Intelligent fault diagnosis for locomotive speed sensor based on LKJ data analysis

In view of the disadvantages of low diagnostic efficiency, long time of diagnosis and high dependence on the experiences of data analyzers in present locomotive speed sensor fault diagnosis with manual work according to LKJ data, the weighted K-nearest neighbor classifier was introduced into fault diagnosis for locomotive speed sensor based on LKJ data. Based on the analysis of the causes of several faults, by combining expert experience, and through the data analysis on LKJ data for different fault types, four failure laws were summed up to eventually derive fault feature vector. Computer simulations show that the LKJ-based WKNN diagnosis method proposed in this paper for locomotive speed sensor is effective and has high failure recognition rate and short diagnosis time. Therefore, compared with artificial fault diagnosis, the fault efficiency with WKNN classifier increases. © 2015, Science Press. All right reserved.