Integrating neural networks and Mamdani fuzzy logic for predicting colour coordinates in trichromatic natural dyeing of cationic dyeable polyester fabric

Abstract

This study involves the dyeing of 66 samples with various combinations of madder, barberry root and woad (0%, 10%, 20%, …, 100%) as red, yellow and blue colourants, respectively. The subsequent analysis of variance of the colour coordinates ( L *, a *, b *) indicated that all three components significantly influence the observed coordinates. Quadratic regression, artificial neural network (ANN) and Mamdani‐type fuzzy logic approaches were employed for modelling, while genetic algorithms were utilized to enhance the performance of the neural network and fuzzy logic. The findings demonstrated that all three approaches could simulate the L * index with a respectable level of accuracy. The ANN prediction was, of course, more accurate than the other two models. Moreover, the ANN model was the only one that could accurately predict the other two colour coordinates. For L *, a * and b * colour coordinates, the ANN prediction errors were 0.52%, 5.06% and 2.73%, respectively. For modelling the colour coordinates of polyester fabrics dyed with these natural colourants, the ANN model improved with the genetic algorithm can thus be regarded as an effective and precise tool.