Fault Diagnosis Method of Control System Actuator Based on Particle Swarm and Support Vector Machine

Abstract

The invention discloses a control system actuator fault diagnosis method based on Multi-swarm Cooperative Chaos Simulated Annealing Particle Swarm Optimization-Support Vector Machine (MCCSAPSO-SVM). Through joint noise reduction and improved Empirical Mode Decomposition (EMD) method, noise reduction processing and feature extraction are performed on the collected actuator output signals; optimization is carried out by using multi-group cooperative chaotic simulated annealing particle swarm optimization algorithm (MCCSAPSO) Support vector machine structural parameters, this optimization method can not only effectively avoid particle swarm prematurity, but also overcome the misjudgment problem caused by information exchange of a single particle, and at the same time improve the algorithm convergence speed and optimization accuracy; the use of combined kernel functions ensures that the support vector machine is good Generalization ability and learning ability; using training data to build partial binary tree support vector machine, this structure transforms a complex multi-classification problem into several binary classification problems, reduces the amount of calculation, and improves the real-time performance of diagnosis. The method of the invention processes common state signals that are easily obtained by the control system, and through the output of the partial binary tree support vector machine, it can effectively judge whether the actuator of the control system fails in real time, and can determine more accurately when the actuator fails. Type of failure. The invention is used for real-time fault diagnosis of a high-precision control system.

Description

technical field [0001] The invention relates to a control system actuator fault diagnosis method based on multi-group cooperative chaotic simulated annealing particle swarm optimization algorithm-support vector machine (MCCSAPSO-SVM), belonging to the technical field of control system signal processing and actuator fault diagnosis. Background technique [0002] Due to the complex composition of modern control systems, they often need to work for a long time and with high loads under different environmental conditions, which inevitably leads to various failures in the control system. Especially in aerospace, medical, large-scale mechanical production and other fields, subtle faults sometimes cause extremely serious economic losses and personal injuries, so the status monitoring and fault diagnosis of equipment operation has become an important research topic. Fault diagnosis of actuators in control systems can be divided into fault detection and fault isolation, which are two...

AI Technical Summary

Problems solved by technology

While combining the advantages of each single kernel function in the composition of the combined kernel function to obtain a support vector machine with better performance, there are still two problems as follows: 1) Due to the large number of kernel parameters of the combined kernel function, improper selection is very easy. The performance of the combined kernel function is not as good as that of a single kernel function; 2) When the type of single kernel function and the kernel parameters contained in the combined kernel function have been determined, the selection of the weight of each single kernel function will also affect the performance of the combined kernel function. performance impact

Among them, the particle swarm optimization algorithm (Particle SwarmOptimization, PSO) needs to adjust fewer parameters, and has the advantages of simple structure, easy implementation, and fast convergence speed. Unable to find the global optimal solution

Introducing the idea of ​​chaos into PSO not only improves the ability of the particle swarm optimization algorithm to jump out of local extreme points, but also improves the convergence speed and accuracy of the algorithm to a certain extent. Causes misjudgment of particle information exchange

The simulated annealing algorithm (Simulated Annealing, SA) uses the probabilistic jump characteristics to randomly search for the global optimal solution of the optimization function, and can also avoid falling into the local optimal solution during the search process, and has good global optimization performance, but because The initial value determines the performance of the algorithm to find the global optimal solution of the problem, so the algorithm also has certain limitations

Images