基于ISSA-BP神经网络的纺纱生产工艺参数反演OACSTPCD

In order to solve the problems of lower accuracy of traditional inversion model,excessive randomization of weights and initial threshold values,worse stability and accuracy of traditional BP(back propagation)neural network,a parameter inversion model for spinning production process of BP neural network based on improved sparrow search algorithm(ISSA)was proposed.Ten key process parameters were extracted by grey relational analysis and used as model input.Chebyshev chaotic map,sine cosine algorithm(SCA)and adaptive weight factor were introduced to optimize sparrow search algorithm(SSA).And BP neural network was optimized by the improved sparrow search algorithm(ISSA).On this basis,the parameter inversion model of spinning production process was constructed.ISSA was used to solve the parameter inversion model.Inversion verification was carried out with fiber property and spinning workshop as objects.The experimental results showed that the MAPE,MSE,MAE,iteration time and fitness of ISSA-BP predicted values were all better than those of SSA-BP model.The process parameters after inversion optimization were forecasted.The mean relative error(MRE)between predicted quality index and expected quality index was 5.04%.It is considered that the inversion precision of spinning process parameters based on ISSA-BP neural network is higher,which is helpful to the reasonable design of process parameters.