R., Tay, A., Yeoh, L.W.,Garstka, J., Smith, E. (2003). RealisingIntegrated Knowledge-based Command andControl – Transforming the SAF. PointerMonograph No.2.

Yeoh, L.W., Chung, W.K., Cai, J. (2007).Emulator-Based C2 Development: A SystemsEngineering Approach To Developing C2Systems. Asia-Pacific Systems EngineeringConference, Singapore.

Yeoh, L.W., Lew, C. S., Cheung, D.S.K., Lee,C.C.W. (2007). Using Modelling and SimulationSystem-In-The Loop Solution to EnhanceProductivity and Service Level for MilitaryCommunication Systems Development. Asia-Pacific Systems Engineering Conference,Singapore.

Yeoh, L., Syn, H., Lam, C. (2007). An EnterpriseArchitecture Framework For DevelopingCommand and Control Systems. 17th AnnualInternational Symposium of the InternationalCouncil on Systems Engineering.

This paper was first presented at INCOSE 2008,15 - 19 June 2008 in the Netherlands and hasbeen adapted for publication in DSTA Horizons.

Dr Yeoh Lean Weng is Director (C4I Development and Systems Architecting).He is also concurrently the Deputy Director of the Temasek Defence SystemsInstitute and an Adjunct Professor at the National University of Singapore(NUS). Lean Weng has extensive experience working on large-scale defenceengineering systems. As a systems architect, he played a key role in developingthe Enterprise Architecture for defence applications. He also developed thesystems architecting methodology for masterplanning and defencetransformation. He is also the Vice-President of the INCOSE Singapore Chapter,INCOSE Region VI Representative to Member Board and the Chairman ofSystems Engineering Technical Committee, Institution of Engineers, Singapore.Lean Weng received his Bachelor and Master of Science degrees from NUSin 1983 and 1987 respectively. He further obtained two Masters in 1990 anda PhD in Electrical Engineering from the Naval Postgraduate School (NPS),USA in 1997.

BIOGRAPHY

Teo Tiat Leng is currently Deputy Director (Industry) in the Defence Industry& Systems Office, Ministry of Defence. He has more than 15 years of experiencein software engineering and systems engineering, developing C4I systemsfor the Singapore Armed Forces. His current work encompasses policy matterspertaining to the defence technology ecosystem. He received his Master ofScience (Computer Science) with Distinction from NPS, USA and a Master inDefence Technology & Systems, Master of Technology (Software Engineering)and Bachelor of Science (Computer & Information Science) degree from NUS.

81

Lim Horng Leong is a Programme Manager (C4I Development). He led thedevelopment of several large-scale command and control systems andsuccessfully fielded the systems for the Republic of Singapore Navy. He iscurrently applying Systems Engineering methodologies for C4Iexperimentation. Horng Leong received his Bachelor of Science degree(Computer and Information Sciences) from NUS in 1996. He also holds aMaster of Science in Systems Engineering from NPS, USA.

Software Safety –Back to Basics, Knowing Where to Tap!

~ A DSTA Perspective

ABSTRACT

Software Safety is becoming an important element of System

Safety as more and more hardware and equipment are run

and controlled by software. Customary approaches such

as hazard and fault tree analyses would help one to identify

and assign hazard(s) to either of a hardware or software nature.

This is only the beginning. Once software hazards have been

identified, software architects and developers would have to

design the application architecture to ensure the software

components are easy to test, verify, and maintain. This paper

will share the important aspects of white-box testing, and the

use of metrics as a proxy to measure the testedness, complexity

and maintainability of program codes. The paper will also

briefly describe the DSTA software quality journey, the DSTA

Software Quality Assurance (QA) & Testing Framework, the

governance to ensure compliance and the challenges faced by

the DSTA Software QA & Testing Programme in enforcing

compliance of the processes and quality enhancement activities.

Lian Tian Tse

84

INTRODUCTION

In the majority of accidents in which software

was used to control actions of components,

the cause can be traced to requirement flaws

such as incomplete requirements in the

specified and implemented software behaviour.

For example, wrong assumptions were made

on how the control system operates in the

production environment. However, even if

there exists a methodology or technique that

could identify all software-related hazards, we

could only conclude that “half the battle is

won”, and we are still presented with a huge

risk if the software is not properly designed,

built and tested.

It is common practice to use fault tree analysis1

to identify software-related hazards. The

analysis is usually carried out right down at

the software interface level in order to trace

the hazards into the software requirements

and design implementation. Furthermore,

software can do more than what is specified

in the requirements. Over-zealous developers

(or sales and marketing staff) may try to

introduce more ‘value-added’ features and

functions to impress and please the customers,

or the problem could be having unintended

functions unwittingly introduced to the system.

All these extras may potentially introduce

safety hazards into the system during operation

and therefore have to be treated with great

caution and care.

Software safety is a subset of system safety.

Hence, in the context of safety, we must ensure

that the system is protected against, and also

designed to handle unexpected software

behaviour. This paper will demonstrate how

good practices of software engineering suchas Software Quality Assurance (QA) and Testingare an essential and necessary function to help

enhance and promote software safety withinan organisation. Project teams are known todo a much better job in Software QA & Testingactivities when they know that there is anindependent body to carry out reviews andthat a quantitative criteria have beenestablished to assess whether due diligencehas been carried out.

THE IMPORTANCE OF WHITE-BOX TESTING

Many companies fall into the trap of acceptingsoftware solely based on the user acceptancetest (UAT), also commonly known as black-boxor functional testing in the software industry,where a set of pre-defined scenarios (orfunctional tests) was applied to the softwaresystem to verify actual results obtained againstthe expected results. This practice of softwareacceptance has the following shortcomings:

• We do not know the risks involved whenwe deploy a system in a productionenvironment.

• We are unsure whether all criticalcomponents / modules are appropriatelytested.

• UAT and Systems Integration Test (SIT) donot reveal which part of the program is nottested – there is no visibility of test coverage.

• Given the limited resources available, whereshould one focus his testing efforts to reducethe risk of programme and system failure?

This form of software acceptance is a verycommon practice in the InformationTechnology (IT) industry. The reason is notbecause the industry is unaware of betterpractices. Rather, the crux of the issue lies withthe customers and consumers notunderstanding the software development lifecycle and hence are not knowledgeable enough

Software Safety –Back to Basics, Knowing Where to Tap!

~ A DSTA Perspective

85

to know what to ask for or expect from thesoftware house. There are obvious economicreasons why most software houses would notg o b e y o n d t h e t r a d i t i o n a l U ATfor software acceptance. A holistic andcomprehensive acceptance methodology wouldentail an elaborate development process thatweaves in different kinds of testingrequirements, procedures, as well as qualityassurance and review activities. All thesetranslate into effort and cost. Regrettably, ifone does not know what to ask for, one wouldprobably not get what is rightfully due to him.

Although the system nature of safety impliesthat the black-box and integration testing willbe more relevant, white-box testing (alsoknown as developer testing) can also highlightsafety considerations when the safety-criticalfunctions and constraints have been traced toindividual modules and variables (Leveson,1995). White-box testing is carried out by the

developers as they write the software codes.The key benefit of white-box testing is thatdefects (programming bugs) or violations ofcoding best practices are promptly made knownto the developers. This reduces the likelihoodof allowing defects to be carried over tosubsequent phases of the development lifecycle. Studies have shown that the longer ittakes to detect the defect as it is promotedthrough its life cycle, the more difficult andcostly it is to fix the problem (see Figure 1).

White-box testing comprises static and dynamicanalyses. While it is possible for one to manuallyconduct code reviews, it is often more cost-effective to use automated tools to flag outbad practices and potential defects. Numeroustools exist in the market, for example, Parasoft’sC++ Test and Jtest, Agitar and McCabe IQsuites2. All these tools have a common objectivei.e. to help developers and managers takeguesswork, intuition and other unreliable

Figure 1. Benefits of early detection of defects

86

methodologies out of the picture. What thismeans is that the code itself is the basis forfinal management decisions to accept or rejectthe piece of software. However, every tool hasits strengths and limitations. In most cases,they cannot be used to check and verifyproprietary codes such as commercial-off-the-shelf products. White-box testing tools are alsonot able to check Enterprise Resource Products(ERP) like Oracle and SAP because the sourcecodes are not made available. The acceptanceof such ERP solutions are normally restrictedto functional, performance and security(vulnerability) testing. Hence, the decision toselect which tool to invest in will depend largelyon the goals and objectives of using such toolsin the first place. More importantly, do wehave a set of evaluation criteria that has beencarefully thought through to help us decideon the tool that is the most cost-effective andeasy to use?

Some schools of thought believe that testingis success-oriented. This is true to a certainextent as most testing tools focus on doingwhat the software is supposed to do i.e.providing required outputs, rather than onwhat it is not supposed to do or on the effectsof failures and errors on safety-relatedbehaviour. Some have even gone further tosuggest that erroneous inputs are often nottested. This truth is more evident in the pastbut with the advancement of technology inrecent years, we are beginning to see tools inthe market today that can generate run-timetest data to break software codes (dynamictests). However, operator errors and somesoftware error conditions may not be fullyaddressed by such state-of-the-art testing tools.These are extremely important for safety andhence have to be complemented with othertechniques and methodologies. It is importantto note that the safety of a piece of softwarecannot be evaluated by simply looking at thesoftware alone.

USE OF SOFTWARE METRICS TO ENHANCE SOFTWARE SAFETY

DeMarco’s insight in 1982 has now become awell-known adage in software engineering:“Managers cannot control or manage whatthey cannot measure.” It has always been achallenge to implement software metrics inan IT organisation to control and managesafety and quality. Often, good intentions aremet with unintended results and behaviour.Developers are known to be suspicious ofwhether the software metrics would be usedagainst them in their performance appraisal(though conversely, it could also be used tojustify higher rewards if the metrics showedthat they were producing high quality codes).This common fear reflects the prevailing culturein most organisations (and the industry) andis underpinned by the management’s level oftrust and willingness to forgive whenemployees make mistakes.

Instituting a new process or a new method ofdoing things in a large organisation can be anuphill and challenging task. There will alwaysbe people who support the initiative; thosewho oppose the initiative; and those who areindifferent. If an IT organisation does not havea strong quality culture, then the introductionof software metrics without some form ofchange management framework will lead tolimited success at best or total failure in mostcases. It has been discovered that applying theCMMI’s 10 generic practices (CMM IntegrationProject, 2001) is a good starting point for thosewho are interested in making software metricsor any new process operational in theirorganisation. However, there is no guaranteeof success. It will require more than these 10best practices and additional factors includestrategic planning, timing and leveragingopportunities.

87

One of the most effective tools in makingthings safer is simplicity and building systemsthat are intellectually manageable. In today’senvironment, it is becoming increasinglydifficult for managers to handle softwaredevelopment projects because software isgrowing to previously unimaginable size andcomplexity. There is probably more softwarethan hardware components in any equipmenttoday as compared to a few decades ago.A good example is the aircraft industry.Furthermore, it is increasingly common thatthe codes being managed are produced bysomeone else such as third party contractorsor different business units within the company.Under such environments, the ability tomanage a software application effectivelydepends on accurate and complete assessmentsof the code.

The use of software metrics like cyclomaticcomplexity (McCabe and Watson, 1996)measures the amount of decision logic in asingle software module. It gives the numberof logical paths or the recommended minimumset of test paths required to pass through everydecision at least once in the software module.Another useful metric is coverage analysis.

Coverage analysis can instantly identifyuntested paths and branches, allowing projectmanagers to streamline test plans to addressonly untested parts of the program. Thisprevents common problems of redundanttesting or over-testing program areas that areat low risk of defects. Coverage testing istherefore a very useful and powerful tool foridentifying and analysing untested paths insafety critical modules.

Some tool vendors provide extensivevisualisation capabilities with tools like battlemaps and flow graphs. Battle maps display astructure chart that graphically represents thefunctional hierarchy of the program. Often,the tool has an ‘exclude’ feature for developersand designers to focus on specific segments ofthe program. Flow graphs graphically displaythe control structure of the code and path ofexecution inside each module. It facilitatescode comprehension and is often used for codereviews. They are very good for highlightingunstructured behaviour of the code anddevelopers and users alike can graphically seethe cause of increased code complexity(see Figure 2).

Figure 2. Example of a flow graph

88

The Software Engineering Institute of theCarnegie Mellon University recommends thefollowing benchmarks for the measurementof cyclomatic complexity (see Table 1).

From the perspective of a customer withoutmuch programming knowledge or ITgrounding, one could easily comprehend thecomplexity of the developed codes andwhether they are easy to test and maintain bysimply looking at the graphical display of thebattle maps and flow graphs. These are veryeffective tools to manage the quality andmaintainability of the codes. With the aid ofsuch tools, it is more compelling for softwarearchitects, developers and third partycontractors to refactor their codes when theyare presented with a complex flow graph thatshowed that it is almost impossible to test theirsoftware. Hence, the use of these toolsfacilitates simplicity and the building ofsoftware systems that are intellectually moremanageable and hopefully make things safer.In addition, such visibility also provides a moreobjective assessment of the operations andsupport costs incurred by both contractors andcustomers alike.

THE DSTA SOFTWARE QUALITY JOURNEY

Since the early 1990s, DSTA has been usingISO9001 as the foundation for theorganisation’s quality management system.DSTA has an elaborate quality manual thatarticulates the processes and procedures thatproject teams will have to comply with in thecourse of their work. Despite these processesand procedures, we still encounter quality andperformance issues when systems weredeployed in the production environment. Ithas become clear that a good ISO9001 systemis insufficient – we have to do more, likesubjecting the software codes to screening andquality control.

DSTA started a centralised test service toprovide performance testing services to project

teams in July 2004. This test is usually performedprior to the deployment of the system. Whilewe see a significant improvement in theperformance of the deployed systems thathave undergone the test, there are still issueswith regard to programming bugs and poorquality codes that are too complex and difficultto maintain. Quality has to be built into thesystem at the onset of the project life cycle.One of the most effective ways to address thisproblem is to introduce white-box testingmethodology into the organisation. In June2005, a small team was formed to explore andmake automated white-box testing operationalin the organisation. At the same time, anotherteam was involved in the development of aSoftware QA & Testing Framework (Lian andChew, 2006) which is built on top of theorganisation’s quality management system.The framework was endorsed by DSTA seniormanagement on 8 February 2006 (seeFigure 3).

DSTA was fortunate in that the seniormanagement is very supportive of these qualityinitiatives. In April 2006, DSTA officially set upthe Software QA & Testing Programme(SwQAT) with the vision of providing leadershipin the delivery of quality software that is notonly reliable and easy to maintain but also safeto use. The senior management also endorsed

Table 1. Benchmarks for measuringcyclomatic complexity

Cyclomatic RiskComplexity Evaluation

1 – 10 a simple program,without much risk

11 – 20 more complex,moderate risk

21 – 50 complex, high riskprogram

Greater than 50 untestable program(very high risk)

89

SwQAT’s recommendation for all projects to

undergo pre-deployment code screening. By

the end of July 2007, SwQAT had managed to

perform code screening on more than 70

projects.

SOFTWARE QA &TESTING GOVERNANCE

DSTA outsources a large number of projects

to the industry. In order to ensure that the

systems delivered to DSTA by the contractors

are of high quality, it is important that

quantitative measures are used for the

acceptance criteria. There are many types of

testing that a piece of software has to undergo

throughout its life cycle. The more obvious

ones are normally ‘milestone-kind’ of testing

which includes interface testing, factory

acceptance test, UAT and performance testing.

The less obvious yet important ones like white-

box and application security testing are often

taken for granted and overlooked by

developers. SwQAT has identified the following

types of testing to be included in the tender

contract which contractors must fulfill when

they bid for the projects: (1) white-box testing,(2) black-box testing, (3) coverage testing forsafety-critical and mission-critical components,(4) application security testing, and (5)performance testing.

For white-box testing, SwQAT uses cyclomaticcomplexity and violation of industry codingbest practices as a proxy for systems acceptance.In addition, for safety-critical and mission-critical modules, coverage testing is to beperformed to determine the percentage ofcodes that has been covered by the black-boxtesting. The white-box testing tool that weuse is also able to identify unused codes anduntested codes which are undetectable by thetraditional black-box testing. Unused anduntested codes could be potential sources ofsafety risk and hence should be identifiedand verified.

In DSTA, black-box testing is carried out by theproject teams together with the users to

Figure 3. DSTA Software QA & Testing Framework

90

validate and verify the functional requirements.SwQAT will help to instrument the source codesfor components that need to undergo coveragetesting. The project team will then carry outthe coverage testing as part of the black-boxtesting and SwQAT will analyse and interpretthe results (test logs) for the project teams aspart of the test services rendered to projectteams.

SwQAT will also conduct vulnerabilityassessment on software systems to identify anysecurity risks. Both white-box and black-boxapplication security tools are used for thispurpose. For pragmatic and sometimesbusiness-related reasons, defects and bugs(those which are not catastrophic in nature)may be intentionally left uncorrected in thenext new release. However, for securityvulnerabilities that have been identified, it ismandatory that these defects must be fixedbefore they are deployed.

Performance testing is usually the last test tobe carried out before the system is rolled out.Depending on the objective of the performancetest, SwQAT can conduct load, stress andvolume tests to identify bottlenecks in thesystem. This test is often carried out at a stagingenvironment and also at a pilot trial at theproduction environment without prior noticebefore declaring the system operational. Oncethe bottlenecks have been identified, therespective experts from the server, databaseand network teams would work collectively tofine-tune the system.

SwQAT will produce test reports for white-box,application security and performance testings.Black-box testing report will be done by theproject team. After each test, SwQAT will briefthe project teams on the findings, analysesand recommendations on how to fix theidentified defects. The final approval to releasethe system for deployment is under the purviewof the Programme Director.

CHALLENGES FACED BY SWQAT

In order to be effective, SwQAT needs to havea pool of talented professionals who are notonly technically competent in softwarearchitecture design and coding practices, butalso first class salesmen with the tenacity andpeople skills to influence and convince theirpeers to do things differently to achieve betterresults. This combination of technical and softskills is not easy to find in any organisation.Finding experienced staff with three to fouryears of development experience in theorganisation to staff SwQAT is a big challengeas we are always competing with the in-houseproject teams for talents to deliver systems toour customers. Mid-career recruitment is alsodifficult as many prefer to work in theprivate sector.

SwQAT took almost three years to build up ateam of capable professionals from four to 15staff specialising in white-box, black-box,performance and application security testing.Since early 2007, we have been focusing ourefforts on building up our skills in applicationsecurity testing. The initial startup waschallenging but we were fortunate to be ableto tap on two software architects to help kick-start the team on a part-time basis. Thesoftware architects were invaluable to theteam in helping to verify and select the ruleset to be used for flagging out the codeviolations and security vulnerabilities as wellas developing guidelines and standardoperating procedures.

Although SwQAT contributes directly to thebranding of the organisation, we were notalways taken seriously by the project teams at

large. As a QA & Testing outfit, SwQAT is

viewed as a cost driver and may potentially be

blamed for causing delays in schedule.

Furthermore, our contributions may not always

91

be appreciated. Nonetheless, SwQAT’s staff

know that their work and contributions to the

organisation are important and valued by

DSTA’s senior management, and our customers,

the Ministry of Defence and the Singapore

Armed Forces.

SwQAT has taken great pains to inculcate the

quality culture in the organisation. For teams

who are willing to work with us, we will be

there to help them ‘smoothen’ their learning

curve and also provide technical support to

resolve any technical issues or problems

encountered. SwQAT also conducts regular

training courses to educate staff on the DSTA

Software QA & Testing Framework and

governance, and hands-on technical training

on the use of the automated tools. Through

our sincerity and professionalism, we are

beginning to win over the project teams. By

word of mouth, news of the value-add

provided by SwQAT has spread and we are

beginning to see an influx of requests for our

services. For SwQAT to be successful, we operate

on the principle of ensuring that our

customers i.e. project teams are successful.

To meet the next challenge in achieving

software safety capability, SwQAT has to train

her staff to understand system safety concepts

and techniques. This is the simple part as we

can easily set aside time for the staff to attend

relevant courses, workshops and conferences

as well as provide them with opportunities to

work on system safety projects to gain valuable

hands-on experience. On a larger scale, the

more challenging task is to train our software

developers to understand not only software-

related hazards, but also any software

requirements plays in the implemention ofinterlocks and other safety design features.The question is whether we can train thesoftware developers such that they will notinadvertently disable or override system safetyfeatures and implement software-controlledsafety features incorrectly. The other difficult

challenge is to get system safety engineers tobe more involved in software developmentand to include software in the system safetyprocess. This is an aspect SwQAT has to workon as we begin to explore and move into thisnew domain of work.

CONCLUSION

There are so many software paths, possibleinputs and hardware failure modes that testingfor all of them is not feasible. A good testingstrategy may chance upon a few hazardoussoftware errors or behaviour, but this is farfrom a rigorous way to identify them. Thispaper advocates sound Software QA & Testingpractices as a means to complement softwaresafety.

In a nutshell, with the aid of advancedautomated test tools, software safety can easilybe complemented and enhanced through (1)the identification of unused and untestedcodes; (2) the highlighting of unstructuredbehaviour, as well as (3) the use of an excellenttool to identify the impact on change analysisgiven that software systems do change overtime due to new or defunct requirements bothduring software development and operationsand support phases. Also, central to anyorganisat ion engaging in softwaredevelopment activities is the need to have asystem for configuration management, a goodtraceability matrix to trace requirements forthe determination of cases to test scripts, andthe execution of the tests.

REFERENCES

92

Arthur H. Watson, Thomas J. McCabe. (1996).Structured Testing: A Testing MethodologyUsing the Cyclomatic Complexity Metric. NISTSpecial Publication 500-235, 1996.

Lian Tian Tse, Chew Wee Hui, Defence Science& Technology Agency. DSTA Software QA &Testing Framework, 2006.

Nancy G. Leveson, (1995). Safeware: SystemSafety and Computers. University ofWashington. Addison-Wesley PublishingCompany, 1995.

SEI CMMI V1.1: CMM Integration Project, 2001.R e t r i v e d o n O c t o b e r 2 0 0 3 f r o mhttp://www.sei.cmu.edu/cmmi

1 There are many types of hazard analysisranging from a simple checklist to moresophisticated techniques like fault tree analysis.Selection of which technique to use dependslargely on the goals of the analysis as eachtechnique has its corresponding strengths andweaknesses.

2 Disclaimer: Certain trade names and companynames are mentioned in the paper. In no casedoes such identification imply recommendationor endorsement by DSTA, nor does it implythat the products are necessarily the bestavailable for the purpose.

This paper was first presented at the International

Systems Safety Conference (13 -17 August 2007)

and has been adapted for publication in DSTA

Horizons.

ENDNOTES

Lian Tian Tse is Assistant Director (Command and Control InformationTechnology Competency Community) and is concurrently heading the SoftwareQA & Testing Programme in DSTA. He has extensive experience in thesolutioning and implementation of Decision Support Systems for the Ministryof Defence and the Singapore Armed Forces. For the last 15 years, he hasbeen working in the IT arena leading projects and programmes that covera wide domain ranging from Logistics, Operations, Manpower, Training andEducation. A recipient of the Defence Technology Group Scholarship, heobtained his degree in Mechanical Engineering from the National Universityof Singapore. He also holds a Master of Science in Operations Research fromthe Naval Postgraduate School, USA.

BIOGRAPHY

93

Hazard Re-classificationof 76mm Naval Gun Ammunition

following UN Test Series 6

Audrey Lao Linmei

Yen Chong Lian

Ng Cher Chia

ABSTRACT

The hazard classification of ammunition has a significant impact

on the maximum quantity allowable and minimum safety

distance required for its storage and transport. The 76mm

Multi-role Oto Munition and Semi-armour Piercing Oto

Munition rounds are gun ammunition used onboard the

Republic of Singapore Navy platforms. The manufacturers

assigned the rounds to Hazard Division (HD) 1.1. In comparison,

US Navy 76mm rounds with identical Comp A3 high explosive

fillings were classified as HD 1.2. DSTA suspected that the

76mm round’s hazard classification was conservative. Thus, a

series of confined and unconfined sympathetic detonation and

fast cook-off tests were conducted on rounds in storehouse

packaging (wooden crates) and storage afloat packaging

(plastic containers, referred to as octovals) in December 2007.

The test results indicated that the hazard classification could

be reduced to HD 1.2. Other than safety improvements and

risk reduction, the benefits of a reduced hazard classification

were enhanced emergency response and platform survivability,

easing of berthing constraints and increased storage flexibility

and capacity. This paper describes the motivating factors, test

programme and results.

96

INTRODUCTION

Class 1 Explosives comprise six hazard divisions(HD), HD 1.1 to 1.6, with HD 1.1 as the mosthazardous and requiring the largest separationdistances for its safe storage. According to theUN Model Regulations (United Nations, 2007),it is mandatory for explosive substances andammunition to be classified into one of the sixhazard divisions before they are transported.

Hazard Classification Methods. With referenceto current practice in the US and the UK, thehazard classification of new ammunition shouldbe assigned by a qualified explosives experton the basis of testing or by analogy.

a. Hazard Classification by Testing. The testprocedures from the UN’s Manual of Tests andCriteria (United Nations, 2003) provide thenecessary information for the competentauthority of each country to assess the hazardof explosive substances and articles, such thatan appropriate classification for transport canbe made.

b. Hazard Classification by Analogy. The hazardclassification of new ammunition may beassigned by analogy in terms of similarities inexplosive filling, design features and packaging(Department of Defense Ammunition andExplosives Hazard Classification Procedures TB-700-2 Review Draft, 2005) to those that havealready been hazard classified. This is theprerogative of the country’s competentauthority. Analogy using analytical studies isa common way of assigning a hazardclassification to ammunition.

c. Hazard Classification by Conservatism. Whenneither tests nor analytical studies have beendone, ammunition may be conservativelyassigned to HD 1.1 – the most hazardousdivision in Class 1. A conservative hazarddivision leads to larger safety distances requiredfor the storage and transport of ammunitionand greater operational constraints than arenecessary based on the ammunition’s truehazard.

Review of Applicable Tests for 76mm RoundsHazard Classification Reduction. A series oftests (Series 1 to 8) described in the Manual ofTests and Criteria have been devised todetermine if substances or articles arecandidates for Class 1 Explosives and the hazarddivision they belong to. Series 1 to 8 alsodetermine how sensitive explosives are toenergetic stimuli such as friction, impact, heat,and shock. However, it was determined thatonly Series 6 tests were applicable to justifythe revision from HD 1.1 to HD 1.2.

MOTIVATING FACTORS FOR THE 76MMROUNDS HAZARD RE-CLASSIFICATION TEST

Cost. Additional expenditure was incurred onlyfor actual hazard re-classification testing andpackaging re-labelling. The 76mm roundstested were reaching the end of their shelf-life and due for disposal. Therefore, theammunition cost was not a significant factor.

Benefits. One of the benefits of hazardclassification reduction is the relaxation of thequantity distance (QD) criteria required duringthe ammunition life cycle shown in Figure 1(Parsons et al., 2000). QD is the safe separationdistance required during the storage andtransportation of explosives to minimise injuryto people and damage to property in the eventof an explosion. The QD criteria are greatlyinfluenced by the explosive hazardclassification. For the Republic of SingaporeNavy, hazard classification reduction of the76mm rounds resulted in greater operationalflexibility during ammunition storage andwharf-to-ship transportation, since the safeseparation distance required for those activitieswas decreased.

Reduced hazard classification of ammunitionconfers advantages other than safetyimprovements and risk reduction. Enhancedplatform survivability, increased storagecapacity, storage flexibility and easing ofberthing constraints are the other benefits(Barnes & Cheese, 2000).

Hazard Re-classificationof 76mm Naval Gun Ammunition

following UN Test Series 6

97

a. Safety improvements/risk reduction. HD 1.1has a mass detonation hazard, whereas HD 1.2has a projection but not a mass explosionhazard. Therefore, the risks to personnel fromstorage, handling and operations involving HD1.2 are lower.

b. Emergency response. Sympatheticdetonation (i.e. unintended detonation of anexplosive charge by exploding another chargeadjacent to it) and fast cook-off (i.e. directexposure to an intense fire e.g. liquid fuel fire)timing and consequences may be derived fromthe mandatory Series 6 tests. The cook-off timeand response of ammunition may be listed ina compendium of safety data which would aidthe ship’s crew during an emergency. Inparticular, the US Navy and Royal Navy havealready established information flow channelsand awareness in the ship’s crew on the cook-off behaviour of specific shipboardammunition. For example, the Royal Navycompiles ammunition safety data as a ShipExplosives Safety Store Instruction (JSP 430,2005).

c. Easing of warship berthing constraints.According to rules governing the mixing ofhazard divisions, regardless of the quantity ofexplosives, when HD 1.1 is stored together

with lower hazard divisions, the entire load issubject to HD 1.1 QD criteria. This poses aconstraint on the ship’s berthing requirements.With hazard classification reduction, the QDcriteria for storage and transport would berelaxed. Explosive loading piers and wharvesin proximity to on-base or publicly exposedsites will also benefit from the easing of QDrequirements.

d. Increased ammunition storage flexibility.With hazard classification reduction, 76mmrounds may be stored with HD 1.2, 1.3 and 1.4items and the load treated collectively as HD1.2. This represents a greater storage flexibilitythan a HD 1.1 load, particularly when themajority of the items belong to a lower hazarddivision. Storage of lower hazard divisionammunition also requires less facilityhardening. This translates to lowerinfrastructure demand.

e. Increased storage capacity. Risk reductionleads to less stringent QD criteria. With theexisting munitions infrastructure i.e. unchangedseparation distance and building structures,more HD 1.2 explosives may be stored thanHD 1.1 and more combat ships may besupported.

Research & Development

Production

Handling

Transportation

Storage

Handling

Transportation

Operations Disposal

Figure 1. Ammunition life cycle

98

f. Enhanced platform survivability. Theoperational asset loss (i.e. ship’s damage)incurred in the event of an accident or a hostileaction is smaller for HD 1.2 than HD 1.1. Hazardclassification reduction would lead to lowerconsequential risks to the ship from the storedammunition.

g. Overall benefits. These are shown in Table1 using generic inputs. K22.2 refers to thesafety factor applied for inhabited buildingswhile K16 refers to the safety factor appliedfor warship loading and unloading.

TEST METHOD

The UN Series 6 tests were planned forimplementation as follows:

Type 6(a): Single Package Test. A test on asingle package to determine if there is massexplosion of the contents.

Type 6(b): Stack Test. A test on packages ofexplosive articles to determine whether anexplosion is propagated from one package toanother.

Type 6(c): Liquid Fire/External FireTest. A test on packages ofexplosive articles to determine ifthere is mass explosion or a hazardfrom dangerous projections,radiant heat and/or violentburning or any other dangerouseffects when involved in a fire.

Possible Waiver. According to Series 6procedures, it is not always necessary to conductall the tests. The Stack Test can be waived if inthe Single Package Test:

a. the exterior of the package is notdamaged by internal detonation and/orignition, or

b. the contents of the package fail toexplode, or explode so feebly thatpropagation of the explosive effect fromone package to another in the Stack Testis excluded.

Test Packaging. It is a UN Series 6 requirementto test articles in their realistic forms oftransport . Packaging and stack ingconfigurations affect results due to thedifferential absorption of energy. The 76mmall-up rounds are stored in two types ofpackaging: wooden crates in storehouses andoctovals onboard ships. The wooden cratepackaging (see Figure 2) is from the originalequipment manufacturer and comprises tworounds in a head-tail (projectile to cartridge)layout, each packed in a fibrous cylinder. One

Parameters HD 1.1 HD 1.2 Remarks

K22.2 Assuming an earth-covered QD Criteria for Storage = 403m 350m storehouse with 6 tons Net

Explosives Quantity

QD Criteria for K16 90m Assuming a ship loaded withWarship (Un)loading = 291m (fixed) 6 tons Net Explosives Quantity

2.1e-7 4.5e-8 Assuming a 50-feet separationIndividual Risk fatalities fatalities between an open storage site

per year per year and an open exposed site

Table 1. Benefits of hazard classification reduction from HD 1.1 to HD 1.2

Figure 2. Wooden crate packaging schematic

99

wooden crate contains two cylinders. Octovalpackaging (Figure 3) comprises two rounds ina head-head (projectile to projectile) layoutpacked into a high impact acrylonitrilebutadiene styrene plastic container.

SINGLE PACKAGE SYMPATHETIC DETONATION TEST

Definitions. The 76mm Multi-role Oto Munition(MOM) round contains pre-formed fragmentswhile the 76mm Semi-armour Piercing OtoMunition (SAPOM) round contains apenetrating warhead. All 76mm roundscomprise a projectile (containing the highexplosive warhead which delivers the weaponeffect) joined to a cartridge case (containingthe propellant which propels the round). Eachtest package comprises two rounds in eitherthe wooden crate or octoval packaging. Oneround is known as the donor, since it ispurposely initiated and ‘donates’ the blast.The other round is known as the acceptor.During the sympathetic detonation testing ofrounds, the assessment is based on the responseof the projectile and not of the cartridge. This

Figure 3. Octoval packaging schematic

is because the high explosive is contained inthe projectile. There are three types of testsprescribed by UN Series 6, namely 6(a) SinglePackage Test, 6(b) Stack Test and 6(c) LiquidFuel/External Fire Test. However, the SinglePackage and Stack Test have multiple trials(e.g. Single Package Test trial 1, Single PackageTest trial 2). The Liquid Fuel/External Fire Testhas one trial each. UN Series 6 prescribes thenumber of trials for each test type.

Set-up. A mild steel witness plate was placedbelow the test package (see Figure 4). The testpackage and witness plate were confinedwithin a metre of sand on all sides (see Figures5, 6 and 7) and the donor projectile wasinitiated. The witness plate and acceptor round

Figure 7. Completed confinement set-upFigure 6. Levelling sand

Figure 5. One-metre sand confinementFigure 4. Test package on witness plate

100

were later recovered for analysis. Four sets oftests were conducted (MOM in wooden crate;MOM in octoval; SAPOM in wooden crate;SAPOM in octoval) with three trials eachaccording to the requirements of UN Series 6.Of the three trials, two were confined and onewas unconfined.

Summary of Results. The 76mm rounds passedthe single package sympathetic detonationtests. Only one crater with a diameter notsubstantially larger than 76mm was found oneach witness plate. This indicated that onlythe donor projectile detonated. All acceptorprojectiles were recovered dislodged from theirrespective cartridge cases. For the MOM, rowsof preformed fragments surrounding the highexplosive core were observed, which showedthat the acceptor high explosive did not react.For all confined tests, there was a gentledisruption and scattering of sand. Greater sanddisruption was observed for MOM due to itshigher Net Explosives Quantity (NEQ).

For the MOM, the acceptors were recoveredintact for the wooden crate tests and shearedfor the octoval tests. This was due to the head-head ammunition configuration in the octoval,as compared to the head-tail ammunitionconfiguration in the wooden crate, as well asthe thin skin of the fragmenting warhead.

For the SAPOM, the acceptors were recoveredintact in both wooden crate and octoval singlepackage tests. This was due to its lower NEQand the thicker skin of the penetratingwarhead.

MOM in Wooden Crate – Test Results

a. Confined trials – A white gas followed bya brown gas was observed. This is due toburning of the nitrogen-based propellant.Large pieces of the cartridge case, cardboardcanister and the intact acceptor wererecovered (Figure 8). Identical results wereobserved for trials 1 and 2.

b. Unconfined trial – The intact acceptorprojectile was thrown eight metres away.Large pieces of the wooden crate, cartridgecase and cardboard canister were recovered(Figure 9).

MOM in Octoval – Test Results

a. Confined trials – Identical results wereobserved for trials 1 (Figure 10) and 2(Figure 11). The projectile case was rupturedand displaced. However, the high explosivedid not react.

b. Unconfined trial – The acceptor projectilewas found intact on the sandbank 12metres away (Figure 12). The projectile casewas displaced and rows of preformedfragments were revealed (Figure 13).

Figure 9. MOM in wooden crate – confined trial

Figure 8. MOM in wooden crate –confined trial 1

101

SAPOM in Wooden Crate – Test Results

a. Confined trials – A crater not substantiallylarger than the diameter of SAPOM wasfound on the witness plate (Figure 14). Largepieces of the cartridge case and intactacceptors were recovered (Figure 15).Identical results were observed for trials 1and 2.

b. Unconfined trial – The intact acceptorprojectile was dislodged from its cartridgeand thrown onto the left sandbank 10metres away. Large pieces of the woodencrate, cartridge casing and cardboard canisterwere recovered (Figure 16).

Figure 16. SAPOM in wooden crate –unconfined trial

Figure 15: SAPOM in wooden crate –confined trial 2

Figure 14: SAPOM in wooden crate –confined trial 1

Figure 13. Close-up of projectile case

Figure 12. MOM in octoval – confined trial

Figure 11. MOM in octoval – confined trial 2Figure 10. MOM in octoval – confined trial 1

102

SAPOM in Octoval – Test Results

a. Confined trials – A crater not substantiallylarger than the diameter of SAPOM wasfound on the witness plate. Large pieces ofthe cartridge case, octoval and the intactacceptor projectile are shown in Figures 17,18, 19 and 20. Identical results were observedfor trials 1 and 2.

b. Unconfined trial – The intact acceptorprojectile was dislodged from its cartridgecase, as shown in Figure 21. The acceptorcartridge in its octoval was found 60 metresfrom the test site as shown in Figure 22.

STACK SYMPATHETIC DETONATION TEST

Since the 76mm rounds passed the SinglePackage Test, the Stack Test was not conducted.This is permissible according to UN TestMethods & Criteria. Figure 22. SAPOM in octoval – unconfined trial

Figure 21. SAPOM in octoval – unconfined trial

Figure 20. SAPOM in octoval –fragments from confined trial 2

Figure 19. SAPOM in octoval –crater from confined trial 2

Figure 18. SAPOM in octoval –fragments from confined trial 1

Figure 17. SAPOM in octoval –crater from confined trial 1

103

LIQUID FIRE/EXTERNALFIRE (FAST COOK-OFF) TEST

Set-up. A stack of nine packages (18 rounds)

was encircled with a steel strip for support and

positioned on a grid test stand above a fuel

receptacle, which extended beyond the grid

by one metre. It was filled with sufficient Jet

A1 fuel for a fire lasting 30 minutes. The

distance between the grid platform and

receptacle was half a metre. Vertical witness

screens were erected on three quadrants, four

metres from the edge of the ammunition stack.

When the wind speed was less than 6m/s, the

fuel was ignited simultaneously on two sides,

one on the upwind side. Cook-off was indicated

by flying burning debris seen or explosions

heard during the test. After the test, the

remaining ammunition was collected foranalysis. The passing criteria was no massexplosion (no simultaneous detonation of morethan 50% of acceptors). Four tests were done(MOM in wooden crate; MOM in octoval;SAPOM in wooden crate; SAPOM in octoval)as required by UN Series 6. Figure 23 showsthe test set-up.

Summary of Results. For all tests, sporadicexplosions were observed and more than nineacceptors out of the 18 rounds were recovered.This meant that less than 50% of the acceptorsexploded simultaneously. Thus, there was nomass explosion. This showed that the 76mmrounds passed the fast cook-off test. Table 2summarises the quantity of projectiles whichreacted for each fast cook-off test. The fastcook-off response was an explosion of thepropellant in the cartridge case. The fast cook-off time was less than two minutes.

Table 2. Summary of liquid fuel/external fire (fast cook-off) test results

Figure 23. Fast cook-off test set-up

S/N Ammunition Packaging Initial Quantity of Quantity of ResultType Quantity of Projectiles Projectiles

Rounds Recovered Reacted

Inside OutsidePit Pit

1 MOM Wooden 12 2 4Crate

2 MOM Octoval 6 3 918

3 SAPOM Wooden 11 3 4Crate

4 SAPOM Octoval 10 7 1

NoSympatheticDetonation

104

REFERENCES

Barnes, P. and Cheese, P.J. (2000). IM TechnologyApplied to Large Calibre Naval Ammunition,Part 1: Strategy and Implementation. DefenceLogistics Safety, UK MoD. A paper from the29th DDESB Seminar.

Department of Defense Ammunition andExplosives Hazard Classification Procedures TB-700-2 Review Draft. (2005). Departments ofthe Army, the Navy, the Air Force, and theDefense Logistics Agency, Washington, DC.

Table 3 summarises the UN Series 6 test resultsfor the 76mm MOM and SAPOM rounds inthe wooden crates and octoval packaging.

CONCLUSION

The 76mm MOM and SAPOM rounds in thewooden crates and octoval packaging passedthe UN Series 6 tests. This justified a reductionof the hazard division from HD 1.1 to HD 1.2.Safety data on the fast cook-off timing andresponse were obtained as a useful by-product.The land use savings enabled by the hazardclassification reduction is significant for a smallcountry such as Singapore. The planning andconducting of the full UN Series 6 tests was avaluable learning opportunity. Future workincludes the testing of other ammunition thatmay be over-classified.

Table 3. Summary of UN Series 6 test results

S/N Ammunition Packaging Single Stack Fast HazardType Package Test Test Cook-off Classification

1 MOM Wooden Crate

2 MOM Octoval

3 SAPOM WoodenCrate

4 SAPOM Octoval

NoSympatheticDetonation

NoSympatheticDetonation

Confined Unconfined

NotRequired(passedSingle

PackageTest)

JSP 430 MoD Ship Safety Management. (2005).Part 3 Procedures Chapter 8 Explosives. UKMoD Ship Safety Management Office, SeaTechnology Group, Defence ProcurementAgency.

JSP 482 MoD Explosives Regulations. (2006).Explosives Storage and Transport Committee,UK MoD.

Parsons, G., Hayles, D., Carlton, P.N., Osborn,D.M., Davis, J.E., Hester, R.E. (2000). HazardClassification Reduction Assessment. Air ForceResearch Laboratory Energetic MaterialsBranch, KBM and Sverdrup Technology. A paperfrom the 29th DDESB Seminar.

ST/SG/AC.10/11/Rev.4. (2003). Recommendationson the Transport of Dangerous Goods, Manualof Tests and Criteria. United Nations.

ST/SG/AC.10/1/Rev.15. (2007). Recommendationson the Transport of Dangerous Goods, ModelRegulations. United Nations.

This paper was first presented at the 2008 United

States Department of Defense Explosives Safety

Seminar on 14 - 16 August 2008, Palm Springs,

California, and has been adapted for publication

in DSTA Horizons.

HD 1.2

105

Audrey Lao Linmei is a Senior Engineer (Armament Systems – ExplosivesSafety). She has prepared safety assessment reports for the safe storage ofexplosives onboard surface ships and submarines and has conducted hazardclassification testing of ammunition. She is presently developing a hazardclassification implementation policy and is leading the development of therisk-based explosives safety siting in the Singapore Armed Forces (SAF). Shegraduated with a Bachelor degree in Chemical Engineering (Honours) fromthe National University of Singapore in 2005.

BIOGRAPHY

Yen Chong Lian is a Senior Principal Engineer and Head (Armament Systems– Explosives Safety). He manages the licensing of military explosive facilitiesin the SAF and is the Chairman of the Explosives Safety Technical Sub-Committee of Explosives Fire and Chemical Safety Committee. He is alsothe President of the Institute of Explosives Engineers (Singapore) and aconsultative member for the Control of Strategic Goods. He holds apostgraduate engineering diploma in Explosives Ordnance Technology fromCranfield University, UK.

Ng Cher Chia is a Deputy Commander and concurrently the Head of theSystem Office in the SAF Ammunition Command. He has substantial workexperience from his appointments in operational units, infantry brigades,general staff departments, Army Logistics and in the SAF AmmunitionCommand. Cher Chia was awarded the Defence Technology Prize Award in2008 for his contributions to the Underground Ammunition Facility project.He has a Master in IT Management and a Master of Engineering (TechnologyManagement) from the University of South Australia.

Business Intelligencein Government Procurement

ABSTRACT

Over the years, Singapore's Government E-Business (GeBIZ)

Portal has accumulated a knowledge base of procurement

data. This data is a valuable source of market knowledge and

intelligence. The paper will detail how Business Intelligence,

via a three-pronged approach of Intelligent Procurement,

Portfolio Management and Performance Management, has

helped GeBIZ users exploit past procurement experiences to

reduce turnaround time, increase productivity and ensure

accountability of public funds.

Carol Lo S Chia

Leo Chen Hong

108

INTRODUCTION

Launched in 2000, the Government ElectronicBusiness (GeBIZ) Portal is an integrated portalfor use by all Singapore government agenciesto conduct business electronically with theirsuppliers. There are more than S$10 billionworth of business opportunities publishedannually to 30,000 suppliers registeredwith GeBIZ.

Over the years, GeBIZ has accumulated aknowledge base of procurement data. Thisdata is a valuable source of market knowledgeand intelligence. To exploit this untappedsource of information, a Business Intelligence(BI) platform was introduced to help improvethe efficiency and effectiveness of governmentprocurement and raise it to a more strategiclevel of intelligent sourcing and purchasing.This paper will detail how BI has helped usersof the GeBIZ Portal reduce turnaround time,increase productivity, ensure accountability ofpublic funds and highlight the challenges inthe implementation.

BUSINESS INTELLIGENCE

The term ‘Business Intelligence’ dates back atleast half a century when Hans Peter Luhndefined it in 1958 as:

“... business is a collection of activities carriedon for whatever purpose, be it science,technology, commerce, industry, law,government, defense, et cetera. Thecommunication facility serving the conduct ofa business (in the broad sense) may be referredto as an intelligence system. The notion ofintelligence is also defined here, in a moregeneral sense, as "the ability to apprehendthe interrelationships of presented facts insuch a way as to guide action towards a desiredgoal.” (Luhn, 1958)

Broadly speaking, BI describes a framework ofpolicies, concepts and methods (Power, 2007)

realised by a system of tools for gathering,storing, analysing and providing access to datato help organisations make better businessdecisions faster. BI systems today comprise aset of tools that can typically perform reporting,budgeting, statistical analysis, data mining,forecasting and visualisation. These toolsrevolve around a data warehouse to facilitateinformed and cross-business decision makingin Supply Chain Management, Business ProcessManagement, Customer RelationshipManagement, Enterprise Resource Planningand other related business processesthroughout an organisation. These tools extractfacts and metrics from raw transactional dataas well as processes and integrate them intoa consistent information schema. This enablesusers to analyse information collaboratively,gaining insightful situation awareness to makeinformed decisions.

THE BUSINESS

With these notions in mind, the project teamenvisaged that BI could be harnessed to bringabout significant improvements in productivityand effectiveness in Government Procurementthrough:

• Intelligent Procurement – BI can be used toextract market intelligence on suppliers,products and procurement activities,enabling buyers to make intelligent decisions

Business Intelligence in Government Procurement

109

Government Procuring Entities (GPE) up theprocurement value chain from a transaction-centric model to one that is proactive,collaborative and efficient.

DEVELOPMENT OF BIIN GEBIZ

The development of BI in GeBIZ started in early2006 when the Ministry of Finance (MOF) feltthat GeBIZ had matured and stabilised sinceits inception. GeBIZ BI initiatives can be broadlydivided into two areas. The first area entailsthe development of GeBIZ InSIGHT which is atool developed internally that leveragesArtificial Intelligence (AI) to deliver a set of BItools, which enables individual procurementusers to research historical buys to gain marketinsights for Intelligent Procurement. The secondportion covers the development of GeBIZManagement Console (GMC). GMC enablesmacro-level portfolio management andperformance management in the public sector.

throughout the procurement process. Withsignificant market intelligence at theirfingertips, buyers will be able to developbetter sourcing strategies.

• Portfolio Management – BI can also provideoverall visibility of spending patterns,facil itate demand forecasting andaggregation as well as assist in procurementplanning.

• Performance Management – Lastly, BI canhelp monitor procurement service level,policy compliance, supplier performance andmeasure procurement efficiency.

Conceptually, Intelligent Procurement involvesthe use of BI at the operational level to facilitateindividual decisions on specific procurementactivities. On the other hand, the applicationof BI in Portfolio Management andPerformance Management addressesGovernment Procurement at the managementlevel. With the overall visibility of governmentand organisation-wide spending, BI can move

Figure 1. Overview of Business Intelligence components within GeBIZ

110

GEBIZ INSIGHT

GeBIZ InSIGHT was made available to all publicsector agencies in December 2006 to allowprocurement officers to mine all of the thenunavailable transactional data stored withinGeBIZ portal. Designed to be pervasive andavailable to all, it is embedded into theprocurement officers’ workflow. By providingthe best suggestions and alternatives ifrequired, it helps the officer make intelligentand cost-effective decisions in their daily work.This tool, through the provision of advice asand when needed, could turn an averagepurchasing officer into a highly informed andefficient buyer. For example, GeBIZ Portal cannow suggest to the purchasing officer how toclassify the procurement category of histransactions and even recommend supplierssuitable for his current procurement objectivebased on their history.

GeBIZ InSIGHT allows users to:

• Research past tenders, quotations and periodcontracts of similar purchases across theentire public sector to determine prices

• Extract purchase references, documents,specifications and requirements of similarpurchases. This information will help buyersplan and structure future purchases andsource for reliable suppliers.

• Appraise suppliers by retrieving supplierfinancial information and transactionalhistory

• Evaluate current tenders and quotations byguiding buyers through the tender/quotationevaluation process in a step-by-step manner.Information on similar historical purchasesand suppliers are also extracted for buyersto refer to.

• Recommend appropriate procurementcategories and suggest possible suppliers toinvite during the tender notification process

With the success of GeBIZ InSIGHT, the projectteam embarked on the next phase of BIdevelopment in June 2007, named GeBIZManagement Console.

GEBIZ Management Console

GMC involves a platform that integrates a setof BI methodologies such as Online AnalyticalProcessing (OLAP), Data Visualisation and DataWarehousing to provide decision makers invarious government agencies with intuitiveaccess to effective Portfolio Management andPerformance Management.

Some key capabilities of GMC include:

• Providing overall spend visibility ofgovernment procurement

• Demanding forecasting and identificationof potential areas for Demand Aggregation(DA) across the public sector

• Facilitating the establishment of customisedprocurement Key Performance Indicators(KPI) for individual agencies and the entiregovernment

111

IMPLEMENTATION CHALLENGES

The business of government procurementdiffers significantly from that of commercialentities. For example, commercial entities aresegmented by their potential value for targetedmarketing campaigns. Preferential treatmentof different customer segments is a norm incommercial practice. Government procurementis different by nature: it has to operate basedon the principles of openness, fairness andtransparency.

Most BI solutions in the market today are builtwith commercial applications in mind, andtheir fundamental principles are different.There is a lack of expertise and off-the-shelfsolutions for government procurement. Theproject team has to do its own research anddevelop the BI solution for governmentprocurement from first principles. The majorchallenges encountered are discussed here.

Classification

GeBIZ Portal was originally built to facilitatetransactional activities such as Tenders andQuotations. These transactions were notclassified into detailed categories for analysis.The result is a database of millions oftransactions but without the means todetermine what was bought or sold. Themitigating approach was to partition the databy taxonomies such as IT Equipment,Transportation as well as Building andConstruction. These categories were derivedfrom the Procurement Supply Heads from MOFand the Building and Construction Authorityin consultation with DSTA as well as otherGPEs.

An off-the-shelf and in-database AI enginebased on the Support Vector Machinesalgorithm was customised and deployed toaddress this classification problem. With the

help from cataloguing specialists from theSingapore Armed Forces Cataloguing Authority,25,000 record samples in GeBIZ were extractedand manually categorised. These sets ofcategorised records called ‘training sets’ wereused to train the initial AI engine to categoriseall historical transactions in GeBIZ. By observingthousands of public officials conduct their dailyprocurement tasks, the engine automaticallylearnt and corrected its previous mistakes andalso discovered new categories as they arose.

However, a classification engine is ineffectiveif it cannot understand the unstructuredinformation within each procurementtransaction. For example, the system neededthe capability to know that printing paper andwatermarked paper, specified by differentprocurement officers, are both paper purchases.It also had to be able to distinguish thedifference between the word ‘plane’ in e.g. a‘woodwork plane’ and an ‘aero-plane’. Toaddress this contextual and disambiguationissue, the project team applied NaturalLanguage Processing methods such asStemming and Concept Extraction to the rawdata as the AI engine was executed.

Performance

Machine classification and text analytics requiresignificant processing time. For a webapplication, system performance is a majorchallenge in the development of GeBIZ's BIplatform. The team, after usability studies,decided on a maximum response time of 12seconds per multi-year analysis and 100ms foranalysis embedded in the workflow. To meetthis performance requirement, classificationand text analysis are performed natively withinthe database. The database and hardwarearchitecture also has been significantly tuned

to help overcome the performance problem.

At present, tens of millions of items are now

conceptually analysed across all of GeBIZ's

transactions and results returned daily without

significant waiting times.

112

Data Quality

GeBIZ portal is a system of many sub-systems

built in phases over a period of a few years.

As a transactional system, each component

performs its work autonomously and these

sub-systems contain silos of data representing

islands of business and data rules, making it

difficult to explicitly describe information across

multiple business areas.

System changes due to policy and capability

additions impose increased demands on the

information schema. The information schema

evolves over time and sometimes results in

data redefinition and inconsistencies.

Furthermore, government procurement is not

an isolated process. GeBIZ receives data

constantly from various external systems for

processing. The data quality from these external

systems is beyond GeBIZ’s control. This

combination of factors makes the reporting

of historical data difficult.

Information Accessibility

Businesses change and evolve over time, often

rapidly. To allow key decision makers insights

into the market, they must be able to access

relevant information immediately as and when

they need it. This requirement called for a

need to rapidly deploy cross-business reports

or answer ad hoc queries. Before GMC was

available to end users, ad hoc reports were

generated manually, and that could take days

or even weeks.

GEBIZ BI CAPABILITIES

Today, there exists a data warehouse where

all aspects of GeBIZ transactions reside. Before

entering the warehouse, data is cleansed of

discrepancies and conformed into a singleconsolidated business schema. Data Martsrepresenting business areas such as tenders orpurchase orders provide reports and analyses.These Data Marts feed off the warehouse and

deliver a single version of the truth. The qualityof data is monitored constantly and isolatedfor further processing if found to bequestionable. To accomplish this, all businessand data rules need to be explicitly identified,documented, encoded and stored in ametadata repository. With a metadatarepository, the derivation of any field can beexplicitly specified and rationalised.

Information Retrieval

Simple yet effective information retrieval andaggregation is necessary in GeBIZ to facilitateefficient and accountable decision making. AllBI components are built with this premise inmind. Through a simple query field, differenthierarchies of procurement information canbe retrieved and aggregated from concepts tospecific purchases across all agencies. Securityis applied transparently depending on the user.The user's affiliated organisations and existingrights automatically determine the level ofdata he or she may view.

Demand Aggregation

DA is a means of facilitating efficientprocurement. By consolidating frequentpurchases into a contractual agreement, thegovernment can exploit economies of scale toobtain favourable prices and reduce thetransactional overhead of subsequentacquisitions of the same item by performing

113

it upfront. However, demand has to be analysedbefore it can be aggregated. Yet as previouslymentioned, transactions of governmentprocurement are unstructured. It is difficult toclassify millions of transactions into varioustaxonomies, and even harder to identifyparticular items in demand. The text analyticand classification approach described in thisarticle was deployed to allow users to derivetransaction trends at any hierarchy andresolution. In fact, any ad hoc queries that canbe performed using the OLAP concept such asdrilling, pivoting, dicing and aggregating canbe applied to the unstructured content foundin the databases of GeBIZ.

Enforcement

Even with effort spent on DA contracts,government users may not realise its existenceand continue to make ad hoc purchases. Thisresults in contract leakages. The BI platformcan facilitate contract leakage detection by:

• Allowing procurement experts to quicklyreview all kinds of ad hoc purchases

• Check for patterns of contract leakages suchas the referencing of a period contract

• Performing brute force searching of alltransactions across contracts

Multi-faceted Analysis

BI can offer many unique and differentperspectives to its users. The BI solution ofGeBIZ provides a multi-level analysis in theform of scoreboards, dashboards and alerts.Through scoreboards which typically measureKPIs, top-level management can view theprocurement performance of their organisationat a glance. These KPIs are colour-coded todepict the various aspects of procurementhealth they measure.

In addition to scoreboards, dashboards are alsoavailable to show aspects such as currenttransaction status and period contract

utilisation to aid operational users. Users canmonitor their dashboards and react to eventsaccordingly. Even with detailed reporting,information may not be timely enough forspecific operational requirements. To that end,an alerting ability integrated with every BIcapability was deployed. This alerting capabilitycan automatically email, page, broadcast orpush information to its users when a conditione.g. a purchase is met.

Automation

Lastly, a recurring theme of many BI missionstatements is that ‘the right Information shouldbe pushed at the right time to the right person’.In GeBIZ's BI solution, this capability is called‘GeBIZ Delivers’ and it is a scheduling engine.Reports can be sent at pre-determined intervalsto dynamically determined individuals basedon various customisable criteria. Furthermore,the system is able to trigger procurementworkflow processes and execute user-definedscripts to perform tasks on behalf of its users.

BI MATURITY

In 2007, on the 50th anniversary of Hans PeterLuhn's definition, The Data WarehousingInstitute (TDWI, 2007) published its annualBenchmark Report. With a scale analogous tohuman maturity (Infant, Child, Teenager, Adult,Sage), it is reported that "two-thirds oforganisations are Teenagers of the TDWI BIMaturity Model". Of the remainder, 20% werein the Child stage, 15% in the Adult stage, andSages were almost non-existent at 0.6%.Furthermore, the report also stated that "thereis a correlation between the age of a BIprogramme and its maturity. More than two-thirds (68%) of organisations that score in theChild stage have BI programmes that are lessthan two and a half years old, whereas 78%of Adult-stage BI programmes have existed formore than five years. Just as in humandevelopment, maturity comes with age". Theprocess of BI development never ends. BI mustreact in cadence with its host business and thenature of any business always changes. To that

REFERENCES

114

Eckerson, W. (2007). 2007 The DataWarehousing Institute BI Benchmark report.

Luhn H. P. (October 1958). A BusinessIntelligence System (PDF). IBM Journal.Ret r ieved on 10 Ju l y 2008 f romhttp://www.research.ibm.com/journal/rd/024/ibmrd0204H.pdf

Power D. J. A Brief History of Decision SupportSystems, version 4.0. DSSResources.COM.Ret r ieved on 10 Ju l y 2008 f romhttp://dssresources.com/history/dsshistory.html

end, continuous reactive and predictivedevelopment efforts are required for any BIproject team.

Acknowledging the immense challenges anduncertainty involved in the development of aBI solution, the project team adopted thephilosophy that a BI solution is not a one-offproject. When harnessed correctly, a BI situationbecomes a cornerstone of its host system.Hence, BI must evolve together with the systemand be ready to support new decisions in aconstantly changing environment.

CONCLUSION

Incorporating BI in GeBIZ is a complex andchallenging undertaking, but the returns oninvestment are enormous. The introduction ofthis BI platform in GeBIZ will result in acapabi l i ty increase for governmentprocurement agencies and elevate Singaporegovernment procurement to a higher paradigmof intelligent sourcing and purchasing. Thiswill raise government procurement to astrategic level and bring about morecollaboration and partnership amonggovernment agencies and the industry. A BI-enabled GeBIZ will not only increase theefficiency of government agencies but alsoimprove the effectiveness of governmentprocurement, resulting in huge cost savingsfor the public sector as a whole.

Carol Lo S Chia is the Assistant Director of the Electronic Business Centre.Besides managing the development and operation of the GovernmentElectronic Business (GeBIZ) Portal, she oversees the development andimplementation of the Electronic Procurement System for the Ministry ofDefence (MINDEF) and the Singapore Armed Forces. Carol spearheaded thedevelopment of the MINDEF Internet Procurement System in 1998 and alsoimplemented a few large-scale IT systems related to Procurement andCataloguing before initiating the development and implementation of theGeBIZ portal for the public sector. Graduating with a Bachelor degree inPhysics (First Class Honours) from the National University of Singapore, shewent on to obtain a Master of Science in System Analysis and Design fromthe London School of Economics and Political Science, University of Londonin 1983 under the Engineering and Scientific Professional Training Award.

BIOGRAPHY

115

Leo Chen Hong is an Engineer (Enterprise IT) and was previously part of theteam that developed and operates GeBIZ InSIGHT and GeBIZ ManagementConsole. He had been involved in the back-end development and front-lineoperations of many aspects of business intelligence for governmentprocurement. These included requirement analysis, data modelling andprofiling, Extraction Transformation Loading development, dashboard andscorecard design as well as end-user application development and support.Chen Hong holds a Bachelor degree in Linguistics and Computer Science fromthe University of Michigan.

Mine Clearing Technology

Introduction to

ABSTRACT

This paper presents the technologies and methods developed

for mine clearing operations currently used by the military and

humanitarian demining organisations. In any mine clearing

operation, the operating environment and the type of threats

are never the same. Thus, a single method or type of equipment

rarely constitutes the most successful means of resolving the

problem in terms of time, cost and effectiveness; a combination

of tools is more commonly employed to ensure a successful

mine clearing mission. This paper aims to give an introduction

to and appreciation of the key mine clearing methods and

equipment, and the key differences and considerations for

military and humanitarian operations. The common methods

of demining such as manual demining, explosive mine breaching

and mechanical demining will be discussed. The design

considerations for mine flails on mine clearing vehicles will

also be presented.

Tan Chun

Gary Wong Hock Lye

Bryan Soh Chee Weng

118

INTRODUCTION

History of Mines

Mines, derived from the Latin word ‘Mina’meaning ‘vein of ore’ was originally used todescribe the digging of minerals from theearth. Over time, it has become a term usedby military engineers to denote the explosivesthey lay in the ground during battles.

Mines were first used in third B.C. in the formof non-explosive metal spikes laid in the earthas a form of defensive tactics. In the 14thcentury, the world ushered in the arrival ofexplosive mines which are commonly foundeven in modern times. The basic design of themines was usually simple explosive matterburied with debris or metallic pieces that wereto be used as fragments during an explosion.

Mines became a regular feature of warfareonly in the 19th century. It was in 1918 thatmines first became used on a large scale as aneffective means against the assault tanksintroduced then.

Since then, mines have become a new threatin war, which also meant that a solution hadto be developed against them. Thus, thissparked the development of various mineclearing techniques.

NEED FOR MINECLEARING

By World War Two (WWII), mines had evolvedto perform two major functions: to kill andmaim personnel, or to destroy and disabletanks and vehicles. The US Army recorded that2.5% of fatalities and approximately 21% oftank losses of WWII were mine-related.

In addition to WWII mines, an estimated 400million mines have been laid in various conflictareas since then.

Despite the initial development of mineclearing concepts as a form of countermeasureagainst mines during wartime, the real needfor mine clearing usually begins after the endof hostilities. This is attributed to the verynature of why mines were laid in the first place– to deter access to and use of land. Mines laidduring conflicts are rarely removed at the endof the conflicts due to the lack of proper minemaps, markings, loss of such maps and markingsor changes in mine location due to soil shifts.Unlike soldiers, mines do not recogniseceasefires and treaties or differentiate betweenfriends, foes or civilians. Mines continue to dotheir job even years after the end of a conflict.

To this day, there are up to 65 countries stillafflicted by mine threats from past conflicts,with the worst regions being Angola, Namibia,Mozambique, Somalia, Ethiopia, Eritea, Sudan,Croatia, Iraq, Afghanistan, Russia, Cambodiaand Vietnam. On average, there areapproximately 70 injuries and death from minesevery day. It is this threat posed by mines thatplaces extreme risks on the civilian populationwho return to the land after a conflict.

The problems caused by active minefields donot end here. There are also socioeconomiceffects as fertile lands are not available foragriculture, roads are not accessible and landcannot be cleared for industries. Theseeventually lead to continued poverty andfamine in the affected areas. Thus, despite thehigh cost of humanitarian demining efforts,Mine Action Co-ordination Centres supportedeither by the United Nations (UN) or hostcountries have been set up in differentcountries to drive the demining efforts.

TYPES OF MINECLEARING METHODSAND EQUIPMENT

The estimated cost of clearing a mined areaof one square kilometre is about US$1 million,whereas it takes only a few dollars to acquire

Mine Clearing Technology

Introduction to

119

a mine. Thus, the search has always beenongoing to find a more effective solution formine clearing. In this section, the types of mineclearing methods and equipment will bediscussed.

Manual Demining

Manual demining is the most commonly usedmeans in humanitarian demining. It is also theonly recognised method of demining that isable to achieve the 99.6% clearance raterequired by UN standards. Its main function isto detect and locate mine locations fordeactivation or destruction in the later stages.It consists of the use of manual proddingmethods, use of detectors such as metallicdetectors and even animals.

As the name suggests, this is also the mostlabour-intensive and costly method whichrequires a large amount of human effort andspecialised training. In the case of manualprodding and metal detection, large numbersof people from the local population areemployed to work through a selected piece ofland with personnel protection kits and selectedtools (see Figure 1). Although the method iseffective, it is one of the slowest ways ofclearing a piece of land. The progress ofdemining is either hampered by high falsealarm rates due to instances of metallic objectsin the ground or by deeply placed metal mineswhich are hard to detect. Recent developments

include the introduction of Ground PenetratingRadars. Although more effective, thetechnology has its fair share of drawbacks interms of high false alarm rates.

Manual demining also creates a socioeconomicbenefit for the population of the area as itcreates a large number of jobs for people whoare usually living below the poverty line. Thesejobs eventually enable the local population toslowly rebuild their lives.

In the case of mine detection using animalssuch as dogs and rats (see Figure 2), the animalsare carefully trained to detect the explosive

Figure 1. Manual deminer with body armourusing prodding tool

Figure 2. How a mine dog is deployed to detect mines

120

substances from the mines (RDX1 or TNT) andto react in a prescribed manner when theypinpoint the location of the mine. Theeffectiveness of using animals is limited by thetraining they undergo and the duration theycan be deployed in the field. Without sufficienttraining, the effectiveness of the method canbe greatly reduced. Effectiveness is also reducedif the animal is deployed in the field for toolong, as fatigue may also affect its ability todetect.

Use of Explosive Charges inDemining

In essence, the use of explosive charges indemining is usually limited to the destructionof the mines after they have been detectedthrough more traditional means such as manualdemining.

But in the arena of military operations, itprovides a fast and effective way to create apath for vehicles or men through a minefield.This is often referred to as ‘breaching’.

The main underlying concept of this methodis the deliberate triggering or destruction ofmines within the section route or area throughthe introduction of massive blast waves createdby an explosion. The same blast wave will alsoact to trigger any Improvised Explosive Devices(IED) within the selected area.

Although the same concept is employed,different types of equipment have beendeveloped to create this capability. Whatdifferentiates them from one another is usuallythe breadth and length of the cleared paththat is created. Some more common forms ofexplosive charges used are bangaloretorpedoes, man-carried or vehicle-mountedline charges (see Figure 3). The more modernapproach would be the use of vehicle-launchedfuel-air explosive bombs.

Mechanical Demining

Mechanical demining methods deployed sinceWorld War One (WWI) includes the use of tillersystems, mine rollers, mechanical excavation,mineploughs and mine flails. Although themain concepts have not changed much overthe years, the technology behind them has.Recent developments on mechanical demininginclude the use of remote control systems whichallows remote operations of the equipmentfrom as far as 5km away. Additional armourto protect the operator cabin, improved cabindesigns to improve crew survivability andsystem designs to overcome the traditionallimitations of mechanical demining methodshave also been developed.

Although developed initially for militaryoperations, some mechanical deminingmethods have been used in humanitariandemining operations. The advantage of usingmechanica l demining methods forhumanitarian purposes is the ability to clearlarge amounts of land quickly. However, thereare doubts on its effectiveness and ability toachieve the 99.6%2 clearance rate required.Thus, a secondary method is usually employedto validate the cleared land. The followingsection will briefly describe the functions anddesign of each mechanical demining method.

Figure 3. A Cobra Projection Line charge usedby combat engineers

121

Tiller Systems

Tiller systems (see Figure 4) are generallypurpose-built systems designed for destroyingmines or Unexploded Ordnance (UXO). Due tothe nature of their operations, they are usuallyheavy and large in size. Substantial power isrequired to operate the heavy tiller drum aswell as the prime mover, which is most likelya tracked vehicle. The tiller system employsthe use of rotation drums fitted withoverlapped teeth or bits that would grind upthe ground in accordance with the depth they

are required to dig. Mines or UXOs wouldeither be destroyed or activated through directcontact with the teeth of the rotating drums.

Key issues with the tiller system are thepossibility of burying the mines deeper belowthe tiller due to the downward force of thedrum and the collection of mines in the ‘bowwave’ (see Figure 5). The bow wave is thecollection of loose soil in front of the drumcaused by the forward movement of the drum.The tiller system has also been reported tohave lower effectiveness in soft and rock-stewed soil.

Anti-Mine Rollers

One of the older mechanical deminingmethods, the anti-mine roller was first deployedin 1914 on the British tanks. It has been adaptedfor humanitarian purposes over the years.

Although a typical anti-mine roller would bemounted on a modified tractor or wheel loaderand made to run over selected areas todeliberately activate mines that are live andin serviceable conditions, most anti-mine rollersused in humanitarian demining are not capableof surviving an anti-tank mine blast. As such,careful surveying of the ground is necessaryprior to their application. A key issue with themine roller is its inability to deal withunserviceable and serviceable mines that mightbe buried too deep.

Figure 4. The tiller drums of the Rhino system

Figure 5. An example of a mine “surfing”on the bow wave

Figure 6. M1AI equipped with mine roller

122

Mechanical Excavation

Mechanical demining employs the use ofvarious machines to enable the removal of soilfrom suspected areas for inspection of minesand UXOs. Soil is sent for processing via sifting,manual inspection or crushing to identifyunexploded ordnance, which in turn will besent for disposal by the Explosive OrdnanceDisposal (EOD). This method enables soil to becleared to depths that may not be reachableby tillers, flails and mine rollers.

Equipment used in mechanical excavation istypically construction machines installed withmine blast protection and additional armourin the cabin to protect the operator. Suchequipment includes tractors, front-end loaders,bulldozers, excavators and soil sifters. In orderto further protect the equipment, larger UXOor anti-tank mines are identified by metaldetectors for separate disposal.

To date, there is no known critical drawbackof the method as long as appropriate safetymeasures are taken.

Mineploughs

Similar to anti-mine rollers, mineploughsoriginated from military applications. One ofthe earliest applications of mineploughs wasin WWII where they were mounted on Churchilland Sherman tanks as part of the assault onNormandy. The mineplough was used as a‘softer’ approach to mine clearing as unlikethe flail, it does not create large columns ofdust and craters from destroyed mines. Themineplough works by moving a speciallydesigned shovel just below the surface of thesoil in order to plough up mines and push themto the sides of the path created.

The concept of employing the mineplough infront of an armoured vehicle has not changedover the years. As it is still a quick and effectivesolution for creating a vehicular path across adeep minefield, many variants of modern-daymain battle tanks (MBT) are still fitted withthe mineplough system.

Today, the depth of clearance can be variedthrough hydraulic controls and vehicles can beoperated remotely increasing the safety levelfor the operator. One of the most widely usedmineplough systems today is the Pearson FullWidth Mineplough system which can be easilyadapted to various platforms without extensivemodifications.

Figure 7. The Hydrema M1100 excavator withan armoured cabin which can be attached

with different tools for mechanical demining

Figure 8. A mechanical sifter which can beattached to a commercial mechanical

demining machine

Figure 9. A Bullshorn mineplough mounted ona WWII Sherman Tank

123

The key issue with the mineplough is thepossible collection of mines that are pushedto the sides of the path. As the mines are rarelyactivated by the mineplough, secondary meansof ordnance disposal must still be in place andthus, the mineplough has not been adaptedfor humanitarian uses.

Mine Flails

Mine flails were also first developed for militaryapplications and were famously used in WWIIwhere they were fitted on Sherman tanks. TheSherman tank flails are known to be veryeffective and played an integral role in theinitial shore invasion in Normandy.

With the growth of humanitarian deminingin the 1990s, flail systems were developed tomeet the demand for tools against mines andsmall UXOs. In the Singapore Armed Forces(SAF), the Hydrema MCV 910 Series 2 has beenin service since 2005. The SAF, DSTA andSingapore Technologies Kinetics have jointly

developed and built a Tracked Mine Flail whichoffers more protection to the crew andexceptional mobility performance over othercommercial off-the-shelf mine clearing vehicles.The Tracked Mine Flail features a fullyautomatic deployable flail system and theworld’s first hydrostatic tank transmission forboth creep and normal mobility.

Conceptually, most flail systems utilise the sameconcept: a drive system, an extended arm, arotating drum and chains with weights at theends to provide the pounding force. The forceof the swinging hammer is transferred ontothe ground upon impact. It is this same impactforce that either causes the mine to detonateor shatter.

Each flail system is different in terms of clearingspeed, clearing depth, survivability and clearingeffectiveness. The next section will cover anin-depth discussion of the design considerationsfor mine flail systems.

Figure 10. The mine clearing vehicle (Hydrema MCV 910 Series 2)Source: Army news (April/May 2006)

Figure 11. Tracked Mine Flail System

124

DESIGN CONSIDERATIONSFOR MINE FLAIL SYSTEMS

The key components in the design of mine flailsystems are as seen below:

Blast Protection

Blast Shield

Modern mine flails operate with the flail infront of a blast plate system. This blast platedeflects the blast wave and also acts as aprotective layer between the mine explosionand the operator. In some cases, the blast shieldcovers the flail circumference to also preventthrow-out of mines that are not destroyed bythe initial impact of the flails.

Blast shield strength and thickness are mainlydetermined by the threat the flail is designedto meet and the blast shield’s angle ofinclination. Finite Element Method studies aremade in accordance with the expected threati.e. amount of explosives in the mines andquantity of mines to defeat before inspectionand repair to determine if blast shield designscan cope with the blast pressure andfragmentation damage. Actual blast trials areconducted to verify and correlate the stresseson the blast shields with the theoretical results.

Protection of the Cabin AgainstAnti-Tank Mine Fragments

Although the blast effect of a mine is supposedto be contained by the blast shield, there isalways a risk of the vehicle going over a missedmine. In order to protect the operator frompossible fragmentations in a mine explosion,the cabin is required to be suitably armouredfor crew protection.

Protection against Loss of StructuralIntegrity and Lift

The impact of the blast overpressure affectsthe platform in two different situations. Oneis during the detonation of mines by the flailand the other during the detonation of missedmines under either one of the wheels or tracks.

Figure 12. A Finite Element Method study made on ablast load case on a blast shield design

Figure 13. Fragmentation damage from ananti-tank mine with 4.45kg TNT

Figure 14. The blast shield from the Hydrema910 Series 2 – the flexible flap at the top of theshield deflects the blast wave over the cab without

sustaining damage from the blast wave.In addition, the tilted angle of the flap helps

to recirculate dust in blast shield duringflailing operations.

8.34e+02

7.54e+02

6.707e+02

5.869e+02

5.03e+02

4.192e+02

3.354e+02

2.515e+02

1.677e+02

8.384e+01

0.000e+00

Time = 0Contours of effective stress (v-m)max ipt valuemin = 0.0534984#75901max = 3290.35 at elem#68814

Fringe Levels

125

It is extremely important to ensure thestructural integrity of a vehicle against mineattacks. When a mine detonates under thevehicle and should the vehicle belly rupture,the blast energy from the mine explosion wouldcause an overpressure in the cabin. The suddensurge of pressure would injure the unprotectedear and other gas-filled organs e.g. lungs ofthe operator. The physical limits foroverpressure are 200Pa for the ears and 19KPafor the lungs and intestines.

Therefore, to protect the vehicle from mineattacks, forces from the mine blast energyshould either be deflected or absorbed toprotect the vehicle belly from being rupturedespecially at the weld joints. To deflect theforces, a V-shaped hull design can be used suchthat impact forces are deflected off the sides

of the undercarriage of the hull. To absorb theforces, a deflection budget or a sacrificial blastshield can be designed into the hull such thatthe forces are intentionally absorbed withoutcausing damage to the main hull. In thesecases, the deflection budget and the blastshield are usually designed such that they canbe easily replaced.

The usage of V-shaped hull design, deflectionbudget and blast shields are also designfeatures that prevent ‘lift’ i.e. overpressureforces that act on the impact point causing thevehicle to be lifted off the ground. The suddenlifting of the vehicle can lead to acceleration-related injuries on the foot, ankle and spinalcord of the operators. In severe cases, it mayflip the entire vehicle.

Figure 15. An example of adeflection budget

Figure 16. How lift affects a vehicle

Figure 17. The limits of acceleration exposure prior to injury

126

Flail System

In our technical evaluation of mine clearingvehicles and the Design Test and Evaluation ofthe Tracked Mine Flail, we conclude that thereare basically four modes by which a mine flailcan neutralise the mines. This was derived fromextensive testing of the mine clearing vehiclesin various terrain and conditions.

In the simplest mode, the chains and strikers(or hammers and weights) of the flail strikethe activating plate of the mine and detonatethe mine. Mine flails have a typical strike forceof more than 1,000kg and it is more thansufficient to set off the mines which typicallyrequire 150kg to 200kg. However, when a minedetonates under the flail, there may be damageto the flail system and immediate repairs arerequired before further operations.

In the second mode, given the strong strikeforce and intensity of newer flail systems, themines can be smashed into pieces. This is thedesirable mode of defeating the mines as theleast damage will be inflicted on the flailsystem. However, the debris which consists ofexplosives and detonators can still be a threatto unprotected personnel.

In the third mode, the mines are partiallydamaged but not smashed in the less drasticcases. The chances of damaging the firingmechanism when the mines are partiallydamaged are very high. Such mines are alsoconsidered neutralised as they cannot deliverthe full impact of the charges.

Lastly, the flailing is essentially a diggingoperation and mines may get thrown up duringflailing. However, only mines thrown off thepath of the flailing are considered neutralised.

The modes of neutralising the mines are alsodependent on the types of mines. For singleimpulse actuated mines which were commonlyused during WWII, the flail is capable ofdetonating them. For newer anti-shock andanti-flail resistant mines, the other three modes

i.e. destruction, neutralisation and throwingoff are more common.

There have been limited studies made on therelation between flail effectiveness and groundconditions. Thus, when an area is set to becleared, it is compulsory for flail operators toperform a test flail to determine the settingsof the system. Through comparison of testresults from local trials, Swedish EOD &Demining Center and Croatia Mine ActionCenter reports, DSTA engineers derived thatthe effectiveness of the flail is a function ofthe power pack output, flail width, groundcondition and creep speed i.e. speed of vehiclemovement during flailing. The powerpackneeds to be suitably sized in accordance withthe flail width required. Generally, theacceptable power output to flail width ratiois between 50HP to 60HP per metre of flailwidth.

In a given test environment, where the groundconditions are unknown, a test flail must beperformed to obtain the creep speed foroptimal flail performance. The depth of theflailing is measured to ensure that there issufficient dig depth to defeat the mines. Inour local trials, extensive testing using the MCVto determine the optimum flailing performanceon different types of terrain was conducted.With this relationship between flailperformance and terrain established from ourtrials, we can calculate and estimate theperformance of other mine clearing equipmentwithout local testing and this saves considerabletime and cost during the evaluation of thisequipment.

Other Enhancements

Additional enhancement systems are availableto improve the operating capability of the flailsystem. They are as follows:

• Lane marking system allows marking out offlail path to demarcate safe zones to travelin. This is more commonly used as a militaryapplication in breaching operations.

127

• Slow clearing speed – For militaryapplication, the slow rate of clearance viamost mechan ica l means (exceptmineploughing) means that the system isoften exposed to enemy fire.

• Intelligent mines – In military applications,a mine field may not just be filled byconventional mines that are detonated viapressure or magnetic means. There areintelligent mines such as off-route mineswhich use thermal, seismic, or magneticsensors or a pressure sensitive wire laid acrossthe road to detect vehicle movement. Oncea vehicle is detected, a rocket or otherammunition hidden alongside the road isfired at the vehicles. The more advancedversions can not only differentiate betweena tank or other class of armored vehicle, butcan even decipher friend from foe as wellas between models such as the T-80 andT-72 MBTs. With ranges up to 100m,mechanical demining tools would have littleor no protection against them.

• Remote control system allows the remotecontrol of the entire system from as far asa five-kilometre line of sight. This enablesthe operator to operate in high-risksituations without endangering his life.

• GPS position logging systems log the pathcleared by the flail system. The same datacan be transferred to a digital map forconfirmation of the areas that have beencleared.

Challenges for Mechanical MineClearing Systems

Although a large number of mechanical mineclearing systems has been developed toimprove the process of demining, manyavailable systems are still unable to cope withthe following challenges:

• Steep, undulating terrain – As mechanicaldemining requires ground contact to beeffective, the presence of undulating terraincan cause ‘skip zones’ where eitherinsufficient pressure or impact was appliedonto the ground. These skip zones remainas potential mined spots within a clearedarea unless a secondary means is used tovalidate the land.

• Defective mines/mines with delaycharges – The risk comes when the minedoes not detonate during the initial contactwith the mechanical demining system. Thedetonation may only be activated secondslater or when the wheel/track pressure isexerted on the mines. This places theoperator in danger as the blast detonationmay only come after the passing of the blastshield.

• Shaped charge mines – Unlike normal mineswhich cause damage through the transferof overpressure forces and fragments, shapedcharges mines have the capability of piercingthrough the majority of the armourprotection on mine clearing platforms.

Figure 18. A German Parm 1 off-route mine

REFERENCES

128

Axelsson H. 8 January 2003. Mine ClearingVehicles Crew Safety Standard, SwedishDefence Material Administration. Retrievedfrom http://www.itep.ws/

Bishop, C. The Encyclopaedia of Weapons ofWorld War 2, pp 54-59.

Dutch Ministry of Foreign Affairs. MineClearance – Policy Framework for humanitarianmine action. Retrieved on 11 July 2008 fromhttp://www.minbuza.nl/en/developmentcooperation/Themes/HumanitarianAid,mine-clearing

Geneva International Centre for HumanitarianDemining. Which areas are affected bylandmines? Retrieved on 11 July 2008 fromhttp://www.gichd.org/mine-action-and-erw-facts/faq/countries-affected/

Geneva International Centre for HumanitarianDemining, GICHD. (May 2004). A Study ofMechanical Application in Demining, pp 9-41.

CONCLUSION

This paper has discussed the different toolsand concepts employed for demining. Despitethe wide variety of equipment developedagainst mine threats, there is no single methodor tool that is truly a perfect match for allsituations. Careful surveying and situationalassessment are still required to determine themost effective solution for each situation.Further improvement of existing tools mightbe a way forward but the cost of doing so maysimply be beyond the means of those whoneed them. The collaboration of two or moreindependent tools, each operating using itsdistinct detection and clearing method, mightbe a better and cost-effective solution toachieve a more thorough mine clearingperformance in the long run.

Hameed, A. February 2008. Attack of Armourand IEDs. Singapore FVD course presentation,pp 21-34.

History of landmines. (1998-2006). InternationalCampaign to Ban Landmines. Retrieved on 11July 2008 from http:/ /www.icbl.org/problem/history

Paul Hubbard & Joseph Wehland. July 1997.Goliath: Landmines, the Invisible Goliath.Ret r ieved on 31 Ju l y 2008 f romhttp://library.thinkquest.org/11051/history.htm

Schneck, W.C. (July 1998) Origins of militarymines: Part 1, pp 1-7. Retrieved on 11 July 2008f r o m h t t p : / / w w w. f a s . o r g / m a n / d o d -101/sys/land/docs/980700-schneck.htm

Theimer, T. 15 January 2001. MCV MechanicalMine Clearing Report Land Clearance Testfacility WTD 51 Identification Number 235014390. Retrieved from http://www.itep.ws/

1 RDX (Cyclotrimethylenetrinitramine) and TNT(Cyclotrimethylenetrinitramine) are explosivesubstances widely used in military and industrialapplications.

2 All UN-sponsored clearance programmesrequire the contractor to achieve the agreedstandard of mine clearance of 99.6% to a depthof 200mm.

ENDNOTES

Tan Chun is a Principal Engineer and Programme Manager (Land Systems).He has many years of experience in managing projects in the areas of militarybridges, mine clearing vehicles, engineering plants and light sea vessels. Heobtained his Bachelor degree in Mechanical Engineering from the NationalUniversity of Singapore (NUS) in 1993. In 2001, he obtained a Master ofScience in Military Vehicle Technology from the Royal Military College ofScience, United Kingdom.

BIOGRAPHY

Gary Wong Hock Lye is a Senior Engineer (Land Systems). He is currentlyinvolved in the acquisition management of combat engineer systems for theArmy. Gary is also involved in the development of mine clearing vehicles. Hegraduated with a Bachelor degree in Mechanical Engineering in1999 and obtained a Master of Engineering in 2002 from NUS, with the areaof research in characterisation and synthesis of functional materials.

129

Bryan Soh Chee Weng is an Engineer (Land Systems). Currently he is involvedin the acquisition management of combat engineer systems for the Army.He graduated with a Bachelor degree in Mechanical Engineering in 2004from NUS under the Singapore Armed Forces Local Study Award. Between2004 to 2007, he served as a Military Engineering Officer in the army as aregular officer.

Future Energy and Power Challenges

ABSTRACT

This paper highlights the range of energy and power challenges

faced by the Ministry of Defence and the Singapore Armed

Forces. Buildings, infrastructure and the whole range of military

requirements in capabilities such as unmanned systems, soldier

systems and platforms have their unique demands and hence

require different solutions. This paper explores the use of

alternative energies, energy management modes and even

advanced composites to overcome challenges and meet the

energy and power demands in a cost-effective manner.

Gareth Tang Ee Ho

LTC Yeo Chee Beng

Ho See Fong

MAJ Oliver Lan Chi Wai

132

THE GLOBAL BACKDROP

In 2008, the world saw record highs in oil prices(see Figure 1) and various reasons proposedincluded diminishing oil reserves, instability inthe Middle East, oil speculation and theweakening US dollar. Drilling of oil is also morecomplex and costly today as the more accessibleoil fields are being depleted and much drillingnow occurs at oil wells which are deeper oroffshore. Figure 2 gives an indication of theextent of the increase in oil production costover the years.

The rising cost of fossil fuels, coupled with theuneven global distribution of energy resourcesmake energy security an important issue tooil-importing countries. In particular, political

and social instability in the Middle Easterncountries is a cause for worry as approximately70% of global oil reserves are found in thesecountries. The dispute between Russia andUkraine over natural gas in 2005 to 2006demonstrated how external energy suppliescan be volatile, even to countries not directlyinvolved – in this case, Western Europeancountries. China, the growing giant, has alsojoined the competition for energy resourcesin the global market to propel its increasinglypower-hungry economy. Russia, Kazakhstan,Iran, Saudi Arabia and Africa are some of thesuppliers of oil and/or gas to China (Liu, 2006).In fact, in his speech on 23 August 2006,Australian Minister for Resources and Energy,A. M. Martin Ferguson, described theinternational competition for energy as thenew Cold War.

Future Energy and Power Challenges

Figure 2. Graph showingtrend of oil drilling cost(figures obtained fromAmerican Petroleum

Institute (May 2008), 2006Joint Association Survey

on Drilling Costs)

Figure 1. Graph of crude oil pricesfrom 1974 to 2008

‘98 ‘99 ‘00 ‘01 ‘02 ‘03 ‘04 ‘05 ‘06Year

Cost of drilling of crude oil wells

2000

1500

1000

500

0

Tho

usan

d do

llars

(U

S$)

per

wel

l

133

Singapore relies heavily on natural gasimported from Indonesia and Malaysia forabout 80% of our electricity consumption1. In2007, the Ministry of Trade and Industry (MTI)launched a ‘whole-of-government’ approachto deal with the increasingly complex energyissue, leading to restructuring that resulted inthe formation of an Energy Division in MTIand the expansion of the Energy MarketAuthority. At the national level, the strategyis to diversify supply channels of energyresources, adopt renewable energy as well asmaximise energy efficiency.

ENERGY AND POWERCHALLENGES

Almost all the land, air and sea platforms ofour military are dependent on oil-based liquidfuel in one form or another. Diesel fuel powersour armour brigades, heavy oil powers ournavy frigates, kerosene fuel powers our fighterplanes and even our command and controlinfrastructure runs on electrical power that isbacked up mainly by diesel-poweredgenerators.

In recent years, we have modernised as wellas brought in new fighting capabilities thatare more technologically sophisticated. Wenow have armoured fighting vehicles withbetter protection, new generation frigatesequipped with sophisticated weapon andsensor systems and twin-engined F15s. The

ongoing transformation will have a

compounding effect on our total energy

demand and thus our thirst for oil.

A key trait of the Third Generation

transformation of the Singapore Armed Forces

(SAF) is the increasing dependence on networks

and electronic systems. A common vulnerability

of electronic and network systems is their high

dependence on continuous power and cooling.

As the Third Generation SAF expands with

exponential growth in computing and

networking power, the energy and cooling

requirements will follow suit. In 2003, the

average power consumption of a legacy server

was 2.5kW per rack. With the advent of high

density blade servers, the average power

consumption will grow to 8.5kW per rack. The

introduction of high density blade servers are

expected to reduce the server footprint by up

to three times, yet our projected need for Data

Centre space and power demand has continued

to grow.

Unmanned warfare has begun to dominate

the land, air and sea domains. The unmanned

systems complement and supplement manned

systems, and in some applications, they can

replace humans and thus remove them from

harm's way. To fully exploit the potential of

the unmanned systems, new energy sources

are required to increase their endurance and

safety. This would consequently lead to

increased range and payload, hence

operational flexibility.

Even as unmanned systems are increasingly

being harnessed, soldiers continue to play

critical roles in operations. Our Third

Generation soldiers will be equipped with a

myriad of sensor and communication devices

to connect them to the sensor-shooter network,

in order to increase their lethality and

survivability. Unfortunately, the proliferation

of these soldier portable systems also increases

the demand for more batteries.

Taking a long-term view of our energy situation

and challenges, we need to do more to address

the problem of our ever-increasing energy

demand.

Figure 3. Power requirements for different military applications

134

TACKLING THECHALLENGES FOR INFRASTRUCTURES

Since the 1990s, an Energy Management and

Economisation Committee has been set up

under the auspices of the Ministry of Defence

(MINDEF) Economic Drive Committee to

regularly review the energy and water

consumption in all MINDEF building and

infrastructure assets. Over the years, the

committee has generated solutions such as

energy thermal storage, sea water cooling,

building integrated photovoltaic systems andcoastal wind turbines. There is still much thatcan be done to reduce energy consumption,improve energy efficiency and conservationfor MINDEF and the SAF.

Building Technologies

Mechanical & Electrical (M&E) systems in olderbuildings can be further optimised with theinstallation of energy management and controlsystems. MINDEF is taking steps to implementcost-effective measures to improve the energyefficiency of these buildings. There are manyenergy-efficient alternatives today which donot require high capital investments. Examplesinclude variable speed drives for motor controls,

motion-triggered lighting devices and white

light-emitting diode (LED) lighting2. Fluid

dynamics simulation software can now be used

to study airflow vectors to optimise the design

of air ducting or under-floor cooling systems

commonly used in data centres. A facility

management system can be installed to enable

an automated control of M&E systems which

will better match the M&E system performance

to the operation patterns in the building.

Similarly for new buildings, an integrated

building design approach can be adopted

where the architecture and M&E systems of

the building are designed in consideration of

their impact on each other. The architecture

of a building affects the solar heat gain, the

amount of daylight entering the functional

space and the amount of natural ventilation.

The goal is to create an architectural design

that minimises solar heat gain and yet allows

the building to exploit the natural environment

for day lighting as well as for natural ventilation

to reduce air-conditioning requirements.

Renewable Energies

Renewable energy installations in MINDEF

facilities provide us with a test bed for free

energy exploitation by the SAF. They offset

our growing dependence on and offer

alternatives to grid power. There is much

Figure 4. Progress from the 1970s to the modern soldier today

135

potential for MINDEF to adopt renewableenergy solutions to satisfy part of our energydemands, and in particular, solar energy hasbeen identified as a viable alternative. MINDEFmanages approximately 20% of mainlandSingapore and 60% of offshore land, most ofwhich are not densely built up. The longcoastlines of naval bases and islands serve asgood wind catchment areas which can beexploited. Figure 5 shows an example of smallscale wind turbine systems (500W and 1kW)installed at the pier of a coastal facility.

Most of the army camps are also located innon-built-up areas and are suitable sites forthe installation of solar arrays. Since the late1990s, building integrated photovoltaic systemshave been explored using solar panels as partof the building facade and the resultant costsavings help to offset part of the cost of thesolar panels. In some sites, solar panels havebeen installed as part of the building facade,in place of glass or metal panels to harnessenergy, which is then fed back to the grid andused to power lights at the atrium.

Energy Storage

Energy storage is another important aspect ofenergy management. Wind energy, forinstance, is highly variable and thus unsuitable

to be fed directly to loads. Hence, an energystorage device for wind energy is needed. Areliable energy storage system enables thewind and solar energy harnessed to be usedto support remote offshore installations.

Currently, computer applications are supportedby uninterruptible battery power systems (UPS)for instantaneous backup power. The typicalUPS is inefficient, consuming 10W for every100W of power supplied to the computer datacentre. To sustain the operability of the battery,an additional 15W to 20W is required tomaintain the batteries in a cool ambientclimate. Furthermore, most UPS systems arebased on lead-acid battery technology whichcan be costly and has a relatively shortoperational lifespan. DSTA is currently studyingthe application of a regenerative fuel cellssystem as an alternative form of energy storage.The regenerative system consists of a ProtonExchange Membrane (PEM) fuel cells modulecoupled with a PEM electrolyser module. ThePEM electrolyser module directs the electricityfrom a renewable or grid source to break thechemical bond of water to produce hydrogenfor storage, while the PEM fuel cell moduleconverts the energy released from the chemicalreaction of hydrogen and oxygen intoelectricity. In collaboration with fuel cellmanufacturer P21 Gmbh, DSTA has developed

Figure 6. Building integrated photovoltaicpanels installed as part of a building facade

replacing conventional metal or glassmaterials

Figure 5. Small-scale wind turbinesat a coastal facility

136

a prototype self-regenerative fuel cell system(see Figure 7) that is able to function as abackup power system, especially for remoteappplications. The prototype can either beconnected to the grid or to renewable energysources. There are two operating modes –Backup Power Mode and Energy CollectionMode.

Flywheel and super-conducting magneticenergy storage systems promise lower energywastage and are also currently being exploredby others as alternatives to the UPS.

Enabling Persistent Surveillance

It was shared at a Defense Advanced ResearchProject Agency (DARPA) conference in 2008that a fleet of eight Global Hawks would berequired to maintain constant monitoring ofa point target located 3,500 miles away. Itwould be extremely costly and henceimpractical to sustain such an operation. Thereare benefits to extend the endurance ofunmanned systems so that they could operatewith fewer platforms over a larger area ofoperations, yet achieving the desiredpersistence over targets.

Unmanned systems can be broadly classifiedinto four categories: Unmanned Ground

Sensors, Land Robots, Unmanned AerialVehicles (UAVs) and Autonomous UnderwaterVehicles (AUVs). Today, these systems havepower demands ranging from 10W up to 10kWand are likely to rise with more sophisticatedsystems. Yet, the payloads are not expected tobecome lighter. For robots, the energy sourceneeds to be ultra-compact due to the limitedspace available. In addition, land robots are tobe ruggedised and semi-submersible to enableuse on all terrain. It is more demanding topower UAVs and AUVs as they require bothcompact and lightweight sources with a highdegree of reliability. Furthermore, a UAV powersystem, for example, must be able to supplypeak power surges at several times the nominalpower during take-offs and dashes.

Hybridisation

It is our aim to power and sustain our UAVs toextend their persistent envelopment. In theshort-term, this can be achieved through ahybrid power system based on a smart powermanager capable of optimising power usage.This hybrid system uses a battery and fuel cellswhile simultaneously harnessing solar energythrough photovoltaic membranes on the bodyof the UAV. A fuel cell-battery hybrid exploitsthe high energy density of fuel cells for longendurance glide and the high power density

Figure 7. A regenerative fuel cell system being tested for its applicationas a remote or offshore power system

137

of a battery to provide peak power duringtake-off. Hybrid systems have the potential tosatisfy power demands and at the same timereduce the gross take-off weight of the UAV.In addition, the harnessing of solar energyfurther reduces the demands on the fuel cellsand batteries.

Fuel Cells

In the longer term, there is a need to developnew power generation technologies for higherefficiency and reliability. Solid oxide fuel cell(SOFC) technology is a potential enabler toultra-efficient electric motors. The primaryadvantage of SOFC is its ability to use a widerrange of fuels, as opposed to other fuel cellsin the market today. However, SOFC requireslong start-up times and it operates at hightemperatures of 800ºC to 1000ºC. Theseunfavourable operating limitations must beovercome before it can be useful to militaryapplications. DSTA and Nanyang TechnologicalUniversity are now experimenting with the

idea of microtubular SOFCs, which have thepotential for higher power densities than whatcan be achieved presently. The aim is to developsmall and lightweight energy solutionsto achieve flight persistence for the UAVs.

Advanced Composites

With the same given energy solution, thelighter the UAV, the longer it can stay in flight.Reducing the weight of platforms is thereforea revolutionary approach to achievebreakthroughs in power efficiency. Lightweightsteel alloys and polymer composites will soonreplace conventional materials, hence allowinglighter platforms to be realised (Lovins, Datta,Bustnes, Kooney and Glasgow, 2005). Currently,such advanced materials are expensive. Weneed to develop new processing andfabrication technologies to drive down thecosts for such materials. DSTA understands thebenefits of low cost composite materials andhas partnered with DSO National Laboratoriesand the local universities to develop new

Figure 8. An illustration of a tri-brid UAV system that exploitssolar cells, fuel cells and lithium-polymer technologies

138

fabrication techniques. Combined with highlyintegrated innovative structures with netshaped composite parts, the aim is to reducethe number of component parts andassembling procedures and work towardslightweight and cost-effective platforms.

SUSTAINING SOLDIERS' PEAK PERFORMANCE

The Third Generation soldiers will be equippedwith an array of advanced soldier systems forsensing, fire, protection and communications.A Third Generation soldier equipped with theAdvanced Combat Man System (ACMS) alonewill need to carry a sizeable amount of batteriesfor his missions. Apart from his basic combatload the soldier needs to be supported by along logistic train. For a seven-day mission, thebattery weight is much greater at 23kg. Thisenergy load is too heavy for our ThirdGeneration soldiers and we need lighter andmore efficient solutions to enable peak soldierperformance during sustained operations(Chua, Tang and Ho, 2007).

Battery Technology

Battery technology has evolved over the yearswith new chemistry and higher energy densitiescontinuously being achieved. Before the adventof lithium-ion (Li-ion) technology, the use ofrechargeable batteries was limited as comparedto primary batteries due to low energydensities. Nickel cadmium (Ni-Cd) and nickelmetal hydride (Ni-MH) batteries have energy

Figure 9. 90% composite airframe on display atthe Singapore Air Show 2008

Figure 10. An illustration of the micro-tubular solid oxide fuel cells that can be developedfor military applications

139

densities below 100Wh/kg. Li-ion batteries,with an energy density (approximately160Wh/kg) comparable to that of primarybatteries, changed the scene. Their introductionencouraged the use of secondary batteries andgradually replaced primary batteries. DSTArecognises this emerging trend and has initiatedthe replacement of primary alkaline batteriescurrently used in 900 series radios to provideweight reductions to combat soldier loads,bringing about significant cost savings forthe SAF.

Unfortunately, the current Li-ion batteries areengineered to lower capacities due to safetyissues and are still too heavy to support theACMS. Research on improving the performanceof lithium-based batteries is intensive andnanotechnology could be the new answer tothis challenge. Going small increases the activesurface area and also enhances the intercalationof metal-ions between the electrodes. A123Systems, for instance, has developed a form ofLi-ion battery that incorporates nanophosphateelectrodes. The improvements brought aboutby this new concept include a high dischargerate and fas t charge capabi l i t y 3 .Nanotechnology holds the promise of increasedstorage capacity, higher power, longer life andimproved safety due to lower heat generation.In the near future, these 'nano-batteries' will

be a viable energy solution for our ThirdGeneration soldiers.

Forward Field Charger

For special operations, the primary concern isgetting re-supplied. Batteries should berecharged in situ by drawing on availableenergy resources. The aim is to plug-and-playwith any type of available power source anddistribute power optimally to a myriad ofequipment. Through this mode, emerging highenergy sources such as fuel cells can beexploited to recharge field batteries. Solarenergy harvesting devices can also be used tosustain prolonged operations which will reduceoverall mission weight and cost withoutcompromising their capability and endurance.

Although high energy sources such as fuel cellsare promising alternatives, batteries will remainthe dominant solution for the army's energyneeds in the near future as they are the most

reliable and rugged. An annual increase of

approximately 5% in terms of energy density

can be expected for lithium battery technology.

On the other hand, fuel cell technology is more

relevant for supporting lengthy missions and

energy intensive operations. In the longer term,

the goal is to have each soldier totally self-

sustained. Imagine low-cost, printable solar

Figure 11. Graph comparing energy densities attained by some batteries today

140

cells with over 50% efficiency woven onto

every soldier's uniform, harvesting power in

situ. This is not an impossible feat4 and we will

continue to work closely with the various

national solar initiatives and overseas solar

research centres to realise this dream.

ENHANCING MOBILITYAND POWER RESILIENCE

Our Third Generation command posts andcritical nodes in the battlefield must have theflexibil ity to disperse and co-locateexpeditiously according to the operationaltempos and scenarios. For critical nodes suchas communication nodes that may be deployedin isolation (such as on a hill top), a high levelof self-sustenance is essential. This may requirethe power sources to be built and sustainedwithin the individual platforms. Althoughconventional diesel generators may be the

immediate solution, the vibration and

intolerable noise level at close proximity to

these generators undermine the operating

conditions and could compromise the locations

of command posts and critical nodes.

Furthermore, the need for a large fuel re-

supply remains, thus curtailing the operational

tempo and flexibility of operations and at the

same time, presenting windows of vulnerability

that could be exploited by the enemy.

Silent Generators

One of the short-term solutions to provide

silent and efficient on-board power is the free-

piston Stirling engine (named after its inventor,

Robert Stirling). Stirling converters were

originally developed for NASA space

applications, and the technology has translated

into power generators that are highly efficient,

reliable, maintenance-free and quiet. At the

Figure 12. An illustration of the benefits of a forward field charger for the advanced soldier –the charger can be built upon various forms of fuel cell technologies

141

DSTA has initiated the development of a smartpower management system capable ofdynamically switching in and out of powersources and loads for minimum disruption ofpower supply. The system is being developedto operate at varying engine speeds in responseto the loads so as to maximise energy efficiency

out in the field. The development of a high

resilience power distribution system will provide

superior field power surety and fuel savings,

hence reducing logistic footprint. In addition,

the ability to plug-and-play with renewable

sources will also be advantageous for extended

disaster relief missions.

43rd Power Conference, the US Army presentedits focus on this emerging technology and alsoannounced that it has achieved a gross fuel-to-electric efficiency of up to 22%. This issubstantially higher than the conventionaldiesel/gasoline engines operating in the 1kWto 3kW power range at an efficiency of 10%to 15%. External combustion Stirling engineshave the advantage of quiet operations whichis essential for military action requiring lowsignatures. They also have the capability toaccept a multitude of fuels – besides militarydiesel, any combustible fuel will work.

Mobile 'Microgrid'

The resilience of future command posts canalso be improved by making use of a smartpower management system, as depicted inFigure 14. Currently, the internal combustionengine-based diesel generators deployed areswitched on continuously and their capacitiesare specified to handle peak outputrequirements. This is an inefficient use ofenergy as the generators are mostly runningon part load. Furthermore, the excess powercapacity of individual generators cannot beshared with adjacent generators to improvethe overall system resilience.

Figure 14. An illustration of power management for field command posts

Figure 13. Next generation 3kW free-piston StirlingGenerator by Infinia Corporation contract by US Army

REFERENCES

142

Cambridge. (2004). Arctic Climate ImpactAssessment 2004. Impacts of a Warming Arctic,Cambridge University Press: UK.

Chua, E.H., Tang, E.H., Ho, S.F. (2007) FeedingPower Needs of the Future Army. InnovationsMagazine, Vol 7, pp 45.

Liu, X. (2006). China's Energy Security and ItsGrand Strategy. The Stanley Foundation, US.

Lovins, A.B., Datta, E.K., Bustnes, O.E., Koomey,J.G., Glasgow, N.J. (2005). Winning the OilEndgame: Innovations for Profits, Jobs, andSecurity.

1 Speech by Prof S Jayakumar, Deputy PrimeMinister, Co-ordinating Minister for NationalSecurity and Minister for Law (8 Nov 2006) atthe Singapore Energy Conference, IslandBallroom, Shangri-La Hotel, Singapore.

2 Though current LED lighting performance is80lumens/Watt, lower than that of fluorescentlighting at 90lumens/Watt, its compact sizeenables it to be located nearer to theillumination target and hence allows moreflexible applications. DSTA is exploring thewider applications of LED technology as itsperformance is expected to match that offluorescent lighting in one to two years’ time.

3 A123 Systems, one of the largest lithium-ionbattery manufacturers for specialty vehicles.

4 In 2007, DARPA announced a break-throughworld record performance of 42.8%.

Zhang S.S., Foster, D., Wolfenstine, J., Read, J.(2008). Safety Issue and Its Solution of Li-IonBatteries. Proceeding of the 43rd Power SourcesConference.

ENDNOTES

CONCLUSION

From the peacetime load growth of buildings

and infrastructures to the operational

sustenance of the soldier, the energy and power

demands of MINDEF and the SAF are wide-

ranging and challenging. It is hoped that this

article will initiate further discourse on the

relationship of energy to our defence

ecosystem. This paper has attempted to suggest

a variety of energy solutions customised for

infrastructures, soldiers, vehicles and some

forms of unmanned systems. Through

innovations, energy management and adoption

of alternative energies, the operational

performance of our fighting forces can be

boosted. There are many other areas of energy

and power management that have yet to be

explored, especially for naval and air platforms.

To manage these issues, DSTA and MINDEF willcontinue to innovate, harness and exploitscience and technology, as well as adopt anintegrated and systems solution approach, toreduce consumption, eliminate wastage,improve efficiencies and also power and sustainthe SAF’s capabilities in the battlefield.

Gareth Tang Ee Ho was the Technology Manager in the Directorate of Research& Development. He pioneered the research and applications of alternativeenergy technologies in various military systems and platforms. Gareth hasalso nurtured a team to design and build new energy solutions for soldierportable power, unmanned systems and also ground infrastructures. His workin this area has resulted in reduced soldiers’ battery weight, increased efficiencyof energy and power systems and significant cost savings for the Ministry ofDefence (MINDEF). He holds a Bachelor degree (Honours) and a Master inEngineering from the University of Cambridge, UK.

BIOGRAPHY

143

LTC Yeo Chee Beng is a Domain Head in Defence Research & TechnologyOffice (DRTech). He is one of the pioneers in DRTech and is actively involvedin the shaping of the Singapore's defence technology ecosystem. LTC Yeo isa Signal Officer by training and is an advocate of environmental conservation.He holds a Bachelor degree in Engineering (Honours) from King's Collegefrom the University of London, and a Master of Science in System Engineeringfrom the Naval Postgraduate School, USA.

Ho See Fong is a Senior Engineer (Mechanical and Electrical – ProtectiveInfrastructure). He is involved in the development and project managementof electrical networks in MINDEF facilities and advises on the lightning safety.He has also been involved in projects exploring new technologies in Energyand Power using fuel cells. He received his Bachelor degree in Electrical andElectronic Engineering from the Imperial College, London, United Kingdom,before furthering his training to attain a Master of Science in Electric PowerEngineering from the Royal Institute of Technology, Sweden.

MAJ Oliver Lan Chi Wai is with the Singapore Command and Staff Collegeand is currently a student in the 40th Command & Staff Course. He waspreviously a Staff Officer in Systems Integration Office, General Staff(Development). Oliver graduated with a Bachelor degree in ElectricalEngineering from the National University of Singapore and went on to obtaina Master of Science in Communication & Information Science, KnowledgeManagement from Nanyang Technological University.

Chairman

Sin Boon WahDeputy Chief Executive (Special Duties)Deputy Chief Executive (Strategic Development)[Up to 31 March 2009]

Members

Tay Kok PhuanDirector (DSTA College)

Lee Chee TengDirector (Procurement)

Wee Kok LingDirector (Networked Systems)

Yeo Mei Li RosemaryDirector (Human Resource)

Chew Keng CheowDeputy Director (Guided Weapons & ArmamentCompetency Community)

Lee Yeaw Lip AlexDeputy Director (Systems EngineeringCompetency Community)

Ong Yew HingDeputy Director (Building & InfrastructureCompetency Community)[Up to 15 March 2009]

Seah Peng HweeAssistant Director (Sensing & ConnectivityCompetency Community)

Ho Ai PhangAssistant Director (Corporate Communications)

Quek Bee TinSenior Manager (DSTA College)

Chua Siew Ting PearlySenior Manager (Corporate Communications)

Too Meng YuenSenior Executive (DSTA College)

Chng Wang Chin JoleneExecutive (Corporate Communications)

A C K N O W L E D G E M E N T S

Peer Reviewers

Lee Chee TengDirector (Procurement)

Tan Yang HowDirector (DSTA Masterplanning & SystemsArchitecting)

Tay Kok PhuanDirector (DSTA College)

Wee Kok LingDirector (Networked Systems)

Chew Keng CheowDeputy Director (Guided Weapons & ArmamentCompetency Community)

Lee Yeaw Lip AlexDeputy Director (Systems EngineeringCompetency Community)

Ong Yew HingDeputy Director (Building & InfrastructureCompetency Community)[Up to 15 March 2009]

Seah Peng HweeAssistant Director (Sensing & ConnectivityCompetency Community)

The Editorial Board, DSTA Horizons

Technical Editor

Professor Bernard TanDepartment of Physics, Faculty of ScienceNational University of Singapore