Abrasion Measurement of Spun Yarns by Image Analysis and Artificial Intelligence Techniques

Abstract

Previously, to evaluate the abrasion of spun yarns, ASTM standard D1379-64 (1970) was applied and valid until 1975. After that, much research work has been carried out to study the abrasion resistance of yarns by using different methods. Recently, new methods based on image processing techniques have been developed. In this research, first, to calculate the abrasion indexes for an image of yarns that are wrapped side by side, the inputs for a back propagation neural network are provided and abrasion destruction indexes are the output. The training of the net is done with data from model images. Moreover, the network has been tested with those model images. To design the model images, attempts are made to simulate various types of defects which are made by abrasion on the body of yarn. After that, groups of spun and filament yarns are tested with both a standard and the new intelligent method and the results are compared. The results prove that trained neural nets have the ability to evaluate the images of yarns trained to the net before; in addition, they can evaluate the images which are inserted into the net for the first time.