Multivariate Analysis of Tactile Sensory Data for Fine Lightweight Knitted Fabrics

Abstract

The market for lightweight next-to-skin knitted wool fabrics requires a reliable tactile sensory or handle assessment system. In this paper, eight bipolar pairs of fabric tactile attributes to describe the tactile sensory properties of fifty-two next-to-skin knitted fabrics were evaluated by twelve experienced trained judges. Principal component analysis and partial least squares regression techniques were employed to analyse the tactile sensory data. The results showed that the most agreed tactile sensory dimension amongst the judges only occurred in the first principal component of each of the eight tactile attributes. In conjunction with the judges' loading coefficients on the agreed component, judges were selected for assessing the next-to-skin knitted fabrics. Furthermore, the relationship between the overall tactile quality and the primary tactile attributes was established using a partial least squares regression method, from which the existence of collinearity between the primary tactile attributes was constrained and the overall tactile quality was predicted with a prediction error of 0.48 subjective units on a scale of 1-10.