Fault Diagnosis of Subway Plug Door Based on KPCA and CS-LSSVM

Abstract

Fault diagnosis of the subway plug door is an indispensable part of ensuring the safe operation of the city subway system. Taking the developed digital signal processing technologies into consideration, a novel fault diagnosis method for subway plug doors based on Kernel Principal Component Analysis (KPCA) and Least Squares Support Vector Machine (LSSVM) optimized by Cuckoo Search (CS) is proposed. First, fault features are extracted from the original data, and then the dimension of features is reduced by KPCA. Later, CS-LSSVM is used as the classification model for subway plug door faults. Experimental results indicate that the diagnosis model can quickly and accurately identify different fault status. In addition, CS provides faster convergence speed than Genetic Algorithms (GA) and Particle Swarm Optimization (PSO), and CS-LSSVM has higher accuracy in fault diagnosis than BP Neural Network and traditional Support Vector Machine.