Evaluation of a novel computer dye recipe prediction method based on the PSO-LSSVM models and single reactive dye database

In the process of fabric dyeing production, the color samples are usually obtained from the users first, and then the color-matching personnel conducts proofing according to the color of the samples. Therefore, the efficiency of color matching is low, and many fabric color-matching technologies emerge. In this work, we designed a series of dyeing schemes, and built three single reactive dye databases by dyeing cotton knitted fabrics with Levafix Red, Levafix Blue and Levafix Amber in the pad-dry-pad-steam (PDPS) process, respectively. Furthermore, three prediction models were established based on PSO-LSSVM, which took the color parameter L∗, a∗ and b∗ value of the dyed fabrics as model input and dye concentration as model output. The color parameter L∗, a∗ and b∗ values of the fabrics dyed at the dye concentration predicted from the PSO-LSSVM models are consistent with actual measured values of the tested cotton fabrics. All evaluation indexes show that the Ep of PSO-LSSVM models are above 96%, possessing a high accuracy of dye recipe prediction and strong dye specificity, which can be used in actual dyeing color matching.