Search results

In this research, a new method is developed for grading various types of yarn for appearance using image analysis and an artificial neural network. The images of standard yarn boards were analyzed by image analysis and four different faults factors were defined and measured for each series of yarn counts. For each series of yarn counts, a neural network with one layer was trained by measured fault factors of standard boards. The trained neural networks were used for grading various types of yarns. The yarns were also graded by the conventional standard method.

The results of grading various types of yarns by image analysis and conventional standard method are compared. We found a strong correlation between the results of grading by two methods. Whereas, in the image analysis method, the grading procedure is not dependent on yarn structure and raw materials, we concluded that it is possible to use this method for grading of any types of yarns based on their apparent features.

Fabric fuzziness is a property which affects the appearance, handle, thermal insulation and other fabric features; it also leads to pilling that is a serious problem for the apparel industry. Fuzz on fabric surface has been measured mostly by subjective methods (human vision) rather than objective methods. In this study, an objective method using image analysis techniques is developed for the measurement of fuzz on fabric surface. The effects of fibers blend ratio and abrasion on fabric fuzziness are evaluated. For this purpose, several blended plain fabrics including polyester/viscose fibers with different blend ratios were produced and abraded under various cycles of traversing abrasion. Yarn hairiness was also measured using an Uster tester 4 apparatus before weaving. High correlation between Uster 4 data and the results obtained from the fuzz measurement of fabrics demonstrates that this method has high accuracy and great potential for the determination of fabric fuzziness levels in a quantitative and reliable manner.

To produce wrinkled fabrics, wrinkles are randomly formed without any means of control. There are many research that aim to enhance the surface of fabric in appearance, but there has not been any work carried out to produce wrinkles by using yarn features. The aim of the present research is to produce controlled and custom-made wrinkled fabrics. Wrinkled fabric samples have been produced with high shrinkage polyester-cotton made of intermingled hybrid yarns. Wrinkled weaves have been predicted by using wavelet analysis on sample fabrics and multi-layer perceptron neural networks. The designed network has been trained based on sample patterns and determined a weft insertion plan for producing wrinkles with an acceptable accuracy