Multi-fault diagnosis of rolling mill main drive system based on UIO–DBO–SVM

Abstract

In this paper, the fault diagnosis problem of the main drive system of rolling mill with multiple faults occurring at the same time is studied. Considering the internal equivalent current loop and nonlinear friction damping, the nonlinear mathematical model of the main drive system of rolling mill is established. A new fault diagnosis solution based on model residual and data classifier is proposed to solve the problem of complex fault in this system. In the first stage, the unknown input observer (UIO) is designed for system fault detection. The observer design of the system using the [Formula: see text] index will ensure the robustness of fault diagnosis. Lyapunov theory and linear matrix inequality are introduced to prove the convergence of the proposed observer. In the second stage, each set of coupled residual signals generated by the observer is treated as a separate subsequence and modeled and classified directly using a knowledge support vector machine (SVM). Aiming at the nonlinear separability and complexity of residual data set, dung beetle optimization (DBO) algorithm was used to optimize SVM model parameters. The numerical simulation results of 2030-mm cold rolling mill show that the UIO method can rapidly track the system at a speed of 0.2 seconds, the error of motor angular velocity estimation is 0.33% less than that of the extended state observer, and it is more robust. At the same time, the proposed DBO-SVM is compared with SVM, particle swarm optimization (PSO) algorithm-SVM, and jumping spider optimization algorithm (JSOA)-SVM, and the classification accuracy of the proposed DBO-SVM is 99.86%. This scheme not only provides a solution for the detection and classification of complex faults in the main drive of rolling mill, but also provides a new idea for the fault diagnosis of other complex mechanical equipment.