Vibration States Detection of HTS Pinning Maglev System Based on Deep Learning Algorithm

Due to self-stable levitation, high-temperature-superconducting (HTS) pinning maglev system has a great potential to become a high-speed transportation mode. As its most significant factor, operation stability and safety reflected by monitoring of vehicle vibration states during its running procedure should get more attention. However, both of the inadequacy of previous researches and continuous growth of vibration conditions increase the detection difficulty of HTS pinning system vibration. Therefore, this paper adapts deep learning algorithm in this detection, by virtue of its plasticity and universality. Initially, horizontal, and vertical acceleration data of HTS pinning system under 108 vibration conditions is collected. Subsequently, this aggregated data is divided into four types of datasets performed with or without Gaussian filter and fast Fourier transform (FFT). Then, back propagation neural network (BPNN) is adapted to distinguish maglev vehicle vibration conditions. Eventually, the detection accuracy of this model in four types of datasets is discussed. The results verify the effectiveness of deep learning algorithm, and the BPNN model on dataset considering Gaussian filter shows the best performance. This paper firstly combines deep learning algorithm with vibration detection of HTS pinning system. Additionally, it also provides an efficient vibration detection approach for HTS pinning system, which could lay a simple basement for its future research fields such as fault identification, operation states classification and so on.