Indian J ournal of Fibre & Text i le Research Vol . 3 1 . December 2006, pp. 50 1 -506

Predicting abrasion behaviour of chenille fabric by fuzzy logic

E K (even & b bzdemir" Texti le Engi neering Department, Faculty o f Engineering and Architecture, Uludag University. Gorukle, Bursa 1 6059, Turkey

and

L Dagkurs Vocational School o f Textile, Gazioslllanpa�a U niversity. Ta� I J( ,: i ft l ik. Toka! 60240. Turkey

Received 29 Septelllber 2005; revised received 26 Decelllber 2005; accepted 2 Fe/Jrt/{//Y 2006

The effects of chcn i l l e yarn count. p i le length and yarn twist level on abrasion res istance of chen i l le fabric have been studied using fuzzy logic system. Di fferent chen i l le yarns arc produced with varying yarn count, pi le length and twist levcl on a cheni l l e yarn machine. The v i scose and acryl ic have been used as p i le and core yarns respect ively and the fabrics arc woven from these yarns by us ing them as fi l l i ng yarns i n the weaving construction. Abrasion rcsistancc of chen i l le fabrics is measured with a M art indale abrasion tester. Experimental data arc used i n the establishment o f the fuzzy logic model and the construction o f basic princi ples. I t is observed that the use o f high twist leve ls and pi le lengths brings about an improvement in abrasion resistance and the yarn count has a signi ficant efrect on mass loss. Correl ation analysis (r = 0.978)

confirms strong l inear relationsh i p between measured :lIld predicted mass loss val ues. It i s practical ly poss ible to obtain positive results wi lh ahrasion resistance optimization in a more economical way by fuzzy logic system.

Keywol'(ls : Abra,llIn res istance, Chen i l le yarn, Fuzzy logic, P i le length. Twist level, Yarn count

IPC Code: Int . CLK D02G3/42. D03 D27/00, G06N7/02

1 Introduction Chen i l le yarn is a kind of fancy yarns which i s

fascinati ng about i t s texture and gleam. These yarns are produced with the in tention of producing an enhanced aesthetic impression . They fi nd a wide range of application area i ncluding outerwear fabrics. furnishing fabrics. trimmi ngs and kni twear.

Cheni l le yarns have a pi le protruding all around at right angles. These are constructed by twisting core yarns together in cheni l le yarn machines where pile yarns are i nserted at right angles and cut to within I or 2 mm length to create a surface in which the fibres contained in the pile yarns burst and form a soft pi le surface to the yarn . The basic structure of a chen i l le yarn is shown in Fig . 1 .

The core yarn gi ves the strength to cheni l ie and constitutes 25-30 % of the relative mass. The pi Ie yarn is responsible for the body and bulk of the yarn and constitutes 70-75 {Iii} of the relat ive mass. The pi le yarn should provide the desired aesthetic effect for the resulting chen i l le. I The count and number of the pile yarns and how much of them are fed onto the core determine the count of the cheni l le yarn.2 4

' To whom a l l the correspondence shou ld be addressed. E-ma i l : ozdclll ir@u l udag.edu.tr

Cheni l l e is a diffi c u l t yarn to manufacture,

requir ing great care i n production and conversion

i n to fi nal art ic les . Due to the instabi l i ty of i ts construction, chen i l le i s susceptible to wear in those areas of the garment where abrasion occurs. such as the underarm, neckl ine, and waist. Any removal of the pi Ie yarn forming the beard. either dur ing further processing or dur i ng the eventual end-use. wil l expose

Pile yarns

(a)

Fig. I -Chen i l le yarn structure I (a ) cheni l le yarn components and (b) cheni l le yarn picture]

502 INDIAN J . FIBRE TEXT. RES. , DECEMBER 2006

the ground yarns, which, in turn, wi ll resul t in a bare appearance.5

Despite the fact that chenille yarns are used to produce special fabrics with high added value, there i s l imited research available on the abrasion behaviour of such yarns and fabrics. Recently, several researchers attempted to determine the effect of yarn structure (pile length, twist level , pile material type) on abrasion characteristics.2-6 They used conventional statistical test methods for their assessments and predictions.

An alternative method, fuzzy logic, can be defined as a mathematical model to study and define uncertainties. The importance of fuzzy logic has continual ly increased, especially in engineering studies.7.9 Fuzzy logic uses the fuzzy set theory and approximate reasoning to deal with imprecision and ambiguity in decision making. I t provides intuitive, flexible ways to create fuzzy inference systems for solving complex control and classification problems. For classification applications, fuzzy logic is a process of mapping an input space into an output space using membership functions and l inguistically specified rules. 1 o

The present study i s aimed at analysing abrasion behaviour of chenille fabrics with fuzzy logic method. The effects of the chenil le yarn count, twist level and pile length on the fabric abrasion behaviour have been studied at constant weave construction and yarn material using fuzzy logic . The cheni l le yarn parameters that affect fabric abrasion have been defined and an experimental study program designed.

2 Materials and Methods

2.1 Preparation of Yarn and Fabric Samples The experiments involved cheni l le yarn samples

manufactured from viscose pi le yarns and acrylic core yarns. Since the viscose chenille yarns are highly vulnerable to abrasive forces, these are preferred to use, expecting that the assessment of the results wil l be more expl icit.

Chenille yarns were produced with a final count of 4 Nm and 6 Nm incorporating two different pile lengths (0.7- 1 .0 mm) and two different twists (700-850 turns/m in S direction) on a Gigliotti & Gualchieri cheni l le fancy yarn machine. Pile and core yarn materials were spun into cheni l le yarn under identical conditions on this machine. 4 Nm count chenil le yarns were produced using two core yarns of the count 2011 Ne (yarn twist 385 turns/m in Z direction, staple acryl ic fibre) and one pile yarn of the

count 2011 Ne. The chenil le yarns of the count 6 Nm were produced using two core yarns of the count 2411 Ne (yarn twi st 580 turns/m in Z direction, staple acrylic fibre) and one pile yarn of the count 301 1 Ne.

Thereafter, woven fabrics were produced with these chenil le yarns using them as fi ll ing in the fabric construction on a Dornier weaving machine. For all fabrics, construction was satin [ 1/4 3(S)] weave, warp yarn was polyester ( 1 50 denier, punched) and warp density was 66 ends/cm. Weft yarns were: 4 Nm chenille yarn with weft density of 10 picks/cm and 6 Nm chenil le yarn with weft density of 1 4 picks/cm.

2.2 Measurement of Fabric Abrasion Properties Abrasion tests of the woven cheni lie fabric samples

were conducted on Martindale wear and abrasion tester in accordance with ASTM 4 158-82 (ref. 1 1 ) . Before starting the abrasion tests, primary trials were made to determine the abrasion cycle that would be suitable for the chenil le fabric specimens. According to the results of these trials, abrasion cycles were limited with 1 0000 rubs and the cut samples were weighed at the beginning and end of 10000 cycles. Mass loss ratios were obtained by dividing mass loss after 1 0000 cycles to initial mass of the samples. 1 2

Measurements were repeated three times for each fabric type. The results were also analysed for significance in differences using three-way repeated measure analysis of variance, and the means were compared by Student-Newman-Keuls (SNK) test at 5 % significance level in the COST A T statistical package. Correlation analysis was performed to observe the relationship between actual fabric mass loss values and predicted fabric mass loss values obtained from fuzzy logic .

2.3 Determination of Chenille Fabric Abrasion Behaviour by Fuzzy Logic and Construction of Basic Rules The aim of this study was to check whether the

fuzzy logic system can be used for predicting the chenille fabric abrasion behaviour. To achieve this aim, it was necessary to compare the results obtained by fuzzy logic with those obtained by real measurements.

In fuzzy logic, uncertainty conditions are determined by pre-defined membership functions. If the most approximate element of the results i s defined as 1 , it can be understood that the other elements are between 0 and 1 and they are changing continuously. The values of every element with the variation between 0 and 1 are called membership degrees and the variation in a subset is called membership

(:EVEN el 01. : PREDICTING A B R ASION BEHAV IOUR OF C H ENILLE FABRIC B Y FUZZY LOGIC 503

function. After that, the definit ion of the fuzzy logic ru le base is implemented by using ei ther Suggeno or Mamdani ' s method. After a solution of the implemented fuzzy logic system has been obtained by using the selected method, the solution is defuzzified. The defuzzification is cal led as the inverse operation of fuzzification.s

Yarn count. pi Ie length and twist level were selected as input variables, whereas fabric mass loss was the output variable. The fuzzy model of thi s problem is gi ven in Fig. 2. Real values of fabric mass loss have been obtained from the experimental study for comparison with those obtained by fuzzy logic. The Fuzzy Toolbox in MATLAB 6.5 was used for mathematical calcu lations.

The numbers of the input membership functions and base widths were obtained using the experimental results and expert knowledge (Fig. 3) . As eight output variable specifications that are cal led fabric mass loss were investigated, the output membership function was divided into eight intervals, as shown in Fig. 4. Because of using the experimental results and since the increments in mass loss values are not l i near. base widths of each interval are different. It is necessary to determine fuzzy logic ru les to make obvious the effect of relationships between the input membership functions and the result. Expert knowledge has been used in the formation of basic rules to define enough of them to reach a correct solution. E lements of these rules are shown in Fig. 5.

3 Results and Discussion The calculated 3-D output-input dependency

resu l ts by fuzzy logic system are presented in Fig 6. The figure shows the relationship among twi st level, yarn count and fabric mass loss. Pile length and yarn count which influence the mass loss of fabric are plotted in Fig. 6b. The relationships among pile length, twist level and fabric mass loss are given in Fig. 6c. Yarn parameters, l ike yarn count, pi le length and twist leve l , are the main effective factors which influence cheni lle fabric abrasion behaviour.

InplIt . . twist le-'\'�r'

Fig. 2 -Fuzzy logic model for cheni l le fabric abrasion

The results of the analysi s of variance test (ANOY A) for mass loss values are summarized in Table 1 . The P-values show that there are statistically significant differences between mass loss values for different twi st levels, pi Ie lengths and yarn counts. Interactions between twist level and pi Ie length. between twist level and yarn count and also between pi Ie length and yarn count have significant effects on mass loss values.

3.1 Effect of Yarn Count on Abrasion Behaviour

A comparison of the cheni l le fabrics in terms of abrasion properties according to SNK test results (Table I ) reveals that the abrasion resistance of the

to )

4.2 4.4 4.6 4.0 5.2 5.4 5.6 5.0 6 Yor1l <ollnt. NIl\

O L-__ � ____ � ________ � ________ �

Pil� 1�11l!1h. nun

700 750 800 850 TWIst level, tums!m

Fig. 3 - Membership functions for (a) yarn count (b) pile Icngth and (e) twist level

3.2 5.1 7 .t,.t 5.2 7.1 13.2 23.3

0.5

4 6 8 10 12 14 16 18 20 22 24 Output variable "Mass loss", 8/0

Fig. 4 - Membership functions of fabric mass loss

504 INDIAN J . FIBRE TEXT. R ES. , DECEMBER 2006

2. II (twi.Uevel 850) and (pile. length is 0.7) and [yarn.counl i, 4) then [mass. loss7. is 5.2) [ 1 ) J I I (twi,tlevel 700) and [pile. lenglh i , 1 ) and [yarn.counl i, 4 ) lhen (mass.lo,,7. i, 7. 1 ) [ 1 ) 4. I I [Iwi,t.level 850) and Ipile.length i, 1 ) and [yarn.count i s 4) then (mass. loss7. is 3.71 ( 1 ) 5 . II [twisUevel 700) and [pile. length i., 0.7) and [yarn.count i, 6 ) Ihen (mass. loss::' i , 1 3.2) (1 ) 6. II [t"';,Uevel 850) and [pile. length IS 0.7) and [yarn. count is 6) then (mass. loss::' " 5. 1 ) [1 ) 7. II llwisUevel 700) and (pile lenglh is 1 ) and [yarn counl is 6) then [mass. loss::' is 7) (1 ) B If [t"';,Uevel 850) and (pile length i, 1 ) and [yarn.counl is 6) then [mass.loss7. is 4

.4) (1 )

Fig. 5 - Some of the defined rules

Table I - Results of variance analysis and Student-New man­

Keuls tests (SNK) for mass loss val ues

Variance analysis

Parameter P = value

Main Effects

Twist level, tpm 0.0000***

Pile length, mm 0.0000***

Yarn count, Nm 0.0005* * *

Interaction

Twist level x pile length 0.0000***

Twist level x yarn count 0.0002***

Pi le length x yarn count 0.0002***

Twist level x pile length x yarn count 0.0007***

Student-Newman-Keuls Test

Parameter Mass loss, %

Twist level, tpm

700 1 2.7" 850 4.6b

Pi le length, mm

0.7 1 1 .8" 1 .0 5 . 6b

Yarn count. Nm

4 9.901 6 7 .4b

*** Denote the significance of the effect of the parameters on mass loss.

".b I ndicates that the valucs arc significantly d i fferent from each other at 5 % significance level.

fabrics woven with fine cheni l le yarns and higher weft density is more than that of the fabrics woven with coarse cheni l le yarns and lower weft density. This situation can be interpreted as 'using higher weft densities wil l affect the pile density on the surface of the cheni lle fabric ' . The yarns wi l l be kept more tightly so the fabric wi ll have a more compact structure. Consequently, yarn count affects the fabric abrasion.4

As can be observed in Fig. 6, when the yarn becomes coarser (Nm count decreases) the fabric mass loss increases. This situation is especially visible for lower yarn twist values. But when the twist value

-rn � 1.'$ � 0.7

Plf 0.8 <le lebgth 1 "" ' lflln

Fig. 6 - 3-D relationship among (a) twist level. yarn count and

fabric mass loss; (b) pi le length, yarn count and fabric mass loss: and (e) pile length, twist level and fabric mass loss

increases the difference between the mass loss values for different yarn counts becomes less; this difference is still significant at 5% significance level.

It i s observed from Fig. 6b that whi le yarn becomes finer, mass loss values decrease at lower pi l e lengths. The difference between the mass loss values I S

decreased at higher pi le lengths. This difference I S still s ign i ficant a t 5% significance level.

3.2 Effect of Pile Length on Abrasion Behaviour One of the results of abrasion i s the gradual

removal of cut pi le yarns from the twi sts of lock yarns. Therefore, factors affecting the cohesion of

(EVEN el at. : PREDICTING ABRASION BEHA VIOUR OF C H EN ILLE F A B R IC B Y FUZZY LOGIC 505

yarns wi l l influence their abrasion res istance. Pi le length affects the cohesion -of yarns and the abrasion properties of chen i l le fabrics . Cheni lle yarn with higher pi le length can provide a more abrasive resistant yarn structure by preventing the movement of fibres from the core yarns . According to the SNK test results, fabrics woven with cheni l le yarns of 0.7 mm pi Ie length are abraded more than that of I mm pi le length. The reason i s that the longer fibres i ncorporated i nto core yarns confer better than the short fibres because they are difficult to remove from the twists of lock yarn .

As it i s observed i n Fig. 6b, the increment i n pile length leads to decrease in mass loss values at al l yarn counts. Figure 6c shows that when the pile length increases there i s a significant decrease i n mass loss values. Thi s difference decreases at high twist l evel s but i t i s sti l l significant at 5 % significance level.

3.3 Effect of Twist Level on Abrasion Bebaviour The abrasion resi stance of a yarn is a function of

both the i ntrins ic abrasion resi stance of the fibre composing the yarn and the geometry of the yarn structure. The geometric arrangements of fibres in yarns and yarns in fabrics are equal ly important when design ing fabrics with high abrasion res istance. It i s known that yarn twist i s directly related t o yarn abrasion resi stance.

The SNK test results given i n Table 1 show that mass loss has a tendency to decrease with i ncreasing twi st level . As shown i n Fig. 6a, when the twist value increases, the fabric mass loss decreases for all yarn counts. Figure 6c shows that when the twist value increases the fabric mass loss decreases for all pi le lengths. It can be concluded that the individual pile fibres in low twist chenil le yarns may not be held together well and hence can be eas i ly pulled out. Mass loss is encouraged by inadequate fibre adherence. A l so, higher twist yarns wi l l retain cohesiveness, presenting a smal ler total surface to be abraded.2.4.6

The mass loss values calculated by using the fuzzy logic model are plotted i n Fig. 7. The effect of twist level on mass loss is higher at short pile lengths whi le i t i s lower at longer pile lengths. Increasi ng twist l evel beyond 825 turns/m leads to a reduction in mass loss values at all pile lengths. This s ituation is val id both for 4 Nm and 6 N m yarn counts.

Moreover, correlation analysis was performed to observe the relationship between actual and predicted fabric mass loss values obtained from fuzzy logic.

2:1 10 (a) 1 8

-- - -16 -14 12 . 10 S

-;;c:, 6 C;:, '" :I ..., ..:: 16

'" (b) '" <': 14 � .... .... . - - ....

11 .

10

S 6

4 700 750 800

Fig. 7 - M ass loss values of fabrics woven wi th (a) 4 Nm and (b)

6 Nm chen i l le yarns w ith d i fferent pile lengths determined by fuzzy l ogic [- 0.7 mm fuzzy, - - 0.8 mm fuzzy, -e- 0.9 mm

fuzzy and -.- 1 .0 mm fuzzy]

25 ,----------------------------------, y = 0.8594x + 1 .2274 o� � 20 +-----------------------------���� .. oS � 15 +------------------.r-��----------� g � 1 0 +---------����------------------� t � c5: 5

O +------.-------.------.-------r-----� o 5 10 15 20 25

A.ctl.lal lnaSs loss. evo Fig. 8 - Relationship between actual fabric mass loss values obtained from measurements and predicted fabric mass loss values obtained from fuzzy l ogic

The relationship between actual and predicted fabric mass loss values i s shown i n Fig. 8. In this analysis, i t i s observed that the relationship between measured and predicted fabric mass loss values is positive. The equation for the correlation between the resul ts of measured and predicted mass loss values i s :

y = 0.8594 x + 1 . 2274

where y i s the predicted fabric mass loss; and x, the actual fabric mass loss (measured).

506 INDIAN 1. FIBRE TEXT. RES., DECEMBER 2006

The linear correlation coefficient for the relationship between measured and predicted fabric mass loss values i s r = 0.978. The border value of the correlation coefficient at a random degree n-2=6 and the significance level u=O.05, above which the correlation exists, is 0.707 .

4 Conclusions

4.1 The abrasion resistance of the fabrics with finer chenil le yarns is more than that of the fabrics with coarser cheni l le yarns. This difference between the mass loss values is high at lower yarn twist values. But when the twist value i ncreases the difference between the mass loss values for different yarn counts becomes less. This difference is sti l l significant at 5% significance level .

4.2 I t is found that when yarn count increases (yarn becomes finer) , despite the changes in mass loss values at lower pile lengths, the difference between the mass loss values is decreased at higher pile lengths. This difference is sti l l significant at 5% significance.

4.3 Since higher pi le lengths affect the cohesion of yarns, there is a significant decrease in mass loss values with the increase in pile length at low twist levels . The increment in pile length leads to decrease in mass loss values at all yarn counts.

4.4 When the twist value increases, fabric mass loss decreases. A high twist level makes the yarn more compact which retains cohesiveness, presenting a smaller total surface to be abraded and increases the pile fibre adherence.

4.5 The fabric mass loss can be easily determined in dependence on yarn count, pile length and twist level by fuzzy logic. Correlation analysis confirmed strong l inear relationship with high value of correlation coefficient (r = 0.978) between measured and predicted mass loss values. I t i s practical ly possible to obtain approaches giving posi tive results with fabric mass loss optimization in a more economical way, without carrying out additional production.

4.6 I t will be useful to carry out further studies on the effect of increasing twist level and pile length on production costs and the comparison of the results.

Usage properties and hand properties must be borne in mind here. Also, it wil l be useful to investigate the effect of cheni l le yarn parameters (twist level, pile length, pile fibre fineness, pi le yarn type and fibre material) on the dimensional and physical properties of chenil le yarns (yarn shrinkage, dye absorption. etc . ) by fuzzy logic.

Acknowledgement The authors wish to thank the authorities of Erol

TUrkUn Textile, Industry and Trade Co. for the opportuni ty of the experimental study and to Prof. Recep EREN for very useful di scussion.

References I Clerck D D, Puissiant P & Langenhove V L, Weavabil ity of

cheni l le yarns on air- jet weaving machines, Int Text Bul/, (4) (2004) 980-984.

2 Kalaoglu F & Demir E, Cheni l le yarn propel1ies and performance of cheni l le upholstery fabrics, Text Asia, 3 (200 I ) 37-40.

3 Kalaoglu F & Dzdemir D, A study of wool cheni l le yarn propertics, Proceedings. I" International Textile, Clothing & Design Conference (University of Zagreb, Croatia), 2002. 1 95- 1 98.

4 Dzdemir D & <;even E K, Influence of chenille yarn manufacturing parameters on yarn and upholstery fahric abrasion resistance, Text Res J, 74(6) (2004) 5 1 5-520.

5 Gong R H & Wright R M, Fancy yams, their manufacture and application (Woodhead Publishing Limited, U.K), 2002, 5 5-56. 8 1 -84.

6 Dzdemir D & Kalaoglu F, The e ffect of material and machine parameters on cheni l le yarn properties, Proceedings, 1st Autex Conference Tecnitex (The University of Minho, Portugal), 200 I , 1 84- 1 89.

7 Zadeh L A, Fuzzy sets, Information Control, 8 ( 1 965) 338-353.

8 Zadeh L A. Fuzzy sets as a basis for a theory of possibil ity. Fuzzy Sets Systems, I ( I ) ( 1 978) 3- 1 0.

9 Branco P J & Dente 1 A, Fuzzy systems model l i ng in practice, Fuzzy Sets Systems, 1 2 1 ( I ) (200 I ) 73-93 .

1 0 H u B , Dale D S , Huang Y & Watson M D , Cotton color classification by fuzzy logic, Text Res J, 72(6) (2002). 504-509

I I AS7M Standard: D 4158 - 82, Vol. 07 .0 1 (American Society of Testing and Materials, USA), 1 995.

1 2 <;even E K , A n Investigation About the Effect of Some Production Parameters on Yarn Properties at Cheni l le Spinning Machines, Masters thesis, University of Uludag. Bursa, 2002.