Fabric Handle Clusters Based on Fuzzy Clustering Algorithm

Fabric handle depends on physical and mechanical properties, which are measured by the two well-know systems, i.e., Kawabata´s Evaluation System for Fabrics (KES-F) by Japan and Fabric Assurance by Simple Testing (FAST) by Austria. However, the two systems are too expensive and time-costing. A new comprehensive handle evaluation system for fabrics and yarns (CHES-FY) developed based on the single-test multiple indicator measuring principle can reflect the weight, bending, friction and tension basic mechanical behavior as well as handle through only one pulling-out test. The present paper is to select characteristic indexes from the pulling-out force and distance curve acquired by the CHES-FY system, and the characteristics make a vector which expresses the handle of the fabric based on Karhunen-Loeve (K-L) transformation; the other is to make clustering on handle of measured fabrics based on K-means fuzzy clustering. The experiments of thirty fabrics and comparisons between experimental and theoretical results were conducted, which shows that K-means fuzzy clustering algorithm is effective and accurate in sorting fabric handle based on the features selected from the pulling-out force and distance curve of the CHES-FY system.