Damage identification of composite based on moving principal component analysis and support vector machine

The vibration response of composite under dynamic excitation is used for structural condition monitoring and damage identification, which has important engineering application value. In this paper, a Particle Swarm Optimization based Support Vector Machine (PSO-SVM) damage identification method is used to classify and predict the location and degree of volume loss damage of Carbon Fiber Reinforced Polymer (CFRP). Through numerical simulation and experimental verification, the feature parameters were extracted by principal component analysis, and then the number of input feature parameters was optimized by PSO-SVM, and the damage was classified and predicted. In this paper, the classification performance of several common Machine Learning (ML) models is compared. The results show that the average accuracy of the PSO-SVM model is 97.22% and 81.94%, both higher than that of Back Propagation Neural Network (BP) and Particle Swarm Optimization based Back Propagation Neural Network (PSO-BP).