Dye uptake prediction of Disperse Yellow 23 dye on PET fibers in supercritical carbon dioxide using neural networks

Abstract

In the research, the dye uptake of PET fibers treated with Disperse Yellow 23 dye in supercritical CO₂ was systematically explored across a range of parameters including dyeing time (10–90 min), dyeing temperature (70–130°C), and dyeing pressure (15–30 MPa). The models based on back propagation neural network (BPNN), genetic algorithm-BPNN (GA-BPNN), particle swarm optimization-BPNN (PSO-BPNN), and Sparrow search algorithm-BPNN (SSA-BPNN) were trained with input layer encompassing dyeing time, dyeing temperature, and dyeing pressure, and the output layer composed of dye uptake of Disperse Yellow 23 dye. The results show that the SSA-BPNN model demonstrated higher prediction accuracy, as evidenced by metrics including mean absolute error, mean squared error, root-mean-squared error, mean absolute percentage error, and correlation coefficient (R²), with values of 0.823, 1.139, 1.067, 3.565%, and 0.994 for the training set and 0.641, 0.626, 0.792, 3.558%, and 0.995 for the testing set.