Research on fuzzy neural network control strategy of carbon fiber multilayer diagonal loom

Abstract The conventional tension control techniques, like Proportional Integral Derivative (PID) controllers, have been inadequate in dealing with the nonlinear, multivariable, and dynamic characteristics of carbon fiber weaving processes. This paper covers these limitations by creating a new hybrid tension control mechanism that combines the predictive ability of Back-Propagation Neural Networks (BPNNs) and real time decision making Fuzzy Logic Controllers (FLCs) to deliver hardy, adaptive, and real-time tension control. To address the issue of tension control in carbon fiber weaving looms, a comprehensive approach was adopted. First, a mathematical model was established to describe the relationship between various parameters such as the speed of the loom, the properties of the carbon fiber, and the tension. These two are integrated into one another to complement their respective capabilities. By MATLAB-based simulation and verification, the hybrid system demonstrates higher tension stability performance, defect minimization, and performance improvement compared to traditional control methods. The simulation has been done with 10 different parameters. The response speed is fast, overshoot is low, which indicates the control process is relatively smooth, and the fluctuation error is minimum. The system became stable within 1.5 s, and the overshoot amount is reduced to 19% compared to traditional methods. The results in this research are beneficial to enhance the precision and reliability of carbon fiber weaving processes and thus provide valuable potential for industrial-scale applications.