Fault Diagnosis for Railway Switch Point Using PSO-LSSVM Algorithm

This paper proposes principle and algorithm which combines Least Squares Support Vector Machine (LSSVM) and Particle Swarm Optimization (PSO) to identify the classification of Railway Switch Point faults. First of all, PSO is used to optimize main parameters of the LSSVM kernel function through iterative search, then by using the optimized parameters, LSSVM trains a diagnosis model that can be applied to classify each failure states with the fault free state of the switch point based on training dataset. The experimental result shows that this PSO-LSSVM method obtains a better diagnosis model with higher precision of classification comparing with the 5-fold Cross Validation (CV)-LSSVM model and it provides a feasible method for the future fault diagnosis research on Railway Switch Point System.