LSSVM-based color prediction for cotton fabrics with reactive pad-dry-pad-steam dyeing

Abstract Least-square support vector machines (LSSVM) have been strongly moving from preliminary theoretical investigations into practical industrial applications. Successfully attaining a required color is a crucial task in the dyeing industry. In this paper, an LSSVM-based color quality prediction model for cotton fabrics dyed with a reactive dye using the conventional pad-dry-pad-steam (PDPS) process is proposed. The concentrations of the dye, sodium carbonate and sodium chloride, as well as the drying and steaming times were selected as the model inputs, while the CIE Lab value (L*, a*, b*) and the K/S value of the dyed fabric were used as the model outputs. The two prominent reactive dyes, Remazol Red and Remazol Navy, were used separately to dye cotton fabrics. The proposed LSSVM-based model gave mean absolute errors of less than 0.5 and 1.5 for predicting the K/S value and the CIE Lab value (L*, a*, b*), respectively. As well, the model had an acceptable average color difference ΔE* of less than 2.0 for the fabrics dyed using the PDPS process with both reactive dyes.