Application of ANN Weighted by Optimization Algorithms to Predict the Color Coordinates of Cellulosic Fabric in Dyeing with Binary Mix of Natural Dyes

Abstract

Cotton is one of the most important fibers used in the textile industry. The dyeing of cotton with synthetic anionic dyes consumes large amounts of salt and alkali, which makes it a challenge for the environment. Furthermore, the relatively high percentage of synthetic dyes remaining in the dyebath is a potential threat for the environment and human health. The application of plant-derived natural dyes has recently been considered as a promising approach to overcome this problem. Optimization of the dyeing process and prediction of the values of the color coordinates of dyed textiles have always been among the most pronounced challenges in the textile industry, especially when a mixture of dyes or mordants is used. In this study, alum was used for mordanting of cotton and two natural dyes—namely, weld and madder—were used for the dyeing. The samples were dyed with various combinations of mordant, weld, and madder for the weight of the fabric and statistical analysis revealed that all three mentioned parameters were effective in determining the color coordinates. To determine the best model to predict the color coordinates of cotton fabrics, the regression method and ANN models weighted with back-propagation (BP) and optimization algorithms, such as the genetic algorithm, particle swarm optimization, gray wolf optimization, FMINCON (a built-in function of MATLAB software) and a combination of particle swarm optimization and FMINCON (PSO-FMIN), were employed and compared based on the mean squared error (MSE). The obtained results revealed that using the PSO-FMIN algorithm for ANN weighting led to higher accuracy in the prediction of color coordinates. The MSEs obtained for ANN outputs and the corresponding actual values reached 2.02, 1.68 and 1.39 for the l*, a* and b* coordinates, which were 44%, 23% and 26% better than the result obtained with BP, respectively.