Fault Diagnosis for RPMs Based on Novel Weighted Multi-Scale Fractional Permutation Entropy Improved by Multi-Scale Algorithm and PSO

Railway point machines (RPMs) are important to make the trains operate efficiently and safely. Thus it is vital to research fault diagnosis. Nowadays contactless fault diagnosis attracts more and more attention. In this paper, a sound-based contactless fault diagnosis method is proposed for RPMs by using novel weighted multi-scale fractional permutation entropy (WMFPE) realized by multi-scale algorithm and synchronous optimization strategy. Firstly, to enhance the fault discrimination ability, novel feature named multi-scale fractional permutation entropy (MFPE) is proposed by introducing the idea of fractional calculus to multi-scale permutation entropy (MPE), which is a more powerful tool. Secondly, considering the effect difference of different permutation, new feature named WMFPE is proposed by introducing weight idea. Then, a multi-scale algorithm is developed to reduce information loss during the coarse-grain process, which can improve the diagnosis accuracy. Finally, a synchronous optimization strategy is proposed to optimize the weights of WMFPE and hyperparameters of support vector machine (SVM) at the same time using particle swarm optimization (PSO). The effectiveness of the proposed method is verified by comparison experiments. Results show the proposed method performs best on fault diagnosis of RPMs with an overall accuracy of 99.33% on reverse-normal and normal-reverse switching processes. The proposed method opens a new window for fault diagnosis on RPMs.