1-s2.0-s1526612514000462-main

8/10/2019 1-s2.0-S1526612514000462-main

1/17

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),http://dx.doi.org/10.1016/j.jmapro.2014.05.004

ARTICLE IN PRESSG Model

JMP-244; No. of Pages17

Journal of Manufacturing Processes xxx (2014) xxxxxx

Contents lists available at ScienceDirect

Journal ofManufacturing Processes

journal homepage: www.elsevier .com/ locate /manpro

Modeling and optimization ofturning duplex stainless steels

Rastee D. Koyee a,, Uwe Heisel a,1, Rocco Eisseler a,2, Siegfried Schmauder b,3

a Institute forMachine Tools, Stuttgart University,Holzgartenstr.17, D-70174 Stuttgart, Germanyb Institute forMaterials Testing, Materials Science and Strength of Materials, Stuttgart University, Pfaffenwaldring 32, D-70569 Stuttgart, Germany

a r t i c l e i n f o

Article history:

Received 14 January 2014Received in revised form 25 March 2014

Accepted 9 May 2014Available online xxx

Keywords:

Duplex stainless steelsResponse surface methodologyCuckoo searchTOPSISOperational sustainability index

a b s t r a c t

The attractive combination ofhigh mechanical strength, good corrosion resistance andrelativelylow costhas contributed to making duplex stainless steels (DSSs) one ofthe fastest growing groups ofstainless

steels. As the importance ofDSSs is increasing, practical information about their successful machiningis expected to be crucial. To address this industrial need, standard EN 1.4462 and super EN 1.4410 DSSsare machined under constant cutting speed multi-pass facing operations. A systematic approach whichemploys different modeling and optimization tools under a three phase investigation scheme has beenadopted. Inphase I,the effect ofdesign variables such as cutting parameters,cutting fluids and axial lengthofcuts are investigated using the D-Optimal method. The mathematical models for performance charac-teristics such as; percentage increase in radial cutting force (%Fr), effective cutting power (Pe), maximumtool flank wear (VBmax) and chip volume ratio (R) are developed using response surface methodology(RSM). The adequacy ofderived models for each cutting scenario is checked using analysis ofvariance(ANOVA). Parametric meta-heuristic optimization using Cuckoo search (CS) algorithm is then performedto determine the optimum design variable set for each performance. In the phase II, comprehensiveexperiment-based production cost and production rate models are developed. To overcome the conflictbetween the desire ofminimizing the production cost and maximizing the production rate, compromisesolutions are suggested using Technique for Order Preference by Similarity to Ideal Solution (TOPSIS).The alternatives are ranked according to their relative closeness to the ideal solution. In the phase III,expert systems based on fuzzy rule modeling approach are adopted to derive measures of machining

operational sustainability called operational sustainability index (OSI). Artificial neural network (ANN)based models are developed to study the effect of design variables on computed OSIs. Cuckoo searchneural network systems (CSNNS) are finally utilized to constrainedly optimize the cutting process pereach cutting scenario. The most appropriate cutting setup to ensure successful turning ofstandard EN1.4462 and super EN 1.4410 for each scenario is selected in accordance with conditions which give themaximum OSI.

2014 The Society ofManufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction

Duplex stainless steels (DSSs) are chromiumnickelmolybdenumiron bi-phased alloys in which the proportionsof the constituent elements enable the optimization of the balanceof the volume fractions of austenite and ferrite [1]. In many ways,the duplex stainless alloys represent a best of both worlds incombining traits from the austenitic and ferritic alloys. They have

Corresponding author. Tel.: +49 711 685 84558; mobile: +49 163 9187997.E-mail addresses: [email protected] (R.D. Koyee),

[email protected](S. Schmauder).1 Tel.: +49 711 685 83860; fax: +49 711 685 73860.2 Tel.: +49 711 685 83876; fax: +49 711 685 83858.3 Tel.: +49 711 685 62556; fax: +49 711 685 62635.

a structure that is roughly a 50% mix of austenite and ferrite; thus,two hardness materials with different hardnesses coexist side byside. The presence of ferrite however, means that DSSs have aductile brittle transition temperature. During machining process,the tool will alternate cutting between soft and hard grains ofthe duplex structure, leading to an automatic tendency to initiatechatter in the cutting system. They possess quite a bit highermechanical strength and lower ductility than standard austeniticstainless steels. Therefore, DSSs are expected to be more difficultto machine than the standard austenitic stainless steels of similarcorrosion resistance [2].

In the last ten years, several studies on machining of stainlesssteels have been conducted [38]. The cutting process modelingby response surface methodology (RSM) using statistical design ofexperiments based on D-optimal designis provedto be an efficientmodeling tool [911]. Researchers have considered the machining

http://dx.doi.org/10.1016/j.jmapro.2014.05.0041526-6125/ 2014 TheSociety of Manufacturing Engineers. Publishedby Elsevier Ltd. All rightsreserved.
http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004http://www.sciencedirect.com/science/journal/15266125http://www.elsevier.com/locate/manpromailto:[email protected]:[email protected]://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004mailto:[email protected]:[email protected]://www.elsevier.com/locate/manprohttp://www.sciencedirect.com/science/journal/15266125http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004

8/10/2019 1-s2.0-S1526612514000462-main

2/17



JMP-244; No. of Pages1 7

2 R.D. Koyee et al./ Journal of Manufacturing Processes xxx (2014) xxxxxx

of metals with multiple performance characteristics and applica-tion of one or more multiple attribute decision making (MADM)techniques to optimize the process parameters [1218]. Consider-ing the drawbacks of traditional optimization techniques, attemptsare being made to optimize the multi-pass machining prob-lem using evolutionary optimization techniques [1924]. Logicalfuzzy reasoning has been applied to multiple output optimiza-tion of machining processes [2527]. As investigated by previousresearchers, the artificial neural network (ANN) is a powerful toolfor dealing with the complex nature of the cutting process [2830].A number of attempts have also been made to study the machin-ing of DSSs experimentally [3135]. For instance, Nomani J. andhis coworkers have conducted machinability tests on duplex alloysSAF 2205 and SAF 2507, while employing austenite stainless steel316L as a benchmark during drilling. Both duplex alloys displayedpoorer machinability responses, with 2507 being worst [36]. DeOliveira et al. studied the turning operation of SAF 2507, and itsinfluence on the alloys corrosion resistance in practical applica-tions. The results indicate that turning with PVD-coated insertsunder high-pressure cooling resulted in long tool lives, good work-piece roughness and high corrosion resistance of the material aftermachining. The most frequent wear mechanism found during thetests was notch wear, while the main tool wear mechanism was

attrition [37]. Selvaraj et al. are optimized dry turning parametersof two differentgrades of nitrogen alloyed duplex stainless steel byusing Taguchi method. Their results revealed that the feed rate isthe more significant parameter influencing the surface roughnessand cutting force. The cutting speed was identified as the moresignificant parameter influencing the tool wear [38]. Most recentworks of Krolczyk et al., in their fourth contribution, identifiedmicrohardnessof surface integrity(SI) afterturningwith wedges ofcoated sintered carbide. The investigation included microhardnessanalyses in dry and wet machining of duplex stainless steel. It hasbeen shown that wet cutting leads to the decrease of SI harden-ing depth, while increasing the rounded cutting edge radius of thewedge increasesthe maximum microhardnessvalues andthe hard-ening depth [39]. Finally, some recently conducted studies have

focused on machining technologies in order to achieve sustainabledevelopment objectives [4044].

In majority of the aforesaid past researches, very few effortshave been made toward application of recent modeling and opti-mization techniques foroptimizing the machiningof DSSs. In orderto provide a scientific approach to select the most proper cuttingconditionfor machining DSSs,this paperpresents theapplication ofrelatively new statistical techniques in experimental design, arti-ficial intelligence, MADM and recently developed meta-heuristicoptimization algorithms in the context of a three phase investiga-tion. The flowchart of the study is shown in Fig. 1.

2. Materials and methods

2.1. Workpiece material

Although there are many varieties of duplex stainless steelsbeing cut in industry, the current research selected two com-mon typesof these materialsfor machining investigation. Standardduplex EN 1.4462 and super duplex EN 1.4410 stainless steel rodsof diameter 55mm were used for all the turning experiments.The chemical composition and mechanical properties of workpiecematerials are listed in Table 1.

2.2. Machine tool and cutting tool

CNC lathe CTX 420 Linear V5 with maximum drive power

25kW and a speed range of 357000rpm is used to perform the

D- Optimal design

Categorical factors: Cutting Fluid Type

vcm/min

fmm/rev

apmm

Lcmm

Response Surface Modeling

VBmm

%Fr R PcW

OSI

Ranking

Preference

Optimum

Cutting set

min(MSE)?No

Yes

Feasible region C1Feasibleregion C2

Feasibleregion C3

Feasibleregion C4

Solution

MADM(TOPSIS)

Optimized

CuttingParameters

CSNNS

OSIpredict.

MachineOccup.Time

(t )bB

M. Time-Related

Cost

K1

C. Tool-Related

Cost

K3

FuzzyLogic

Neural Network

Model

Inputs

Outputs

Lc vcfap

OSIactual.

Adjust Network

via Taguchi

Fig. 1. Flow chart of thestudy.

experiments. Coated carbide inserts with ISO code of CNMG120408-QM 2025 are used in this study. The QM designation hasa sharp positive edge and open chip breaker which promoteslow cutting force and more stable cutting operation. The insertshad a multilayer CVD coating (TiN/MtTiCN/Al2O3) ona cementedcarbide substrate and were mounted on a right hand style PCLNL-2525M-12 ISO tool holder with a tool geometry feature as follow:including angle = 80, back rake angle=6, approach angle = 95and clearance angle = 5.

2.3. Experimental design

Optimal design methods use a single criterion in order toconstruct designs for RSM; this is especially relevant when fit-

ting second-order models. D- and A-optimality criteria provide a
http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004

8/10/2019 1-s2.0-S1526612514000462-main

3/17

8/10/2019 1-s2.0-S1526612514000462-main

4/17





External effective-power modules are installed in machine toolsbetween the frequency converter and the motor. To read the cur-rent, hall sensors are used that measure the current along with itsphasing via the magnetic field surrounding the conductor in a cir-cular shape. The effective power is obtained from these using Eq.(2) [46].

Pe=

3.U.I. cos (2)

where U is the line voltage in volts, I is the line current in ampereand cos is the power factor. The average Pe values of cuttingpasses were added and later divided by total number of passes todeterminethe averageeffectivecuttingpowerconsumptionin eachexperiment run.

At the endof each experimentalrun thetoolwearwas measuredusing an optical microscope and the tool life was estimated basedon flank wear criterion ofVBmax =600m on the major cuttingedge. During cutting tests the metal chips posed great challengesto the machine tool and the workpiece. To deal with this prob-lem, the chips were collected after each machining trial and lateranalyzed. Chip volume ratios (R) were then calculated by; (a) mea-suringthe mass of thesample chip, (b)dividing the mass by densityof the steels hence calculating the volume of the cut material priorto machining, (c) gently placing the chips inside plastic bags, (d)gently vacuum packaging which removed air from the packageprior to sealing, (e) measuring the volume of the gently packedchips using the water displacement procedure and (f) dividing thevolume of the gently packed chips by volume of the cut materialprior to machining. The ratio corresponds to the intended R. Theexperimental results as shown in Table 3 were utilized for the nextanalyses.

2.3.2. RSMmodeling of performance characteristics

Response surface methodology (RSM) is a collection of statisti-cal and mathematical techniques useful for developing, improving,and optimizing processes. It also has important applications in the

design, development, and formulation of new products, as well asin the improvement of existing product designs. Successful useof RSM is critically dependent upon the experimenters ability todevelop a suitable approximation for response function. Usually,a low-order polynomial in some relatively small region of theindependent variable space is appropriate. In many cases, eithera first-order or a second-order model is used. Often the curva-ture in the true response surface in machining experimentationsis strong enough that the first-order model (even with the interac-tion term included)is inadequate. A second-order model will likelybe required in these situations. The second-order model is widelyused in response surface methodology for their flexibility, easinessin estimating the constant parameters, and indications of consider-ablepractical experiences which confirm thatsecond-ordermodels

work well in solving real response surface problems [47]. However,when the entire design space is used to develop the models, then,second-order models, modified second order models, cubic mod-els, modified cubic models, response transformations and modeldesign reductions were seen not satisfactory to capture accuratelythe highly non-linear relations between the cutting variables andperformance characteristics. Therefore, an intermediate and rea-sonable solution were found for all the cases when second-ordernon-linear mathematical models in terms of natural variables areselected to predict theperformance characteristics(Y)foreachcon-dition category in separate, which were of the following form:

Y = b0 + b1vc+ b2f+ b3ap + b4Lc+ b11v2c+ b22f2 + b33a2p+ b44L2c

+b

12v

cf+b

13v

cap +

b14

v

cLc+

b23fa

p +b

24fL

c+b

34apLc

(3)

The values of the regression coefficients as defined above aredetermined using the following formula:

b = (XTX)1XTY (4)

where b, matrix of parameter estimates; XT, transpose of theindependent variables matrix (X); Y, matrix of the performancecharacteristic.

2.4. Cuckoo search

Cuckoo Search is one of the latest nature-inspired meta-heuristic algorithms, developed by Xin-She Yang and Suash Deb[48]. Cuckoo Search is inspired by lifestyle of a bird family calledcuckoo. Specific egg laying and breeding of cuckoos is the basisof this optimization algorithm. Like other evolutionary algorithms,the proposed algorithm starts with an initial population of cuck-oos. These initial cuckoos have some eggs to lay in some host birdsnests. Some of these eggs which are more similar to the host birdseggs have theopportunity to grow upand becomea maturecuckoo.Other eggs aredetectedby host birds andare killed. The growneggsreveal the suitability of the nests in that area. The more eggs sur-vive in an area, the more profit is gainedin that area. So theposition

in which more eggs survive will be the term that cuckoo optimiza-tionalgorithm is going to optimize[49]. For simplicity in describingCuckoo Search the following idealized assumptions are made:

a. Each cuckoo lays one egg at a time, and dumps its egg in a ran-domly chosen nest.

b. The best nest with high-quality eggs will be carried over to thenext generation.

c. The number of available host nests is fixed, and the egg laid bya cuckoo is discovered by the host bird with a probability Pabetween 0 and 1. In this case, the host bird can either get ridof the egg, or simply abandon the nest and build a completelynew nest.

3. Results and discussions

3.1. Effect of independent cutting variables on performance

characteristics

The values of regression coefficients in natural form and vari-ous statistics about the statistical validity of the developed modelsat a 95% confidence interval are given in Table 4. The R2 and R2

adj .

values indicate that the models fit the data well. As per analysis ofvariance (ANOVA) technique, since the calculated value ofFratioofall developed second-order models are greater than the standardtabulated value of the table (F14,9,0.05 = 3.03), then the models areconsidered adequate within the confidence limit. The adequateprecision (Aprec.) for all models are greater than 4, which indicate

an adequate signal to noise ratio, thus the models can be used tonavigate the design space.

3.1.1. Effect of independent cutting variables on the percentage

increase in radial cutting force (%Fr)

In phase I, the developed RSM models were utilized to studythe interaction effects of selected independent variables on %Fr.To analyze the interaction effects, three dimensional plots weregenerated considering two parameter at a time while the otherparametersareheldconstantattheirrespectivecenterlevels.Theseinteraction plots are presentedin Fig.4a. Fromwhich, the followingobservations can be made:

a. For a given depth and length of cut, %Fr approaches its mini-

mum value as cutting speed and feed rate increase to certain

8/10/2019 1-s2.0-S1526612514000462-main

5/17

8/10/2019 1-s2.0-S1526612514000462-main

6/17

Pleasecitethisarticleinpressas:KoyeeRD,etal.Modelingandop

timizationofturningduplexstainlesssteels.JManufProcess(2014),

Table 4

Models coefficients estimates and ANOVA.

Coef. EN 1.4462 EN 1.4410

%Fr Pe VBmax R %Fr Pe

Wet Dry Wet Dry Wet Dry Wet Dry Wet Dry Wet Dry

b0 70.6 69.4 1700.9 2165.7 467.2 604.2 29.8 88.9 75.8 22.3 1765.7 2013.1b1 0.59 0.48 20.63 5.30 3.03 1.91 0.67 0.32 0.51 0.21 25.55 18.75b2 386. 435 1384.5 23259 1550 4341.5 412.3 401.5 253.4 160.1 5243 14,453b3 9.22 0.74 811.25 0.52 69.37 257.56 144.9 173.5 10.17 21.75 1274.9 1963.4b4 1.58 1.70 61.81 46.41 10.37 19.05 13.30 31.54 1.94 2.01 59.15 181.15b11 0.00 0.00 0.13 0.07 0.01 0.01 0.00 0.00 0.00 0.00 0.16 0.06b22 931 1235 3539.4 50,842 2176 11,795 95.65 737.5 358.9 425.8 21,964 8438.1b33 6.29 1.34 88.09 8.97 27.36 103.39 118.20 166.96 4.15 9.63 236.10 869.17b44 0.10 0.08 3.74 0.54 0.08 0.44 0.90 1.60 0.03 0.11 2.81 9.98

b12 0.38 0.01 11.96 36.67 3.88 0.95 2.46 1.27 0.62 0.28 3.69 25.30b13 0.03 0.03 1.78 0.23 0.06 0.16 0.65 0.65 0.01 0.04 1.52 3.76b14 0.00 0.01 0.08 0.06 0.01 0.07 0.00 0.02 0.00 0.00 0.04 0.42b23 22.1 31.8 3421.3 3590.9 73.52 580.6 117.5 93.95 12.25 35.85 4036.3 3235.8b24 5.81 5.36 16.90 97.45 48.25 58.30 5.89 26.88 8.00 4.00 9.61 360.96b34 0.02 1.64 3.13 36.41 3.56 14.68 0.02 7.57 0.58 0.60 11.06 54.40

ANOVA

R2 0.96 0.97 0.99 0.98 0.92 0.98 0.97 0.93 0.92 0.95 0.99 0.97R2

adj 0.90 0.93 0.98 0.95 0.80 0.95 0.91 0.83 0.80 0.87 0.97 0.93

RMSE 2.09 3.2 179 274 18.6 21.4 8.33 16.8 3.01 4.86 212 423Aprec. 14.9 17.7 24.5 17.1 9.82 22.91 15.57 9.66 9.82 12.67 21.59 14.67Fratio. 15 22.3 79.9 34.8 7.57 32.9 17.8 8.89 7.62 12.4 61.3 21.9Reg. DF 14 14 14 14 14 14 14 14 14 14 14 14Err. DF 9 9 9 9 9 9 9 9 9 9 9 9

RMSE, root mean squared error;Reg. DF, regression degrees of freedom; Err. DF, error degrees of freedom.

8/10/2019 1-s2.0-S1526612514000462-main

7/17




R.D. Koyee et al./ Journal of Manufacturing Processes xxx (2014) xxxxxx 7

Fig. 4. Interaction effects of independent cutting variables on the performance characteristics.

specified limit.Beyond that limit it startsto increase again.Whenthe materials are cut dry, lower cutting speed and feed rates arepreferred to minimize %Fr. The values of obtained %Frwhen EN

1.4410 are cut were higher than EN 1.4462 due to the generallyhigher mechanical strength of EN 1.4410 and fewer weight per-centage of assisting machinability elements such as sulfur andphosphorus, see Table 1.

b. For a specified cutting speed and feed rate, depth of cut valuesof (1.251.5 mm) were seen to give the minimum %Fr, as theyreduce theploughing effect, produce friendlier-to-machine chipsand exert a more stable cutting process. Generally, lower valuesof depths of cuts are recommended when process conditions aredry and/or super DSS EN 1.4410 is machined. Values lower thantool nose radius of 0.8 mm have to be avoided.

c. Whencuttingspeed, feed rate and depth of cut are set at a spec-ified value, %Fris expected to increase with the increase of totallength of cut. However, the slopes of %Frversus total length ofcut were seen steeper at higher levels of cutting speeds.

3.1.2. Effect of independent cutting variables on the effective

cutting power (Pe)

Effect-power measurement systems are characterized by thefact that they do not affect the mechanical properties of themachine tool. The machining torque can be measured during theoperation without the integration of external sensors into theelectric flux of the machine. The primary area of application isthe recognition of tool fractures and collisions in the workspace.Sufficiently large force changes are necessary for efficient wearmonitoring. When monitoring cutting processes, usually externaleffective-power measurement tools are used,often withassociated

evaluation software and visualization unit [46].

Fig.4b illustrates theinteraction effects of theindependent vari-ables on the effective cutting power. It is seen from this collectivefigure that:

a. Effective cutting power is most sensitive to cutting speed varia-tions when other parameters are kept constant. Maximum andminimum consumptions in effective cutting power are seen atwet cutting of EN 1.4462 and dry cutting of EN 1.4410 respec-tively.

b. Effective cutting power is less sensitive to the length of cut vari-ations as far as the cutting tool does not suffer the catastrophicchipping in the cutting edges.

c. Minimum consumption in effective cutting power results whenthe feed rate is in the range 0.1250.175 mm/rev of dry cuttingand 0.10.15mm/rev of wet cutting.

d. Two-factor interaction models were seen accurate enough toexplain theinteraction effects of cuttingparameters on theeffec-

tive cutting power.

3.1.3. Effect of independent cutting variables on the maximum

tool flankwear (VBmax)

Tool wear has a remarkable influence in tool life, cutting forces,vibration, quality of themachinedsurfaceand itsdimensional accu-racy,and consequently, the economics of cutting operations. In thisstudy off-line modeling of the maximum flank wear lands (VBmax)are described as functions of independent cutting variables. Fig. 4cdepicts the interaction contour plots ofVBmax as per RSM modelstabulated in Table 4. It can be found that:

a. During the course of the dry cutting of DSSs, an increased builtup edge formation on the tool, obvious inferior chip morpholo-

gies, more aggressive notchwearon the major and minor cutting

8/10/2019 1-s2.0-S1526612514000462-main

8/17





1 3

1

7

13

19

25

31

37

43

48

AlternativeRank

5 7 9 11 13 15 17 19 21 23 2

Alternativ

5 27 29 31

ve No.

33 35 37 39 41 43 45 47 48

EN 1.446

EN 1.441

2

0

Fig. 5. Alternatives ranks.

edges and inferior surface qualities were noticed. Therefore,when cutting DSSs, generous amount of cutting fluid is alwaysrecommended.

b. Wet cutting at cutting parameter ranges ofvc

100160m/min,f0.150.25mm/rev, ap0.751 mm for machining EN1.4462 and vc120160m/min, f0.150.20mm/rev,ap1.251.5mm for machining EN 1.4462 were seen optimumin minimizing the tool flank wear.

c. Dry machining is possible when cutting parameters are setappropriately. The settings include; machining of EN 1.4462at vc75110m/min, f0.150.2mm/rev, ap11.5mmand maximum Lc/ap6, and machining of EN 1.4410 atvc7590m/min, f0.1250.175 mm/rev, ap0.81.35mmand maximum Lc/ap4.

d. Directlinearcorrelations between %FrandVBmax with reasonablecoefficients of determinations (R2) were noticed:

Cutting of DSS EN 1.4462

Wet: R2 =0.70

VBmax= 5.4198(%Fr) + 35.149 (5)

Dry: R2 =0.92

VBmax= 7.7696(%Fr) 18.549 (6)

Cutting of DSS EN 1.4410

Wet: R2 =0.71

VBmax= 6.5767(%Fr) 16.793 (7)

Dry: R2 =0.93

VBmax= 9.918(%Fr) 63.494 (8)

3.1.4. Effect of independent cutting variables on the chip volume

ratio (R)

The chip shapes are assessed according to two criteria: trans-portability and danger for the machine operator. It is no problemtomoveshort,brokenchips,suchasfragmentedspiralchips,incon-tainers. By contrast, this is impossible for ribbon chips, since theyalways demand special treatment (breaking in the chip breakeror briquetting) in order to make them ready for transport. Ina plant with automated manufacturing equipment, where manychips occur, these treatment procedures are very expensive. Con-sequently, as analternative,always theaimis toproduce chip forms

that can be handled easily. Since long ribbon chips and entangled

chips, whose edges are very sharp, could possibly endanger themachine operator and cause safety risks [50]. In this study, utiliz-ing the chip volume ratio R, the spatial requirement for the chipsis considered. Each chip form is assigned to a chip volume ratioR, which defines by what factor the transport volume needed forthe specific chip form exceeds the intrinsic material volume of thechip. According to these assignments, R100 for snarled and rib-bon chips, 60R100 for coiled, flat helical and cylindrical helicalchips, 30R60 for short coiled chips, 10R30 for spiral chipsand 3R10 for short chip particles. It is clear from Fig. 5d that;

a. Higher values of cutting speeds and feed rates are necessary tominimize the chip volume ratio.

b. In wet cutting of DSSs, friendlier-to-machine chip forms wereproduced because of the less encountered friction in the contactarea between the chip and rake face.

c. Ribbon and snarled chip forms were common when the

metals are machined at feed rate and depth of cuts lowerthan 0.15mm/rev and 1 mm, respectively. At intermediate feedranges, the produced chips were rather of flat-helical andcylindrical-helical forms. The chip formswererather short coiledchips at higher feed ranges.

d. Under the same machining conditions which might producecontinuous chips, different chip forms were possible due to con-tinuous entanglement of chips especially when Lc/ap6.

e. Dueto thehigherproof and tensile strength, and, lower sulfurandphosphor constituents to assist chip breaking, R values of cuttingEN 1.4410 were generally higher than of cutting EN 1.4462.

f. The tendency to enhance chip segmentation is expected to risewhen the chip curl radius is reduced, the coolant pressure isincreased and the coolant restricted to smaller area of the chip.

3.2. Parametric optimization of performance characteristics

The first optimization process in this study is formulated as fol-lows:

3.2.1. Objective function

Formulation of optimizationmodel is one of the most importanttasks in optimization process. The type of optimization modelingtechniques used to express the objective function determines itsaccuracy and the possibility of reaching a global optimum solu-tion. The developed response surface models, expressed by Eq. (3)and presented previously in Table 4 are used as both objective and

constraint functions.

8/10/2019 1-s2.0-S1526612514000462-main

9/17

8/10/2019 1-s2.0-S1526612514000462-main

10/17





Production time or time per unit tecan be expressed in terms ofthe basic time tgand the idle time as follow:

te= tg+ ti (11)

The basic time tg is the sum of the main process time th andauxiliary process time tn.

tg= th + tn (12)

The main cutting time for constant cutting speed facing pereach cutting trial in Table 3 can be calculated using the followingexpression:

th= tc+ Cf (13)

Here, the cutting time tcis the time in which the tool is actuallycutting and determined by:

tc=

4000 (Lc/ap) ((2lr+ d)

2 + (1000vc/n)2)f vc

(14)

where d is the diameter of the workpiece inmm, lris the additionaltravelingradial distance of the cutting tool inmm and n is the max-imum rotational speed of the workpiece which set as 3500rpm.

The constant Cfis function of cut lengths which are considered inconjunction with the respective feed velocity.

Theauxiliaryprocesstime tn isthetimeduringwhichallindirectprocesses arising during the machining operation (e.g. tightening,measuring, adjusting, pro rata tool change and workpiece change)are executed. The idle time ti takes all pauses into considerationduring which the machine tools are not in operation and the totaltime required for all irregular events such as procuring necessaryresources. In this study, the following relation is considered validfor calculating the idle time:

ti= 0.3(th + tn) (15)

Total setup time trB refers to the time required for machine set-up t

vM, tool change t

rWand nonproductive set-up activities t

rV. The

latter is often estimated as 30% of the machine set-up and toolchange time.

trB= tvM+ trW+ trV (16)

For a batch ofm workpieces per a machine, total tool changetime is defined as:

trW= m tWZthT

(17)

where tWZis the time that passes till a single tool is changed, andboththe position correctionand positioning for re-entry have takenplace. The tool life T is calculated based on tool wear criteria ofVBmax =600m. The developed RSM models ofVBmax are utilized

to find the maximum allowable length of cuts Lcmax when VBmaxis 600m. Once the length of cuts are determined, they are sub-stituted into Eq. 14 to find the tool life for each cutting trial. Theoverall working time taper x number of machines is described as:

ta= te m

x (18)

Finally, the followingrelation is true for the machine utilizationtime per workpiece or process:

tbB= m

1.3

4000 (Lc/ap) ((2lr+ d)

2 + (1000vc/n)2)f vc

+Cf+ tn+ tWZthT+ trV+ tvM (19)

3.3.2. Production cost

A typical production cost for a workpiece produced by turningoperations is comprised of machinecosts, labor costsand toolcosts:

KF= KM+ KL + KW (20)The machine hour-rate describes the costs to be calculated of a

machine tool per hour. The Machine hour-rate KMis calculated asfollow:

KM= 1JAS

kbBtl

+ kbW+ kbZ+ kbR+ kbE (21)

Annual operating hoursJASof a machine is defined as:

JAS = No.of operationhoursWeek(w.)

No.of workingshifts

No.of working weeksYear

(22)

The yearly machine runtime JAS amounts for example to16001800h/a for single-shift operation. In the case of multi-shiftoperation, the runtime is increased proportionately (e.g. two-shiftoperation ca. 3200h/a or three-shift operation ca. 4800h/a). Theprocurement cost kbB covers the purchasing, transportation and

installation costs. The timetlis defined as a timeframe atwhich themachine is economically utilizable. The cost of maintenance andrepair services kbWcan be expressed in terms of the percentage %pof the procurement cost kbB:

kbW=p

100 kbB (23)

Assumed interest rates can be set at the current value of themachine also of a non-depreciated element as a calculated averagebased on the procurement price at the full interest rate (%q):

kbZ= 0.5 %q kbB (24)Tocalculate the space costkbR, planningestimationsshould con-

sider themachine required areaQin m2 andthe monthly rentA perm2:

kbR= 12 Q A (25)The operating costs kbE include the costs of operation energy,

illumination and coolant kk. The electricity costs can be estimatedbased on effective cutting power PM, the standardcost of electricityEc in Euro/kWh and percentage %C2of being in ON state:

kbE= PM Ec %C2 + kk (26)Labor costsKLare calculated as follows:

KL= Lm(1+ r) (27)where Lm is the gross hourly wage and r as the amount of non-wage costs of the operator. It is appropriate when a newfactor thatsummarizes the machine cost and the wage rate per hour in termsof previously defined time scales is formulated as:

KML= KM+trB + tatbB

KL (28)

The costs of typical indexable carbide cutting tools are com-prised of the cost of tool holders kWH, inserts kWPand spare partskET:

KW= kWH+ kWP+ kET (29)The total insert costs are defined as:

kWP=m th KWSP

0.8 T Zs(30)

whereKWSPisthecostofaninsert, Zs is thenumber of usablecutting

edge per insert and 0.8 is a safety factor accounts for uncertainty

8/10/2019 1-s2.0-S1526612514000462-main

11/17





in tool life. The spare parts costs are often expressed in terms ofpercentage of the tool holder and insert costs. The final productioncost per unit is obtained by adding the terms in Eqs. (20)(35):

KF=1

m

tbB KM+ (trB + ta) KL + kWH+

m th KWSP0.8 T Zs

+ kET

(31)

Employing the previous relationships, one can define the maintime-related costs as:

K1=KM+ 1

xKL th (32)

The fixed or the workpiece-related costs as:

K2=1

m

KM+

1

xKL

(trM+ trV) +

KM+

KLx

(tn + tb + tvB) (33)

and the tool related costs as:

K3=thT

(KM+ KL)(tW) +1

mKW (34)

The production cost can also be determined using [46,51]:

KF= K1 + K2 + K3 (35)

3.3.3. Case study

To illustrate the application of the equationspresented in previ-ous two sections,productions of 12,000 identical components wereconsidered as a case study. The machining of each unit has to beaccomplished through removing a total cutting length of 12mmunder constant cutting speed facing operation. Eqs. (10)(35) wereemployed to define each machining economics terms presentedin this study. The objective is to simultaneously minimize theconflicting objectives such as machine utilization time tbB, maintime-related costs K1 and the tool related costs K3 using a mul-tiple attribute decision making (MADM) method called TOPSIS. Inconjunction withexperimentations andutilizing thefeatureof sim-ulation on the CNC-controller, the main cutting times of the trialswere accurately recorded. The data considered for the case studyare given below:

kbB= 100,000D Q=40m2 KWSP= 6 Dm= 12,000 A= 30 D/m2.Mo. x= 1tn= 1.5min Ec= 0.130D/kWh PM=25kWtvM=30 min kk= 0.90 D/h %C2= 30tWZ= 0.5min Lm= 34 D/h h/week= 40tl=2 years

d=55mmr=2.5kWH= 80D

week/annum= 40

lr=6.875mmp= 10

Zs= 4q=10.5

working shifts/day=1.9

The problem can be defined in the context of multi-objectiveoptimization. A multiple attribute decision making method knownas Technique for Order Preference by Similarity to Ideal Solution(TOPSIS) is proposed to convert the multi-objective optimization

of three objectives into a single objective optimization problem.The technique is based on the concept that the chosen alternativeshould have theshortestEuclideandistance fromthe ideal solution.The ideal solution is a hypothetical solution for which all attributevalues correspond to themaximum attributevalues in the databasecomprising the satisfying solutions; the negative-ideal solution isthe hypothetical solution for which all attribute values correspondto the minimum attributevalues in theabove-mentioned database.TOPSIS, thus, gives a solution that is not only closest to the hypo-thetically best, but which is also farthest from the hypotheticallyworst. The steps involved for calculating the TOPSIS values are asfollows:

Step 1. Determination of decision matrix: A decision matrix (D)

i s a (mn) matrix whose element (xij) indicates the performance

rating of alternative i, Ai, with respect to attribute j, Xj. Hence Ai,for i=1, 2, 3, . . .,m is denoted by:

AF= (xi1, xi2, ,xin) (36)

andXjfor j= 1,2,3, . . .,n is denoted by:Then the decision problem could be shown in matrix form:

Xj= (x1j, x2j, ,xmj) (37)

(38)

Rij=xijmi=1x

2ij

(39)

Step 2. Construct the normalized decision matrix, Rij. This can berepresented as:

vij= wjrij (40)

Step 3. Construct the weighted normalized decision matrix. Thisis obtained by the multiplication of each element of the columnofthe matrix Rij with its associated weightwj. Hence, the elementsof the weighted normalized matrix vijare expressed as:Step 4. Determine the ideal (best) and negative ideal (worst)

solutions. The ideal (best) and negative ideal (worst) solution canbe expressed as:

A= {maxi(vij)ifj J;mini(vij) ifjJ, i = 1,2, . . . ,m}

Positiveideal solution =

v1 , v

1, , vj, , vn

(41)A= {mini(vij) ifjJ;maxi(vij) ifjJ, i = 1,2, . . . ,m}

Negativeideal solution =

v1 , v

1 , , v

j , , vn

(42)Step 5. Obtain the separation measures. The separation of eachalternative from the ideal one is given by Euclidean distance.Therefore, the separation from positive ideal alternative is:

Si =n

j=1(vij vi )

2

(43)

similarly, the separation from the negative ideal alternative is:

Si =

nj=1

(vij vi )2

(44)

Step 6. The relative closenessof a particular alternative to the idealsolution C

i can be evaluated as:

Ci=

Si

Si + S

i

i=

1,2, . . . ,m; 0Ci

1

(45)

8/10/2019 1-s2.0-S1526612514000462-main

12/17





Fig. 6. The overlapping circles model of sustainable development.

Step 7. Rank the preference order, so that the alternative that hasthe shortest distance to the ideal solutionis rankedfirst [17,52,53].

The data given in Table 3 are represented as matrix D484 foreach material case. The matrix is not shown here as it is nothingbut the repetition of data given in Table 3, which is represented ina matrix form. The attribute weights for machine utilization time

tbB, main time-related costs K1 and the tool related costs K3 were0.45, 0.3, 0.25, respectively. The normalized decision matrix andthen the weighted normalized matrix are determined by using Eqs.(39) and (40) respectively. The positive ideal solution (A*) and thenegative ideal solution (A) could be found by Eqs. (41) and (42).Eqs. (43) and (44) are used to determine the separation measures.Finally, Eq. (45) is utilized to calculate the relative closeness to theidealsolutionC

i. The results of performing TOPSIS aresummarized

in Table 6.The rankings of alternatives by their corresponding TOPSIS

indices (see Fig. 5.) revealed the favourability of adapting wetcutting for a simultaneous minimization approach. This is mainlyattributed to the fact that when the components are dry machinedthe cost of increasingly tool wear rate and power consumption

overtakes the advantage of not employing cutting fluids. It is alsoseen that the order of ranking of both materials is almost identical.Because of the higher effective power consumption and tool wearsrate at higher cutting speed and higher feed rate ranges and theworst chip morphology at lower ranges, an intermediate range hasgiven the preference over both ranges. It is also noticed that thelower cutting speed when dry cutting EN 1.4410 and the highercutting speed when wet cutting EN 1.4462 has given higher pref-erence than the similar dry cutting EN 1.4462 and wet cutting EN1.4410, respectively.

3.4. Phase III: Operational sustainability index (OSI)

Originally, sustainability relates to forestry. In a very broad and

fuzzy definition, forestry is called sustainable, if just as much tim-ber is cut down as can replenish to maintain the basis of life forfuture generations [54]. A more widely accepted general defini-tion of sustainable development is provided by the United NationsBrundtland Commission in 1987: development that meets theneedsof the present without compromising the ability of future generations

tomeet their ownneeds [55]. Based on this view, theUnited Nations2005 World Summit Outcome document refers to the indepen-dent and mutually reinforcing pillars of sustainable developmentas economic development, social development and environmentalprotection (see Fig. 6.) [56].

In the context of manufacturing, sustainability means the abil-ity to produce specific product operations and the circulationof resources at the rate of production. Although improving sus-

tainability through manufacturing process optimization is far from

Fig. 7. Fuzzy Inference System prediction of OSI.

simple, it is appealing because manageable number of variables,and as relatively low uncertainty can be achieved by measure-ment of local manufacturing operations [57]. In this study, theapplications of sustainabilityprinciplesin manufacturing processesare presented using machining as an example. With the imple-mentation of sustainability principles in machining technologies,end-users have the potential to reduce the cost, enhance opera-tional safety and reduce power consumption.

Operational sustainability in machiningcan be definedin termsof the cost of machining, power consumptionand chip volumeratiowhich greatly affect the waste management. Based on these threeinteracting and contradicting elements, a compromise solution hasto be introduced for a comprehensive evaluation of machiningoperational sustainability. To obtain this sustainability measure,fuzzy logic system (see Fig. 7.) is employed to combine the totalproduction cost per unit KF, effective cutting power Pe and chipvolume ratios R of each experimental trial into a single sustaina-bility characteristics index called Operational Sustainability Index(OSI). Eqs. (10)(34) and constants presented in the case study ofphase II were respectively employed to calculate the total produc-tion cost per unit KF. Matlab software was used to construct theinference model of the OSI. The three performance values were

first adjusted to a notionally common scale between null and one,using simple normalization methods, so that the digit one repre-sents the most desirable andnull is the least desirable alternative.Small (S), Medium (M) and Large (Lg) fuzzy sets are assigned tothe performances. The sustainability index has the following sevenlevels: Very Low (VL), Low (L), Lower Medium (LM), Medium (M),Upper Medium (UM), High (H), Very High (VH). Mamdani impli-cation method is employed for the fuzzy inference reasoning. Therelationship between system input and output is expressed by anIf-Thentype. Totally 27 fuzzy rulesper material wereformulated.

The predicted values of OSI are presented in Fig. 8 and the fol-lowing conclusions are extracted:

a. Generally, higher OSI values were noticed when cutting EN1.4462 is performed.

b. The average wet cutting OSI is 10% higher than dry cutting dueto the fact that the production cost, the effective cutting powerand the chip volume ratios were lower in the wet cutting.

c. Lower cutting speed, intermediate feed rate and depth of cutranges, and higher cutting speed, intermediate feed rate andlower depth of cut ranges tend to maximize OSI in dry and wetcutting, respectively.

d. OSI deteriorates with increasing number cutting passes (Lc/ap),since all non-beneficial performances are expected to increaseas Lc/apincreases.

e. For the equal cutting length scenario of facing EN 1.4462 andEN 1.4410, the maximum OSIs are obtained when alternatives

number 8 are selected.

8/10/2019 1-s2.0-S1526612514000462-main

13/17

8/10/2019 1-s2.0-S1526612514000462-main

14/17





Fig. 8. Process parameters interaction effects on OSI.

3.4.1. Optimization of OSI using CSNNS

Due to the highlycomplex relation between cutting parametersand OSI, the RSM was seen no longer efficient to accurately predictthevalues of OSI.Serious divergences werenoticed between exper-imental data and predicted values for several points. Therefore, amultilayer perceptron(MLP) artificial neural network (ANN) whichcan describe the relationships with more precision was integratedwithCuckoo Search meta-heuristicalgorithm to perform thetask ofmodeling and optimization of OSIs. The neural networks have twolayers: one hidden layer and one output layer. The hidden layeruses a sigmoid-type transference function:

f(x) = 11 + exp[(b+

wixi)]

(46)

while the output layer uses a linear function:

Outputpredicted= b+

wixi (47)

where w an b are the weights and biases of the network respec-tively. To facilitate the neural network training process, all theinputs were normalized using the following equation:

xi=2(xxmin)

(xmax

xmin

) 1 (48)

This normalization maps all the inputs and OSI between 1 and+1. The ANN architecture consists of 4 neurons in the input layer, 1neuron in the output layer. The weights and biases of the networkare initialized to small random values to avoid immediate satura-tion in the respective functions. The network was trained by usinggradient descendent withmomentum backpropagation algorithm.In this algorithm four parameters must be tuned: learning rate LR,momentum constantMc, training epochsEpand number of hiddenneurons Hn. For this tuning the Taguchi design L9(34) was used tofind the most convenient values for achieving no only lower rootmean square, but also good generalization capability, giving thefollowing values: LR =0.0705,Mc= 0.5895,Ep = 5000 andHn = 7.Theperformance of the ANNs was statistically measured by the rootmean squared error (RMSE), the coefficient of determination (R2)and the absolute average deviation (avg) obtained as follows:

RMSE=

1n

ni=1

[(yactual)i (ypredicted)i]2 (49)

R2 = 1 n

i=1[(yactual)i (ypredicted)i]2

ni=1

[(yactual)i

(ymean)i]2

(50)

8/10/2019 1-s2.0-S1526612514000462-main

15/17





Cuckoo SearchAlgorithm

MaximumOSI

Neural NetworkModel

Neural NetworkModel

Independent Variables

OSILc

vc

f

ap

Fig. 9. Cuckoo Search Neural Network System.

avg=1

n

ni=1

(yactual)i (ypredicted)i(yactual)i

100 (51)

whereyactualrepresents the actual,ypredictedobtained by neural net-work and ymean the mean of actual OSI values respectively, and n

is the number of experimentation trials. It must be remarked thatnumber of hidden neurons guarantee that there are more trainingsamples than the total amount of free parameters, thus the train-ing process is mathematically determined. TheR-squared statisticsof the models were generally greater than 0.99, which indicatethat the models as fitted explain over 99% of the variability inOSI. The trained networks achieved RSME and avgvalues belowtarget (0.0001) and (1%) respectively. It must be mentioned thatthe relationships between variables are complex, which prove theapplication of artificial neural networks very advantageous.

Neural network models are then integrated with the cuckoosearch optimization algorithm, so thatsolutions which willprovideuseful information to the user during the selection of machiningparameters are obtained. The architecture of the cuckoo search

neural network system (CSNNS) is shown in Fig. 9. CS outperformsmany existing algorithms such as genetic algorithm and particleswarm optimization. This superiority can be attributed to the factthat cuckoo search uses a combination of vectorized mutation,crossover by permutation and Levy flights and selective elitismamong the best solutions. In addition, the not-so-good solutionscan be replaced systematicallyby new solutions,and new solutionsare often generated by preferring quality solutions in the solutionsets. Thus, the mechanism of the overall searchmoves is more sub-tle and balanced, compared with the simple mechanism used inparticle swarm optimization [58]. The number of host nests (or thepopulation sizen) and the probabilities (Pa) were tuned using trialand error method. The population size of (25) and probability of(0.25) were found sufficient in this case.

Theselecteddecisionvariables werecuttingparametersfor eachprocess condition. They were defined for the ranges between theminimum and maximum experimental levels presented in Table 2.OSIs were selected as the objective functions to maximize:

OSIi= i(vc, f ,ap, Lc) (52)

Fig. 10. Maximization of OSIwith CSNNS.

where i are the neural-network-based models. The consideredconstraints are the percentage increase in radial cutting force (%Fr)and the arithmetic average roughness (Ra). The latter is defined as:

Ra=0.032f2

r (53)

where Ra is in m and r is the tool nose radius in mm. In thiscase,the optimum cutting settingsshould always lead to theresultsof preferably smaller or equal to the respective values of 15% and2m.

CSNNS optimization of OSI has to yield minimum productioncost, minimum effective cutting power and the best chip morphol-ogy, while considering technological constraints. Fig. 10 shows theperformance of proposed CSNNS. Total computation timewerelessthan 3min with an Intel Xeon CPU 3.47GHz and 24GB RAMcomputer. Less than 2000 iterations were sufficient to reach tothe global optimums in each case. It is evident that the devel-oped CSNNS is very efficient and highly reliable approach for theselection of optimum control parameters.

The obtained optimization results are listed in Table 7. Gener-ally, the following conclusion point can be depicted:

a. Higher optimum cutting speeds were more often to observe inwet cutting process than in dry cutting process which promoteshigher production rates.

b. Based on the estimated optimum OSIs, the machinability of EN1.4462 is higher than the machinability of EN 1.4410.

c. Wet cutting of EN 1.4462 and EN 1.4410 outperforms theirrespectivedry cutting in operational sustainabilityby 9.768% and12.383%, respectively.

d. While wet cutting generally gives lower non-beneficial perfor-

mancevalues,thisconclusionis hardto noticedue tothe differentadopted optimum cutting speeds and feed rates.

e. The magnitudes of surface roughness and chip volume ratio areshowing similar trends.

f. The coincidence between optimum cutting conditions of the firstrank TOPSIS and those of the optimized OSI is observed.

Table 7

Results of CSNNS.

Material Process condition vc(m/min) f(mm/rev) ap(mm) Lc(mm) %Fr Pe(W) VBmax(m) R OSI

EN1.4462

Wet 156.283 0.1359 1.500 12 6.6477 2218.1 104.19 69.073 0.7105Dry 91.793 0.1848 1.500 12 9.9772 1354.6 48.049 105.6 0.6411

EN1.4410

Wet 146.30 0.1286 1.500 12 14.878 2066.7 106.401 97.052 0.6969Dry 82.3259 0.2019 1.500 12 13.204 1.4175 62.135 100.7 0.6106

8/10/2019 1-s2.0-S1526612514000462-main

16/17





4. Conclusions

Machining of duplex stainless steel grades such as EN 1.4462and EN 1.4410 DEDM has been systematically investigated undera multi-pass constant cutting speed facing operation. In the firstphase of the investigation, D-optimal experimental design is usedextensively to investigate the effect of process variables on perfor-mance characteristics such as percentage increase in radial cuttingforces, effective cutting power, maximum tool flank wear and chipvolume ratio. Based on RSM, effective empirical relationships topredict performance characteristics at 95% confidence level weredeveloped. ANOVA used to check the adequacy of the models. Themodels were then analyzed using 3D surface graphs and used tostudy the interaction effects of process parameters. At the end ofthe first phase, constrained cuckoo search algorithm is selected toperform optimization of the performance characteristics therebydefining the optimum process conditions. The following generalconclusions can be drawn from the first phase of the investigation:

a. The values of obtained %Frwhen EN 1.4410 steels are cut werehigher than EN 1.4462 steels. They have shown direct linearcorrelations with maximum tool flank wear and proportional

dependency on Lc/apratio, and approached their minimum val-ues as cutting speed and feed rate increased to certain specifiedlimit. However, when the materials are cut dry, lower cuttingspeed and feed rates than wet cutting are preferred to minimize%Fr.

b. Generally, the two-factor interaction models were seen accurateenough to explain the dependency relation between effec-tive cutting power and independent variables. Their minimumconsumption were seen when the feed rate is in the range0.1250.175mm/rev of dry cutting and 0.10.15mm/rev of wetcutting and the rest of the remained independent variables arekept at their lowest levels.

c. Wet cutting at cutting parameter ranges ofvc100160m/min,f0.150.25mm/rev, ap0.751 mm for machining EN

1.4462 and vc120160m/min, f0.150.20mm/rev,ap1.251.5mm for machining EN 1.4462 were seen opti-mum in minimizing the tool flank wear. Dry machiningis possible when cutting parameters are set appropri-ately. The settings include; machining of EN 1.4462 atvc75110m/min, f0.150.2mm/rev, ap11.5mmand maximum Lc/ap6, and machining of EN 1.4410 atvc7590m/min, f0.1250.175 mm/rev, ap0.81.35mmand maximum Lc/ap4.

d. Ribbonand snarled chip forms were commonwhenthe DSSs aremachined at feed rate anddepthof cuts lower than 0.15 mm/revand 1mm, respectively. In wet cutting, friendlier-to-machinechip forms were produced and the R values of cutting EN 1.4410were generally seen higher than of cutting EN 1.4462. Under the

same machining conditions which might produce continuouschips, different chip forms were also possible, especially whenLc/ap6.

In the second phase of the study, the measurements were usedto develop a comprehensive machining economics model. A casestudy of producing 12000 units of per each experimental run isconsidered and the corresponding machining costs and produc-tion rates were determined. Summarizing the conclusion pointsdepicted at this stage:

a. The optimization conflict between machining economicsattributes, such as simultaneous minimization of machine uti-

lization time tbB, main time-related costs K1 and the tool

related costsK3could be effectively solved employing a multipleattribute decision making approach called TOPSIS.

b. The alternatives were ranked based on their computed relativecloseness to the ideal solution C*. The ranking of the alternativeshas revealed that the intermediate range of cutting speed, feedrates at wet cutting produce the optimum choice to minimizethe considered attributes simultaneously.

c. The order of ranking of both materials is almost identical. How-ever, under the same cutting condition, lower cutting speedwhen dry cutting EN 1.4410 and higher cutting speed when wetcutting EN 1.4462 has shown higher preference than similar drycutting EN 1.4462 and wet cutting EN 1.4410, respectively.

In the third phase of the study, the computed performances inthe first and second phases were utilized to derive a new indexof measuring machining sustainability called operational sustaina-bility index (OSI). Based on the Mamdani implication method forthe fuzzy inference reasoning, normalized production cost per unitto consider the economics of the machining process, normalizedeffectivecuttingpower to assessthe energydemandof themachin-ing process and normalized chip volume ratio to consider the chipmorphology were successfully employed to define the OSI. The

optimal machining parameters were tabulated and many conclu-sion points were extracted:

a. To accurately model and constrainedly optimize the highly non-linear OSIs, neural network models could be integrated withcuckoo search algorithm and formcuckoo search neural networksystem(CSNNS). Results have indicatedthe efficiencyof the pro-posedapproachfor solving theoptimizationproblem effectively.

b. Generally, higher OSI values were noticed when wet cutting ofDSSs areperformed. For instance, in this study,wet cutting of EN1.4462 and EN 1.4410 outperforms their respective dry cuttingin operational sustainabilityby 9.768% and 12.383%respectively.

c. Based on the predicted OSIs values, the machinability of EN1.4462 is higher than the machinability of EN 1.4410.

d. Lower cutting speed, intermediate feed rate depth of cut ranges,andhigher cuttingspeed, intermediatefeed rateand lower depthofcutrangestendtomaximizeOSIindryandwet cutting respec-tively.

e. Results also showedcoincidence between optimum cutting con-ditions in second and third phases.

Acknowledgments

The authors would like to acknowledge the significant con-tribution of the workshop master Rolf Bauer for his importantexperimental contributions as well as the many valuable sugges-tions made by Dr.Michael Schaalat theInstitute for Machine Tools,Stuttgart University.

References

[1] Armas IA, Moreuil SD. Duplex stainlesssteels. UK/USA: ISTE Ltd./JohnWiley &SonsInc.; 2009. pp.1.

[2] McGuire M. Stainless steels for designengineers. OH, USA: ASM International;2008. p. 94.

[3] Chien WT, Tsai CS. The investigation on the prediction of tool wear and thedetermination of optimum cutting conditions in machining 17-4PH stainlesssteel. J Mater Process Technol 2003;140:3405.

[4] Korkut I, Kasap M, Ciftci I, Seker U. Determination of optimum cutting param-eters during machining of AISI 304 austenitic stainless steel. Mater Des2004;25(4):3035.

[5] Endrino J, Fox GR, Gey C. Hard AlTiN, AlCrN PVD coatings for machining ofaustenitic stainless steel. Surf Coat Technol 2006;200(24):68405.

[6] MaranhoC, DavimJP. Finite element modellingof machiningof AISI316 steel:numerical simulation and experimental validation. Simul Model Pract Theory2012;18(2):13956.
http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0030http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0025http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0020http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0015http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0010http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0005http://localhost/var/www/apps/conversion/tmp/scratch_8/dx.doi.org/10.1016/j.jmapro.2014.05.004

8/10/2019 1-s2.0-S1526612514000462-main

17/17




[7] Sahoo AK, Sahoo B. Experimental investigations on machinability aspects infinish hard turning of AISI 4340 steel using uncoated and multilayer coatedcarbide inserts. Measurement 2012;45(8):215365.

[8] AhmadiE, HomamiRM, Rahmati S. Experimental investigation and mathemat-ical modeling of thecomposite ceramic cutting tools with alumina base in themachining process of PH-hardened austenitic-ferritic (Duplex) stainless steel.IntJ Adv Des Manuf Technol 2012;5(2):1725.

[9] Kurama E, Ozcelik B, Bayramoglu M, Demirbas E, Simsek BT. Optimization ofcutting fluids and cutting parameters during end milling by using D-optimaldesign of experiments. J Clean Prod 2013;42:15966.

[10] Iqbal A, Ning H, Khan I, Liang L, DarNU. Modeling theeffectsof cutting param-

eters in MQL-employed finishhard-milling process using D-optimal method. JMater Process Technol 2008;199(13):37990.

[11] Arokiadass R, Palaniradja K, Alagumoorthi N. Prediction and optimization ofend milling process parameters of cast aluminium based MMC. Trans Nonfer-rous Met Soc China 2012;22(7):156874.

[12] KaoP, Hocheng H. Optimization of electrochemical polishing of stainless steelby grey relational analysis. J Mater Process Technol 2003;140(1):2559.

[13] Walia R, Shan H, Kumar P. Multi-response optimization of CFAAFM pro-cess through Taguchi method and utility concept. Mater Manuf Process2006;21(8):90714.

[14] Chang CK, Lu HS. Design optimization of cutting parameters for side millingoperations with multiple performance characteristics. Int J Adv ManufTechnol2007;32(12):1826.

[15] Gaitonde V, Karnik S, Achyutha B, Siddeswarappa B. Taguchi optimizationin drilling of AISI 316L stainless steel to minimize burr size using multi-performance objective basedon membershipfunction. J MaterProcess Technol2008;202(13):3749.

[16] Gaitonde V, Karnik S, Davim JP. Multiperformance optimization in turning offree-machining steel using taguchi method and utility concept. J Mater Eng

Perform 2009;18(3):2316.[17] SinghA, DattaS, MahapatraSS.Applicationof TOPSISin theTaguchimethodfor

optimal machiningparameterselection.J ManufSci Prod2011;11(13):4960.[18] Kaladhar M, Subbaiah KV,Rao CS,Rao KN.Applicationof taguchi approach and

utility concept in solving the multi-objective problem when turning AISI 202austenitic stainlesssteel. J Eng Sci Technol Rev 2011;4(1):5561.

[19] Vijayakumar K, Prabhaharan G, AsokanP, Saravanan R. Optimization of multi-pass turning operations using ant colony system. Int J Mach Tools Manuf2003;43(15):16339.

[20] Samanta S, Chakraborty S. Parametric optimization of some non-traditionalmachining processesusing artificial bee colony algorithm.Eng Appl ArtifIntell2011;24(6):94657.

[21] Zarei O, Fesanghary M, Farshi B, Saffar RJ, Razfar MR. Optimization of multi-pass face-milling via harmony search algorithm. J Mater Process Technol2009;209(5):238692.

[22] DAddona DM, Teti R. Genetic algorithm-basedoptimization of cutting param-eters in turning processes. Procedia CIRP 2013;7:3238.

[23] Yildiz AR. Hybrid Taguchi-differential evolution algorithm for optimization of

multi-pass turning operations. Appl Soft Comput 2013;13(3):14339.[24] Rao RV, Kalyankar VD. Multi-pass turning process parameter opti-mization using teachinglearning-based optimization algorithm. Sci Iran2013;20(3):96774.

[25] TzengY-C, ChenF-C. Multi-objective optimisation of high-speed electricaldis-charge machining process using a Taguchi fuzzy-based approach. Mater Des2007;28(4):115968.

[26] Gupta A, Singh H, Aggarwal A. Taguchi-fuzzy multi output optimization(MOO) in high speed CNC turning of AISI P-20 tool steel. Expert Syst Appl2011;38(6):68228.

[27] Yang YS, Huang W. A grey-fuzzy Taguchi approach for optimizing multi-objective properties of zirconium-containing diamond-like carbon coatings.Expert Syst Appl 2012;39(1):74350.

[28] Chien WT, Chou CY. The predictive model for machinability of 304 stainlesssteel. J Mater Process Technol 2001;118:4427.

[29] Karpat Y, zel T. Multi-objective optimization for turning processes usingneural network modeling and dynamic-neighborhood particle swarm opti-mization. IntJ Adv Manuf Technol 2007;35(34):23447.

[30] Karnik SR, Gaitonde VN, Davim JP. A comparative study of the ANN and RSM

modeling approaches forpredicting burr size in drilling. IntJ Adv Manuf Tech-nol 2008;38(910):86883.

[31] Paro J, Hnninen H, Kauppinen V. Tool wear and machinability of HIPed P/Mand conventional cast duplex stainless steels. Wear 2001;249(34):27984.

[32] Bordinassi E, Stipkovic M, Batalha G, Delijaicov S, de Lima NB. Superficialintegrityanalysisin a super duplex stainlesssteel after turning. J AchievMaterManuf Eng 2006;18(12):3358.

[33] Krlczyk G, Legutko S, Gajek M. Predicting the surface roughness in the drymachining of duplex stainless steel (DSS). Metalurgija 2013;52(2):25962.

[34] Krlczyk G, Gajek M, Legutko S. Effect of the cutting parameters on tool life induplex stainless steel turning process. Tehnicki vjesnik 2013;20(4):58792.

[35] Krlczyk G, Gajek M, Legutko S. Predicting thetool life in thedry machining ofduplex stainless steel. Maint Reliab 2013;15(1):625.

[36] Nomani J, PramanikA, HilditchT, LittlefairG. Machinability studyof firstgener-ation duplex (2205), second generation duplex (2507) and austenite stainlesssteel during drilling process. Wear 2013;304(12):208.

[37] deOliveiraJr CA,Diniz AE,Bertazzoli R.Correlatingtool wear,surface roughnessand corrosionresistance in theturning process of super duplexstainless steel.

J Braz Soc Mech Sci Eng 2013, http://dx.doi.org/10.1007/s11665-013-0832-4.

[38] Selvaraj DP, Chandramohanb P, Mohanraj M. Optimization of surface rough-ness, cutting force andtool wear of nitrogen alloyed duplexstainless steel in adry turning process using Taguchi method. Measurement 2014;49:20515.

[39] Krlczyk G, Nieslony P, Legutko S. Microhardness and surface integrity in tur-ning process of duplex stainless steel (DSS) for different cutting conditions. JMater Eng Perform 2014;23(3):85966.

[40] Marksberry PW,Jawahir IS. A comprehensive tool-wear/tool-life performancemodel in the evaluation of NDM (near dry machining) for sustainable manu-facturing. IntJ Mach Tool Manuf 2008;48(78):87886.

[41] Jayal AD, Badurdeen F, Dillon OW,Jawahir IS. Sustainable manufacturing: mod-eling and optimization challenges at the product, process and system levels.CIRP J Manuf Sci Technol 2010;2(3):14452.

[42] Heisel U, Stehle T. Energy efficiency in metal cutting production. Int Sci-Convenient Conf Mod Eng Sci Edu 2011;5(1):38598.

[43] Hanafi I, Khamlichi A, Cabrera FM, Almansa E, Jabbouri A. Optimizationof cut-ting conditionsfor sustainable machiningof PEEK-CF30using TiN tools. J CleanProd 2012;33:19.

[44] Heisel U, Braun S. Prognose des prozessabhngigen EnergieverbrauchsSimulations- and Prognosemodell des prozessabhngigen Energiever-brauchs von Werkzeugm-aschinen Teil 1. Werkstattstechnik Online2013;103(1):1621.

[45] Anderson-CookCM, Borror CM,Montgomery DC.Responsesurfacedesigneval-uation and comparison. J Stat Plan Inference2009;139:62941.

[46] Klocke F. Manufacturing processes 1, cutting. Heidelberg, Germany: Springer-Verlag; 2011. p. 33951.

[47] Myers RH, Montgomery DC, Anderson-Cook CM. Response surface method-ology process and product and product optimization using designedexperiments. New Jersey, USA: John Wiley & Sons; 2009.

[48] Yang XS. Nature-inspired metaheuristic algorithms. UK: Luniver Press; 2010.p. 6583.

[49] Rajabioun R. Cuckoo optimization algorithm. Appl Soft Comput

2011;11(8):550818.[50] Tschtsch H. Applied machining technology. Heidelberg, Germany: Springer;2008. p. 702.

[51] Paucksch E, Holsten S, Lin M, Tikal F. Zerspantechnik. Wiesbaden, Germany:Vieweg Teubner Verlag; 2008. p. 936.

[52] Lai YJ, HwangCL. Fuzzymultipleobjectivedecisionmakingmethodsand appli-cations. Berlin: Springer-Verlag; 1994. p. 119.

[53] Chittaranjan D, Srinivas C. Evaluation of metalstrip-layout selectionusing AHPand TOPSIS method. Adv Mater Manuf Charact 2013;3(1):4259.

[54] Kandisch D, Hermann P, MeierC. Sustainableprocessmanagement-status Quoand perspectives. In: Advanced Manufacturing and Sustainable Logistics: 8thInternational Heinz Nixdorf Symposium, 44. 2010. p. 94107.

[55] UN World Commission on Environment and Development. Our CommonFuture. Oxford: Oxford University Press; 1987.

[56] Chick A, Micklethwaite P. Design for sustainable change: how design anddesignerscan drive thesustainability agenda. China:AVA Publishing; 2011.

[57] Nahkala S. Aligning product design methods and tools for sustainability. In:Re-engineering Manufacturing for Sustainability:Proceedings of the 20th CIRPInternational Conference on Life Cycle Engineering. 2013. p. 539.

[58] Yang XS,Deb S. Multiobjective cuckoosearch fordesign optimization. ComputOper Res 2013;40(6):161624.
http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0035http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0035http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0035http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0035http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0040http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0045http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0045http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0045http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0045http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0045http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0050http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0050http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0050http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0050http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0050http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0055http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0060http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0060http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0060http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0060http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0065http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0065http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0065http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0065http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0070http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0070http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0070http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0070http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0070http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0075http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0080http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0080http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0080http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0080http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0085http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0085http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0085http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0085http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0090http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0090http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0090http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0090http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0095http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0095http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0095http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0095http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0095http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0100http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0100http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0100http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0100http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0105http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0105http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0105http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0105http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0105http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0110http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0110http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0110http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0110http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0115http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0115http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0115http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0120http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0120http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0120http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0120http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0120http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0125http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0125http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0125http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0125http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0130http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0130http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0130http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0130http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0135http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0135http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0135http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0135http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0140http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0140http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0140http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0140http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0145http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0145http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0145http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0145http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0145http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0150http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0150http://refhub.elsevier.com/S1526-6125(14)00046-2/sbref0150http://refhub.elsevier.com/S1526-6125(1

1-s2.0-s1526612514000462-main

Documents