A novel neural network warp tension control combined with online identification

Abstract

In order to solve the tension fluctuation of long warp yarn during loom weaving, a control method of differential separation I-RBF-PID let-off system based on online identification transfer function was proposed. It is difficult to establish an accurate mathematical model because of the nonlinear, time-varying and complex coupling characteristics of the let-off system caused by the clearance of the transmission system and the change of the reel diameter of the weaving shaft. LMS algorithm was used to identify the time-varying and high-order transfer function of loom, and the PID parameters were set by RBF neural network optimized by improved PSO. Step size optimizer and differential separation PID control strategy were introduced to improve the convergence accuracy and reduce overshoot of RBF-PID. The time-varying transfer function was applied to the I-RBF-PID control system by designing the transfer function updating formula. The results are compared with those of conventional PID controller and RBF-PID controller under the performance conditions of overshoot, arrival time and response speed. The effectiveness of the proposed high-precision tension control scheme was verified by experiments.

Keywords

• online identification
• let-off system
• tension control
• transfer function
• differential separation
• neural network