Optimizing SVM's parameters based on backtracking search optimization algorithm for gear fault diagnosis

The accuracy of a support vector machine (SVM) classifier is decided by the selection of optimal parameters for the SVM. The Backtracking Search Optimization Algorithm (BSA) is often applied to resolve the global optimization problem and adapted to optimize SVM parameters. In this research, a SVM parameter optimization method based on BSA (BSA-SVM) is proposed, and the BSA-SVM is applied to diagnose gear faults. Firstly, a gear vibration signal can be decomposed into several intrinsic scale components (ISCs) by means of the Local Characteristics-Scale Decomposition (LCD). Secondly, the MPE can extract the fault feature vectors from the first few ISCs. Thirdly, the fault feature vectors are taken as the input vectors of the BSA-SVM classifier. The analysis results of BSA-SVM classifier show that this method has higher accuracy than GA (Genetic Algorithm) or PSO (Particles Swarm Algorithm) algorithms combined with SVM. In short, the BSA-SVM based on the MPE-LCD is suitable to diagnose the state of health gear.