JOURNAL OF TEXTILES AND POLYMERS, VOL. 6, NO. 1, JANUARY

Transcription

1 JOURNAL OF TEXTILES AND POLYMERS, VOL. 6, NO., JANUARY Mult-Objectve Optmzaton of Rotorcraft Compact Spnnng Core-Spun Yarn Propertes Parvaneh Kherkhah Barzok, Morteza Vadood *, and Majd Safar Johar Abstract- One way to mprove the propertes of staple yarns s to employ core compact yarn spnnng system. Ths type of yarn s used n a wde range of applcatons and up to now many researchers have studed ts producton process and propertes. However, there s a lack of researches regardng the optmzaton of the propertes of rotorcraft compact spnnng (RoCos) core-spun yarns based on the spnnng parameters. Therefore, n ths paper, the nfluence of some spnnng parameters ncludng the pre-tenson of flament, yarn count and type of sheath fber on the propertes of RoCos core-spun yarns was nvestgated. To acheve the goals of ths research, the physcal and mechancal propertes of RoCos core-spun yarns ncludng the tenacty, harness and abrason resstance were measured, and then modeled by artfcal neural network (ANN). Fnally, to optmze all measured propertes at the same tme the ANN models and non-domnated sortng genetc algorthm (NSGAII) method were appled as a hybrd model. The results showed that the presented method could be successfully used to determne the spnnng parameters to produce RoCos yarns wth desred propertes. The optmzed values of harness, tenacty and abrason resstance for an deal yarn were observed at yarn count of 4.5 tex, flament pre-tenson of 25 g and for sheath fber of vscous/polyester. Keywords: rocos, core-spun yarn, artfcal neural network, non-domnated sortng genetc algorthm, mult-objectve optmzaton I. INTRODUCTION Core spnnng system has two dstnct categores: core and sheath. The core s usually made of flament yarns and the sheath part s made of staple fbers. The man am P. Kherkhah Barzok and M. Safar Johar Department of Textle Engneerng, Amrkabr Unversty of Technology, Tehran, Iran. M. Vadood Department of Textle Engneerng, Yazd Unversty, Yazd, Iran. Correspondence should be addressed to M. Vadood e-mal: mortezavadood@yazd.ac.r of usng the core-spun yarns s to mx the advantage of dfferent propertes of both components. Core yarns are used n dfferent felds and up to now some studes have been accomplshed about core yarns producton system [- 5]. Bhatnagar [6] studed the effects of twst, pre-tenson, and feed postons of the core flament on propertes of corespun yarns. Yuan et al. [7] studed the effects of changng compound spnnng condtons such as the tenson rato between flaments and staple fbers, and poston where the flaments were fed nto the staple fbers. These knds of yarns have varous propertes whch can be predcted wth dfferent models. In recent decades, ANN models have been employed n many publshed works to predct the propertes of varous yarns and many other characterstcs of textle materals precsely [8-]. On the other hand, n many engneerng problems, objectves under consderaton conflct wth each other, so the optmzaton of a partcular soluton wth respect to a sngle specfed objectve may cause to obtan unacceptable results regardng the other objectves. A reasonable soluton to a mult-objectve problem s to nvestgate a set of solutons, each of whch satsfes the objectves at an acceptable level wthout beng domnated by any other soluton [2]. Mult-objectve formulatons are realstc models for many complex engneerng optmzaton problems. Customzed genetc algorthms have been demonstrated to be partcularly effectve to determne excellent solutons to these problems [3-8]. In case of textle engneerng, Chen et al. [9] used mult-objectve optmzaton method for carbon fber drawng process. They proposed a new synergetc mmune clonal selecton algorthm (SICSA) to obtan the optmal process parameters, such as lnear densty, tenacty, and breakng elongaton rato. Han and Wang [20] proposed mult-objectve genetc algorthm based on the numercal smulaton of the polymer flow to optmze the geometry parameters of the coat-hanger de wth unform outlet velocty and mnmal resdence tme. Gu [2] dscussed about the feasblty of adoptng multple objectve

2 48 JOURNAL OF TEXTILES AND POLYMERS, VOL. 6, NO., JANUARY 208 optmzaton to desgn blended fabrc. In ths research, he llustrated that multple objectve optmzaton s feasble for desgnng textle proucts. Recently Majumdar et al. [22] derved a smultaneous optmal soluton of two objectves, namely ar permeablty and thermal conductvty for both sngle jersey and rb kntted fabrcs wth desred ultravolet (UV) protecton. In ths work, Pareto-optmal solutons were derved by an eltst mult-objectve evolutonary algorthm based on non-domnated sortng genetc algorthm (NSGAII), and the effectve knttng and yarn parameters for engneer kntted fabrcs wth optmal comfort and desred level of UV protecton were obtaned. Due to the lack of nformaton on predcton and mult optmzaton about the RoCos (Rotorcraft Compact Spnnng) core-spun yarns, n present study the relaton between the controllable factors and physcal and mechancal propertes of RoCos yarns was modeled, and then the best controllable factors for an deal yarn were determned usng mult optmzaton method. To ths am, usng ANN the tenacty, abrason resstance and harness of yarn were modeled, and NSGAII technque of multobjectve optmzaton has been mplemented wth the am to maxmze the tenacty and abrason resstance and to mnmze the harness, smultaneously. II. ANN ANN models are very useful n modelng nonlnear and complex systems. ANN ncludes three layers namely: nput, hdden and output layers. Neurons n each layer are connected by assocated weghts to other neurons n the next layer. The nput data s receved n nput layer and the output s obtaned n the output layer by mathematcal operatons through hdden layers. The number of hdden layers and neurons n each hdden layer are the most mportant parameters n ANN model [23]. III. NSGAII Generally, n the method of mult-objectve optmzaton, there are some optmal solutons known as Pareto-optmal or non-domnated solutons, so that each soluton can be consdered as a response. The Pareto-optmal solutons plotted regardng the related objectves are called Paretooptmal fronts. NSGAII s one of the most used methods for mult optmzaton and ts purpose s to fnd a set of solutons as much as close to Pareto-optmal front smultaneously wth maxmum varety. All operatons n NSGAII are the same as those n genetc algorthm (GA) ncludng crossover and mutaton, except for the ftness assgnment method whch s modfed by fast non-domnaton sortng and crowdng dstance sortng [7, 24, 25]. A. Fast Non-Domnaton Sortng The populaton generated by GA s sorted nto a herarchy of subpopulatons based on non-domnaton method. In ths method each objectve for a soluton s compared wth the correspondng objectves n other solutons as follows (for example a soluton wth two objectves): O S > O S or O S O S j j O2 O2 O2 > O2 Where, S s the soluton, O s the objectve, and ndces and j are the th and j th solutons, respectvely. Based on the upper constrant j th soluton s domnated by th one, otherwse t s non-domnated. Therefore, the whole populaton s dvded nto dfferent ranks such as solutons n Rank whch are better than ones n Rank 2 and solutons n Rank 2 whch are better than ones n Rank 3 and so on. B. Crowdng Dstance After sortng the populaton, the crowdng dstance s assgned to each ndvdual for each rank. To ths am, regardng each objectve, the ftness values of objectve functons for solutons are sorted n descendng order. The crowdng dstance operator helps n dstrbutng the solutons unformly to the ranks rather than bunchng up at several good ponts. The ndvduals of the next generaton are selected usng tournament selecton functon based on the rank and crowdng dstance. In tournament functon two ndvduals are compared and that wth better rank s selected, and f two ndvduals have the same rank one wth hgher crowdng dstance s selected. IV. MATERIAL AND METHOD In ths study, 56 dfferent types of yarn samples were produced on a compact-core spnnng system. A blended () TABLE I PROPERTIES OF SHEATH FIBERS Elongaton (%) Tenacty (cn/tex) Percentage of fber blendng (%) Dener Length (mm) Type of sheath fber Vscose Polyester Cotton

3 KHEIRKHAH BARZOKI et al.: MULTI-OBJECTIVE OPTIMIZATION OF ROTORCRAFT COMPACT * Traveler type: J Settng parameters TABLE II MACHINE PARAMETERS Value Twst per meter 900 Spndel speed (rpm) Rng dameter (mm) 36 Traveler* ISO No. 50 Fg.. Rotorcraft compact spnnng roller (RoCoS). vscose/polyester and cotton fbers were used as the sheath fber wth propertes presented n Table I. Multflament nylon (30 monoflaments: elongaton-at-break and tenacty of monoflament were 26% and 4600 cn/tex, respectvely) wth count of 00 dener was used as the core flament. The cotton and vscose/polyester rovng count were 0.72 and.09 Ne, respectvely. To produce the samples, RoCos system was nstalled on SKF lab spnner nstead of delvery top roller (Fg. ). In order to produce dfferent types of compact-core yarns, core flament should be pre-drawn before enterng the front rollers (RoCos roller). Flament also should be fed to compactor groove of RoCos roller and to ths am, a gude rod and a pre-tensoner were used. Fg. 2 llustrates the process of core-compact yarn producton and Table II exhbts the machne parameters for producng RoCos yarns. In ths study, the tenacty, harness and abrason resstance were consdered as the physcal and mechancal propertes of compact-core yarns, and to determne the effect of controllable factors on these propertes, 7 levels of flament pre-tenson (25, 50, 75, 00, 25, 60 and 80 g), 4 levels of yarn count (4.5, 43.5, 48 and 59 tex) and two knds of sheath fbers (cotton and polyester/vscose) were chosen. The lst of yarn samples and consdered controllable factors are presented n Table III. The Instron testng machne (Model: M0-820-) was used to measure the tenacty of yarns wth a gauge length of 25 cm accordng to ASTM D2256, and the speed of testng was selected accordng to the breakng tme of 20 s. For measurng harness (number of hars longer than or equal to 3 mm), a Shrley harness tester (Model SDL096/8) was used based on ASTM D5647. The measurement was carred out on 20 m of each yarn sample at the speed of 60 m/mn. Abrason resstance was determned by a Shrley abrason tester (Model: Y027) accordng to ASTM D66. In order to determne the tenacty, each test was repeated 0 tmes, and for abrason resstance and harness t was 5 tmes. All experments were conducted at the condton of 30 ºC and 60 RH%. V. RESULT AND DISCUSSION A. ANN Model To fnd the best set of ANN parameters to predct each property, the tral and error method was appled. Regardng the lterature revew, the number of hdden layers and neurons n each hdden layer were consdered between to 3 and to 0, respectvely. The actvaton functons for all hdden and output layers were consdered tangent hyperbolc and lnear functons, respectvely. Obvously, the nput parameters to ANN models are the pre-tenson of flament, yarn count and the knd of sheath fbers. The ANNs were traned wth the error back propagaton algorthm usng Tranlm functon. 20% of data for test, 20% for valdaton and the rest were selected for tranng set, randomly. As the ntal weghts n ANN were selected randomly, each ANN topology was created fve tmes and the best obtaned result was consdered for that topology. To evaluate the accuracy of the created ANNs, besdes the correlaton coeffcent (R-value) two more ndexes namely MSE (mean square error) and MAPE (mean absolute percentage error) were calculated between ANN outputs and correspondng actual values for testng set (Eqs. (2) and (3)). Fg. 2. Producton of core-compact yarns. y x n MAPE = 00 n = x n MSE = (y x ) n = 2 (2) (3)

4 50 JOURNAL OF TEXTILES AND POLYMERS, VOL. 6, NO., JANUARY 208 No. Flament pre-tenson (g) Yarn count (tex) TABLE III LIST OF YARN SAMPLES AND CONTROLLABLE FACTORS Type of sheath fber No. Flament pre-tenson (g) Yarn count (tex) Type of sheath fber Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Vscose/polyester Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Cotton Where, y and x are the ANN output and correspondng actual values, respectvely. The obtaned results revealed that the use of MAPE ndex leads to hgher accuracy n predcton. Table IV. shows the best topology of ANNs for each measured propertes and Fg. 3 depcts the ANN outputs based on the MAPE ndex along wth correspondng actual values n testng group for dfferent propertes. B. NSGAII algorthm After determnng the best topology of ANN for each property, the value of tenacty and abrason resstance and harness can be predcted accurately by nsertng the pretenson of flament, yarn count and the knd of sheath fber to the ANN models. Therefore, at ths step, the optmzaton of all propertes at the same tme s possble. Clearly, for an deal yarn the tenacty and abrason resstance should be maxmum and harness should be mnmum. Consderng that GA mnmzes the ftness functon, n ths step three objectves are defned as follows (Eq. (4)) and each soluton contans all objectves. Objectve =Tenasty value Objectve =abrason resstance 2 value Objectve 3 = Harness value (4) TABLE IV THE BEST TOPOLOGY OF ANN FOR EACH INVESTIGATED PROPERTIES BASED ON MAPE Measured propertes Hdden layer (best topology of ANN) Accuracy ndex between ANN output wth best topology and correspondng actual values for testng group MAPE MSE R-value Tenacty [8 3]* Harness [6 6] Abrason resstance [2 4 4]

5 KHEIRKHAH BARZOKI et al.: MULTI-OBJECTIVE OPTIMIZATION OF ROTORCRAFT COMPACT... 5 Fg. 4. Result of mplementaton NSGAII (the optmal Pareto front for measured propertes). Fg. 3. ANN outputs based on the MAPE ndex along wth correspondng actual values for dfferent propertes (testng group). resstance ncrease, the harness decreases. In ths study, the Pareto-optmal front contans 56 solutons for RoCos core-compact yarn. It must be noted that the correspondng values of the flament pretenson, yarn count and knd of sheath fbers for each pont n Fg. 4 are presented by NSGAII. As all the solutons n the Pareto front are better than all those n the others, at least n terms of one objectve, any one of them s an acceptable soluton. The choce of one soluton over other depends completely on the end user requrements. For better understandng the relaton between three optmzed mentoned propertes, the Paretooptma fronts are llustrated n 2D-dmensonal fgures (Fgs. 5 to 7). In Fgs. 5 and 6, the trends between the parameters are very clear; one s decreasng whle the other s ncreasng. However, a specfed trend s not observed n Fg. 7, because NSGAII starts wth randomly generated 00 ntal populatons and t ranks the ndvduals based on the domnance. The fast non-domnated sortng procedure fnds out the non-domnaton ranks where ndvduals of a partcular rank are non-domnated by any soluton. In the next step, the crowdng dstance s calculated for each ndvdual. NSGAII steps are repeated and the Paretooptmal front s obtaned at the end of 00 generatons leadng to the fnal set whle the crossover and mutaton rate are 0.8 and 0.2, respectvely. Fg. 4 shows the 3D scattered Pareto-optmal front for tenacty, harness and abrason resstance. The coordnates of some ponts on the axes are dsplayed for better comprehenson of pont s poston. As can be seen n Fg. 4, whle the tenacty and abrason Fg. 5. Optmal Pareto front for harness and tenacty for RoCos corecompact yarn. TABLE V CHARACTERISTICS OF OPTIMUM YARN Property Measured value Observed values for Optmum pont Tenacty (cn/tex) Abrason resstance (cycle) Harness (har/m)

6 52 JOURNAL OF TEXTILES AND POLYMERS, VOL. 6, NO., JANUARY 208 easly wthout need to use tme-consumng and costly experments. Fg. 6. Optmal Pareto front for harness and abrason resstance for RoCos core-compact yarn. the tenacty and abrason resstance should be maxmzed smultaneously. By obtanng the Pareto-optma solutons, the relaton between yarns characterstcs becomes clear and a yarn wth desred values of harness, tenacty and abrason resstance can be produced wth a sutable combnaton of flament pre-tenson, type of sheath fbers and yarn count. For example, a pont has been ndcated n Fg. 4 as Optmum pont. Ths pont has been hghlghted n Fgs. 5 to 7, too. As can be seen, ths pont enjoys from hgher values of tenacty and abrason resstance, whle t has low value of harness. Accordng to the results obtaned from Table III, there s a yarn wth characterstcs very close to the optmum pont characterstcs (Table V). The spnnng parameters for ths yarn were count of 4.5 tex, flament pre tenson of 25 g and sheath fber of polyester/ vscous. So, for an deal yarn n RoCos spnnng system, these parameters can be consdered for spnnng. The method appled usng NSGAII s effectve, less subjectve, more practcal and computatonally effcent. The result of ths method s a set of trade off solutons, and engneers can select ther own desred soluton Fg. 7. Optmal Pareto front for abrason resstance and tenacty for RoCos core-compact yarn. VI. CONCLUSION In ths paper, a database was created by consderng the nfluence of spnng parameters such as flament pre-tenson, yarn count and type of sheath fber on the tenacty, harness and abrason resstance of compact-core spun yarns.then, the relaton between the spnng parameters and measured propertes was modeled usng ANN, accurately. In the next step, to optmze measured propertes smultaneously, NSGAII was appled by the help of ANN models obtaned. The results llustrated the relatons between the measured propertes, therefore, the spnng parameters for a core yarn wth desred propertes could be determned easly. Ths method s effectve, less subjectve, more practcal and not tme-consumng, and can be used to predct the nput parametrs for a product wth desred propertes nstead of dffcult and expensve expermental testng. REFERENCES [] Z. Xa, X. Wang, W. Ye, H.A. Eltahr, and W. Xu, Fber trappng comparson of embeddable and locatable spnnng wth srofl and sro core-spnnng wth flute ppe ar sucton, Text. Res. J., vol. 82, no. 2, pp , 202. [2] J. Chen, Q. Xu, Z. Xa, and W. Xu, An expermental study of nfluence of flament and rovng locaton on yarn propertes durng embeddable and locatable spnnng, Fber. Polym., vol. 3, no. 9, pp , 202. [3] A. Pourahmad and M.S. Johar, Comparson of the propertes of rng, solo, and sro core-spun yarns, J. Text. Inst., vol. 02, no.6, pp , 20. [4] A. Sawhney, K. Robert, G. Ruppencker, and L. Kmmel, Improved method of producng a cotton covered/polyester staple-core yarn on a rng spnnng frame, Text. Res. J., vol. 62, no., pp. 2-25, 992. [5] A. Sawhney, G. Ruppencker, L. Kmmel, and K. Robert, Comparson of flament-core spun yarns produced by new and conventonal methods, Text. Res. J., vol. 62, no. 2, pp ,992. [6] N. Balasubramanan and V. Bhatnagar, The effect of spnnng condtons on the tensle propertes of corespun yarns, J. Text. Inst., vol. 6, no., pp , 970. [7] X. Yuan, S. Mngbao, and S. Shyuan, Text. Inst. 83 rd World Conference, Shangha, Chna, [8] H. Hasan and M. Shanbeh, Applcaton of multple lnear regresson and artfcal neural network

7 KHEIRKHAH BARZOKI et al.: MULTI-OBJECTIVE OPTIMIZATION OF ROTORCRAFT COMPACT algorthms to predct the total hand value of summer kntted T-shrtsˮ, Ind. J. Fbre. Text. Res., vol. 35, no. 3, pp , 200. [9] E. Naghashzargar, D. Semnan, S. Karbas, and H. Nekoee, Applcaton of ntellgent neural network method for predcton of mechancal behavor of wre-rope scaffold n tssue engneerng, J. Text. Inst., vol. 05, pp , 204. [0] M. Vadood, D. Semnan, and M. Morshed, Optmzaton of acrylc dry spnnng producton lne by usng artfcal neural network and genetc algorthm, J. Appl. Polym. Sc., vol. 20, no.2, pp , 20. [] V. Ghorban, M. Vadood, and M.S. Johar, Predcton of polyester/cotton blended rotor-spun yarns harness based on the machne parameters, Ind. J. Fbre. Text. Res., vol. 4, no., pp. 9-25, 206. [2] X.-S. Yang, Artfcal ntellgence, evolutonary computng and meta-heurstcs: n the footsteps of Alan Turng, Sprnger, 202. [3] K. Deb, Mult-objectve optmzaton usng evolutonary algorthms, John Wley & Sons, 200. [4] K. Deb, Search methodologes, Sprnger,?????, 204. [5] R.T. Marler and J.S. Arora, Survey of mult-objectve optmzaton methods for engneerng. structural and multdscplnary optmzaton, Struct. Multdscp. Opt., vol. 26, no. 6, pp , [6] A. Konak, D.W. Cot, and A.E. Smth, Multobjectve optmzaton usng genetc algorthms: A tutoral, Relab. Eng. Syst. Safe., vol. 9, no. 9, pp , [7] K. Deb, A. Pratap, S. Agarwal, and T. Meyarvan, A fast and eltst multobjectve genetc algorthm: NSGA-II, IEEE T. Evolut. Comput., vol. 6, no. 2, pp , [8] B. Najlaw, M. Nejlaou, Z. Aff, and L. Romdhane, Mult-objectve desgn optmzaton of the NBTTL mechansm, n Advances n Acoustcs and Vbraton, vol. 5, Swtzerland, Sprnger, 207, pp [9] J. Chen, Y. Dng, Y. Jn, and K. Hao, A synergetc mmune clonal selecton algorthm based multobjectve optmzaton method for carbon fber drawng process, Fber. Polym., vol. 4, no.0, pp , 203. [20] W. Han and X. Wang, Mult-objectve optmzaton of the coat-hanger de for melt-blowng process, Fber. Polym., vol. 3, no. 5, pp , 202. [2] L.H.G.J.H. Gu, Cotton Text. Tech., vol. 6, pp. 00, [22] A. Majumdar, P. Mal, A. Ghosh, and D. Banerjee, Mult-objectve optmzaton of ar permeablty and thermal conductvty of kntted fabrcs wth desred ultravolet protecton, J. Text. Inst., vol. 08, no., pp. 0-6, 207. [23] D. Semnan and M. Vadood, Improvement of ntellgent methods for evaluatng the apparent qualty of kntted fabrcs, Eng. Appl. Artf. Intell., vol. 23, no. 2, pp , 200. [24] D.E. Golberg, Genetc algorthms n search, optmzaton and machne learnng, Addon wesley, 989. [25] S. Sette and L. Langenhove, Optmsng the fbreto-yarn producton process: fndng a blend of fbre qualtes to create an optmal prce/qualty yarn, Autex Res. J., vol. 2, no. 2, pp , 2002.