Suspension system status detection of maglev train based on machine learning using levitation sensors

The objective of this paper is to design a new method, based on machine learning, to detect the abnormal status of the suspension system of maglev train by using the real-time signals collected by the levitation sensors. The abnormal status is harmful to the operation of maglev train. It is necessary to detect the abnormity timely. Generally, the operation data of train is recorded online, but it is analyzed by experts off-line, the efficiency of this approach is very low. So, an efficient and accurate method is urgently demanded. There are 120 levitation sensors in a carriage, which measure the levitation states with nearly 40khz sampling frequency. In tradition, the information of sensors is just used as the feedback of suspension controllers. This paper will study how to use these information for detecting the abnormal status of suspension system by using support vertical machine (SVM). Firstly, some reasonable features originate from experience are extracted from levitation sensors. Secondly, considering that the fault rate is very low, resulting in less fault samples. So, the SVM method is introduced. At last, the raised machine learning model is applied in the practical usage on maglev train by using the programming in the Digital signal processor (DSP) controller. By this method, the existing abnormal status of suspension system can be diagnosed in real-time. The experiment results show that this algorithm can achieve the detection accuracy of 91.3%.