Study on hybrid yarns integrity through image processing and artificial intelligence techniques

Transcription

1 Indian Journal of Fibre & Textile Research Vol. 35, September 2010, pp Study on hybrid yarns integrity through image processing and artificial intelligence techniques Mehdi Gholipour Baradari, Dariush Semnani a & Mohammad Sheikhzadeh Department of Textile Engineering, Isfahan University of Technology, Isfahan, , Iran Received 12 August 2009; revised received and accepted 23 December 2009 The commingled hybrid yarns of different structures have been used to investigate the variation in their abrasion resistance over those of simple yarns by calculating the abrasion destruction index. The cotton yarns of the counts 20Ne and 30Ne and cotton-polyester yarns of the same counts (20Ne and 30Ne) at 20, 40 and 60 bar pressure, have been commingled using flat and textured polyester yarns of 150 den. The produced samples are then abraded by a standard metallic object at four different stages including 150 abrasion cycles in each stage. Through image analyzing technique, the abrasive damage of the samples has been investigated and the abrasion indexes are calculated. The Kohonen neural network is used to cluster the samples in 5 classes as per their abrasion resistance. The cotton-polyester yarn (30Ne), hybrid samples from cotton (30Ne) and textured polyester at 20, 40 and 60 bar; and the hybrid yarn made from cotton-polyester (30Ne) and flat polyester at 60 bar are found to be the best. Furthermore, the abrasion resistance of samples improves on increasing the pressure of commingling process. Generally, cotton yarn and textured polyester yarn show the better abrasion resistance in comparison with the other samples. Keywords: Abrasion resistance, Artificial intelligence, Cotton, Hybrid yarns, Image processing, Polyester 1 Introduction Abrasion plays a very important role in textiles. Abrasion resistance is one of the most effective properties of yarns that affect weaving ability, serviceability, and durability of the yarns. Yarn is subjected to abrasion during textile processes due to the interaction in between metal, fibre and fluid. This abrasion is high when the yarn is rubbed against a rough harsh metal surface. It is important to know how several yarns produced from different fibres act against this condition. Some procedures such as finishing processes or illustrating on yarn lead to improve the abrasion resistance 1. Properties of spun yarns depend on their nature, methods of production and spinning system used. Hence, the selection of yarn properties and method of evaluation or measurement in relation to abrasion resistance is difficult 2. When a spun yarn gets abraded, it is possible that the fibres pull out and eventually the end of these fibres are abraded; consequently, because there is no other force except the one between yarn and abrader due to friction to stop the movement, fibres situation changes. If this movement is accompanied with a frequent force, it a To whom all the correspondence should be addressed. d_semnani@cc.iut.ac.ir; dariush_semnani@hotmail.com causes the fibres to pill and the yarn to rupture. Changes in yarn appearance, yarn rupturing, yarn pilling and thin places in yarn body are some results of fibres pilling during this process 2. The first standard method to evaluate the abrasion resistance of yarns was developed in 1970 by ASTM. In this method yarn was subjected to a constant rotational abrasion up to its rupture. The number of cycles was reported as the indexes of yarn abrasion resistance 3. More resistant yarns can bear numerous cycles than weaker ones. But this method was unable to evaluate the extension of destruction caused by yarn abrasion. In the present study, it is attempted to compensate this shortage by determining an index of abrasion. There are numerous parameters that affect yarn abrasion resistance, such as fibre quality, yarn twist, yarn count, spinning system, resin on yarn or fibre body and yarn mechanical properties 4. Another study 5 shows that by increasing the yarn twist, the abrasion resistance of yarn increases. Also, by decreasing the yarn count, the friction ratio between yarn and abraded object is increased by the exponent of 2(1-n) than the yarn diameter 6,7. On the other hand, surface roughness, fibre adhesive property, shearing and cutting occurrences during abrasion process have been found to be the most important factors affecting yarn abrasion resistance 8.

2 BARADARI et al.: HYBRID YARNS INTEGRITY THROUGH IMAGE PROCESSING & AI TECHNIQUES 207 Surface roughness acts whenever the textile material contacts the abrader substance. When these surfaces are touched, the snob points of top substance move over the behind one. The necessary power to thrust two sides could be their weights or tension forces. In this condition, the snob's gradient determines the friction coefficient or µ=tan θ. In abrasion process, the initial forces between the molecules lead to the friction force. Whenever two materials have greater molecular adhesive force, they contact more and therefore the contact area and friction force increase. Fibres with the abrader objects are under the alternative pressure, tension, and bending. If at one point of fibre, the tension load on the yarn rises, it is possible that the fibre ruptures. Yarn is always exposed to shearing force due to its structure and fibres assemblies. The prior investigation shows that the shearing force will be greater if it acts at 45º. Spun yarns could be pulled out due to possible emergence of the fibre ends through the yarn body. In this condition, they cause pilling and interfere with the abrasion process. When loose surface fibres contact with the abrader object, the yarn is fibrillated and the yarn destruction possibility increases. This occurs during the weaving process and usage period. In this condition, the cutting process eliminates the loose fibres and therefore reduces the yarn strength. It is possible that all these events occur simultaneously but there is no reason for them to be connected. During shearing process, the short fibres present in yarn structure can emerge and contribute to increase abrasion resistance of yarns but cutting process does the contrary action and eliminates some fibres from yarn structure and reduces the abrasion resistance of yarn. The observations have shown that the fibres leave the yarn body during the abrasion process and most of them are the short fibres present in yarn structure. It is obvious that by increasing the short fibres percentage in yarn structure, the abrasion resistance decreases and vice versa 4. Recently, various studies have been conducted to evaluate the quality of yarn in terms of evenness and appearance aspects. In these studies, a single yarn or a group of yarns have been assessed through several procedures such as image processing method 4,9,10 but there is no research available on the abrasion resistance of hybrid yarns. The yarn produced by blending two or more yarns is called a hybrid yarn. There are several procedures to prepare hybrid yarns such as core spun spinning, intermingling, etc. Intermingling is usually carried out via an air-jet machine that uses cold compressed air. Figure 1 exhibits an air-jet view briefly 11. If two yarns are fed to an intermingling jet, the blended yarn is similar to a multi-strand yarn 2. If the yarn is prepared by intermingling of two different yarns, the produced yarn is called as commingled yarn, that is used in this research. The present work was undertaken to study the abrasion resistance of the commingled hybrid yarns, which exhibit a high strength because of using filament yarn in their structure and acceptable abrasion resistances in some conditions, in term of abrasion index. These yarns can be used in apparel industry because of their suitable performance. 2 Materials and Methods 2.1 Sample Preparation The spun and filament cotton and cotton-polyester yarns of the count of 20Ne and 30Ne were used; flat and textured polyester yarns of 150 den were used. To prepare the hybrid samples, all the abovementioned spun yarns were commingled in different combination at three different pressures (20, 40 and 60 bar) using air-jet machine (Heberline model). In all, 28 samples (24 hybrid and 4 original spun yarns) were prepared (Table 1). The abrasion tests were performed using ASTM standard (D ) 12. As per the ASTM, the temperature was fixed at 21 C ± 1 C and the humidity at 65% ± 2%. The yarn taking up speed was kept at 100 m/min 12. A machine was prepared for performing the abrasion experiments. A schematic representation of this machine is shown in Fig. 2. The yarn was put into the machine in a loop shape between the cylinder and the abrader object and two ends of yarn were knotted precisely. The cylinder was connected to a shaft which was circulated by a converter. The speed was varied from 0 rpm to 200 rpm. The abrasion tests were performed on the 28 samples at 4 different Fig. 1 Schematic diagram of intermingle air jet 11

3 208 INDIAN J. FIBRE TEXT. RES., SEPTEMBER 2010 Sample Table 1 Input data to Kohonen net Abrasion destruction index variation gradient Diameter variation gradient Co Co Co-PET Co-PET Hy Co20, PET flat at 20 bar Hy Co20, PET flat at 40 bar Hy Co20, PET flat at 60 bar Hy Co 20, PET text at 20 bar Hy Co 20, PET text at 40 bar Hy Co 20, PET text at 60 bar Hy Co30, PET flat at 20 bar Hy Co30, PET flat at 40 bar Hy Co30, PET flat at 60 bar Hy Co30, PET text at 20 bar Hy Co30, PET text at 40 bar Hy Co30, PET text at 60 bar Hy Co-PET 20, PET flat at 20 bar Hy Co-PET 20, PET flat at 40 bar Hy Co-PET 20, PET flat at 60 bar Hy Co-PET 20, PET text at 20 bar Hy Co-PET 20, PET text at 40 bar Hy Co-PET 20, PET text at 60 bar Hy Co-PET 30, PET flat at 20 bar Hy Co-PET 30, PET flat at 40 bar Hy Co-PET 30, PET flat at 60 bar Hy Co-PET 30, PET text at 20 bar Hy Co-PET 30, PET text at 40 bar Hy Co-PET 30, PET text at 60 bar Hy Hybrid, Co Cotton, PET polyester, and text textured. stages, each including 150 abrasion cycles. Yarn touches the abrader object during the rotation cycles. After each cycle, yarn exits from the abrasion machine and images are recorded via a scanner with a 300 DPI resolution. Sample yarns were abraded by abrasion machine again to record another abrasion stage results. After finishing the last cycle (600), yarns were abraded till rupturing and the respective images were prepared. Also, the yarns were scanned before abrading to compare the abrasion cycle effects on their body. Overall, 6 images were available for each sample to do the image analysis. The images observed for the hybrid yarn made by 20Ne cotton and textured polyester at 60 bar are shown in Fig Calculation of Abrasion Destruction Index As per previous discussion, the abrasion index has been determined to compare abrasion resistance of samples more accurately. To study the effect of abrasion on yarn body and to evaluate the proper Fig. 2 Procedure of yarn abrasion Fig. 3 Hybrid yarns during abrasion progress [a before abrasion, b after 150 cycles, c after 300 cycles, d after 450 cycles, e after 600 cycles, and f after rupturing] index to investigate the abrasion destruction, image processing technique through Matlab software was used. It is obvious that the abrasion action reduces the yarn diameter because some parts of yarn body are scrubbed during the abrasion process and then eliminated from the yarn body. The intensity of this occurrence was investigated for the hybrid yarns through the image processing technique. An attempt was also made to omit the body of yarns from the images and just the fuzzes that emerged during the progress were retained using another program, and the abrasive parameters were determined. These abrasive parameters are: (i) the thick defects that are fortified on yarn body such as nep or fuzzes tissues (PFF), (ii) the thick distributed defects that are available on the yarn body and are eliminated during the abrasion progress easily (PHF), (iii) the long fuzzes in which one of their ends is located within yarn body (PLF) and (iv) the short

4 [ BARADARI et al.: HYBRID YARNS INTEGRITY THROUGH IMAGE PROCESSING & AI TECHNIQUES 209 fuzzes in which one of their ends is located within the yarn body and the other end is free (PNF). Hence, in all a 4 1 matrix is obtained, as shown below: p1 p 2 P = p3 p 4 where P i is each of the mentioned parameters. Semnani et al. 10 found that each of these parameters has a specific weight to determine the ultimate abrasion destruction index. These weights are shown in Table 2 that can be used in the shape: W= [W 1, W 2, W 3, W 4 ]. However, the mentioned weights (W) have been presented for standard images 11. In this study, the weights can be used in order to compare the abrasion destruction between all samples because these weights depend on yarn count and are independent of yarn kind and type; hence they are worthy to use in this condition. To compare the yarns in aspect of abrasion resistance, an abrasive index was measured. This index implies the quantity of destruction during each stage and presents a suitable parameter to analyze and compare yarn destruction for all samples; this is called abrasion destruction index. It is obvious, when the abrasion destruction value is high, yarn abrasive property is low. The final abrasion destruction index (I) is calculated by multiplying the abrasion matrix P with weight matrix W using the following equation: I = W.P (1) 2.3 Samples Classification using Neural Network Through the above measured parameters, the yarns can be clustered into the proper classes with regard to their abrasive resistance in order to obtain their abrasion destruction index and yarn diameter reduction. To distribute the yarns in appropriate classes, Kohonen net was used. During the selforganization process, the cluster unit whose weight vector matches the input pattern most closely (typically, the square of the minimum Euclidean distance) is chosen as the winner 13. In this net, by competing between neurons, one neuron (yarn) is introduced as the winner and then clusters into the higher class. To enter the data into the Kohonen net to cluster the yarns, line gradient for abrasion destruction index and yarn diameter reduction were calculated. The Table 2 Appropriate weights for each abrasive parameter 16 Yarn count, tex W 1 W 2 W 3 W main reason to choose the gradient for net input is its quality to describe the abrasive cycle on each input parameters. On the other hand, the yarn diameter reduction and the abrasion destruction index as the variables in abrasion cycle are included in the gradient precisely. It is obvious that the gradient for yarn diameter reduction is always negative but for abrasion destruction index, the value could be negative or positive. For each sample, 6 values for abrasion destruction index and yarn diameter were determined through the image processing method and to calculate their gradient. The net input values (28 vectors) are given in Table 1. As mentioned above, Kohonen net algorithm works through minimum distance rule. The best cluster is the nearest one to the point of (0,0), where the yarn diameter and abrasion destruction index do not change because the yarn diameter is reduced during the abrasion process and the abrasion resistance does not increase anymore. Hence, the best point is the condition at which these parameters do not alter. If both these parameters have same sign (negative or positive), Kohonen net can cluster the sample perfectly through its general algorithm. However, in this study they do not have same sign and it is necessary that the original Kohonen algorithm be modified. To make the difference between the distances, it has been assumed that if two input parameters do not have the same sign, the distance is calculated using the following equation: D = x x0 + y y0 (2) where x 0 and y 0 are equal to zero (the ideal point); x and y are delegated as yarn diameter reduction gradient and abrasion destruction index gradient respectively. It is obvious that the calculated value from Eq. (2) is bigger than the Euclidean distance value [ D = ( x y) 2 ], where x is the yarn diameter reduction gradient; and y, the abrasion destruction

5 210 INDIAN J. FIBRE TEXT. RES., SEPTEMBER 2010 index variation gradient. Now, Kohonen net can cluster the samples in the proper classes. The learning rate and the final epoch were chosen as 0.5 and 20 respectively. Five distinct classes were defined in this condition. Class 1 ends with the yarns that have the highest abrasion resistance while the fifth class ends with the yarns that have the lowest one. Figure 4 shows the Kohonen net used in this study. 3 Results and Discussion As anticipated, the mean diameter of yarns at each abrasion stage decreases due to the elimination of a part of yarn body. To cater this, the deviation of yarn diameter has been calculated through the data obtained by image processing technique. Table 1 shows that the diameter variation gradient is negative after abrasion stages. 3.1 Effect of Abrasion Parameters on Abrasion Destruction Index Shearing and cutting are the most important parameters that determine the abrasion resistance of yarns. Basically, shearing leads to raise the fuzzes up and increase the abrasion resistance of yarns, but cutting eliminates some parts of yarn body and decreases the abrasion resistance. The variation in abrasion destruction index is different for each sample and there is not a distinctive increase and decrease trend in each cycle, for example in the hybrid yarn that is produced by cotton (30Ne) and flat polyester yarn at 60 bar pressure, the abrasion destruction index increases at the first stage but decreases within the next cycle. On the other hand, in the hybrid yarn that is produced by cotton (30Ne) and the textured polyester yarn at 40 bar pressure, the index decreases at first and then increases. However, this increasingdecreasing trend has been noticed several times during abrasion in each cycle. Empirically, the result shows that whenever the shearing occurrence appears, the fuzzes rise up and the abrasion resistance index increases but in cutting process the trend is opposite and the abrasion resistance decreases as well. The effect of shearing process and fuzziness occurrence depends on yarn twist, filament part and the utilized pressure of commingling action. 3.2 Yarn Classification in terms of Abrasion Resistance The yarn classification was done using the abrasion destruction index variation gradient and the yarn diameter variation gradient for each sample. The Fig. 4 Kohonen net 13 Table 3 Net output and yarns classes Class 1 Co-PET 30 Hybrid Co 30, PET textured at 20 bar Hybrid Co 30, PET textured at 40 bar Hybrid Co 30, PET textured at 60 bar Hybrid Co-PET 30, PET flat at 60 bar Class 2 Co 20 Co 30 Hybrid Co 20, PET flat t 60 bar Hybrid Co 20, PET textured at 20 bar Hybrid Co 20, PET textured at 60 bar Hybrid Co 30, PET flat at 20 bar Hybrid Co 30, PET flat at 40 bar Hybrid Co-PET 20, PET flat at 20 bar Hybrid Co-PET 30, PET flat at 40 bar Hybrid Co-PET 30, PET textured at 20 bar Hybrid Co-PET 30, PET textured at 60 bar Class 3 Co 20, PET flat at 20 bar Co 20, PET textured at 40 bar Co 30, PET flat at 40 bar Co-PET 20, PET flat at 40 bar Co-PET 20, PET flat at 60 bar Co-PET 20, PET textured at 20 bar Co-PET 30, PET flat at 20 bar Co-PET 30, PET textured at 40 bar Class 4 Co-PET 20 Co 20, PET flat at 40bar Co-PET 20, PET textured at 60 bar Class 5 Co-PET 20, PET flat at 40 bar procedure has already been discussed. Five classes were defined and the net filled the yarns into the proper class through the described algorithm. Table 3 exhibits the yarn classification in 5 classes from high to low abrasion resistance. The results show that the commingling process does not always provide the better condition for yarn abrasion resistance and in some circumstances the original yarn has the better abrasive property. Several

6 BARADARI et al.: HYBRID YARNS INTEGRITY THROUGH IMAGE PROCESSING & AI TECHNIQUES 211 samples of this kind are shown in Table 3. Hybrid yarns spun with yarn component of 30Ne provide better abrasion resistance than the hybrid yarns spun with yarn of 20 Ne. It is due to compaction of thinner yarn than the thicker one that does not allow the short fibres to extract as fuzzes on yarn body because of better fibre migration in thin yarn in comparison with thick yarn. This condition provides a higher abrasion resistance index. Also it is found that components of textured filament yarn improve the abrasion resistance of hybrid yarns in most conditions in this study. Probably, the high contact area that the textured yarns provide as opposed to the flat one prevents the spun yarn destruction and due to its high resistance against abrasion, the yarn resistance increases but in the flat filament, the spun yarn interferes with abrasive progress and damage because of its low contact area. The pressure plays an important role in manufacturing the hybrid yarns; the pressure sets the fuzzes on the yarn body during the commingling progress. The results demonstrate that by increasing the pressure, the abrasion resistance is increased due to the fibres compactness. The results of Kohonen neural network confirm that class 1 contains samples with high abrasion resistance and class 5 contains the sample with low abrasion resistance. Other classes are arranged between them. Figure 5 shows the samples classification in each class. 3.3 Statistical Study Two variable parameters were used for the hybrid yarns abrasion creation, namely yarn kind and producing pressure. To assure that these two parameters are efficient on yarn abrasion behavior, it is necessary to consider whether the pressure and yarn kind affect the yarn abrasion resistance separately and simultaneously. For this, design and analysis of experiments method were used 14. The figured design is shown below: I = K + P + K P ij i j ij ( * i 1,...,8 ) { j = = 1,2,3 (3) where I is the abrasive destruction index; K, the yarn kind; and P, is the pressure. K*P indicates the presence of pressure and yarn kind at the same time. Also I delegates as the kind number and j is nominated as pressure surface quantity. The investigation was done via SPSS software on the statistical surface of 90%. Results show that the Source Fig. 5 Yarn classification using the Kohonen net Table 4 P-value for each experiment variable P-value Kind Pressure Kind and pressure presence of K and P in the Eq. (3) is necessary but K*P could be omitted from the equation. It means that the simultaneous presence of pressure and yarn kind in the equation does not improve the P-value and can be ignored. This issue corresponded with the prior theories 14. According to prior studies, pressure affects shearing action during the abrasion progress and the yarn kind affects cutting action. Hence, it is possible that the simultaneous presence of them interferes with other side and changes its behavior. Table 4 shows the P-value of the statistical variables for this study. This is obvious as 0 and are lower than 0.1(statistical surface) and K and P do not reject but K*P rejects because is greater than Conclusions It is found that the pressure and the types of hybrid component affect hybrid yarn abrasion resistance. Based on the optimum abrasive properties, the experienced samples having the best abrasion resistance are grouped in class I, these are cotton-polyester yarn (30Ne), hybrid samples from cotton (30Ne) and textured polyester at 20, 40 and 60 bar and the hybrid yarn made by cotton- polyester (30Ne) and flat polyester at 60 bar. Result shows that increasing the pressure leads to an increase the contact among elements of yarn structure, increasing the abrasion resistance. Samples of cotton yarn and

7 212 INDIAN J. FIBRE TEXT. RES., SEPTEMBER 2010 textured polyester yarn show better resistance against abrasion mechanism in comparison to other samples studied. References 1 Gupta B S, Friction in Textile Materials (Woodhead Publishing, USA), Saville B P, Physical Testing of Textiles (Woodhead Publishing Limited, Cambridge, UK), Mahli R S & Batra H S, Annual Book of ASTM Standard, Part 64, Section D1379 (1970). 4 Palaniswamy K & Mohammad P, Affect of the single yarn twists and PLY to single yarn twist ration on the hairiness and abrasion resistance of cotton two PLY yarn, Autex Res J, 6 (2006) Barella A, Manich A, Relation between twist and abrasion resistance of rotor yarns Part 1: Cotton yarns, viscose and acrylics, Text Res J, 53 (1983) Gupta Bhupender S & Yehia Elmogahzy, Friction in fibrous materials, Part 1: Structural model, Text Res J, 61(1991) Gupta Bhupender S & Yehia Elmogahzy, Friction in fibrous materials, Part 2: Experimental study of the affects of structural and morpholigical factors, Text Res J, 63(1993) Brorens P H, Lappage J, Bedford & Ranford S L, Studies on abrasion resistance of weaving yarns, J Text Inst, 83 (1992) Ceven E K & Ozdemir O, Evaluatin of chenille yarn abrasion behaviour with abrasion tests and image processing, Text Res J, 76 (2006) Semnani D, Latifi M, Pourdeyhimi B & Merati A A, Grading of the yarn appearance using image analysis and an artificial intelligence technique, Text Res J, 76 (2006) Ramasamy A & Vinayak O, Commingled & air jet-textured hybrid yarns for termoplastic composits, J Industrial Text, 33 (2004) Standard test method for coefficient of friction, yarn to solid materials, ASTM D , Annual Book of ASTM Standard (American Society for Testing and Materials, USA), Fausett L, Fundamentals of Neural Networks (Prentice Hall International, New Jersey), Montgomery D C, Design and Analysis of Experiments (John Wiley & Sons, USA), 1991.