Degraded state assessment of switch machine based on MSET and PELT algorithm

Abstract Aiming at the problem that the maintainers are unable to accurately evaluate the degradation state of the equipment due to the high frequency of action of the switch machine on the metro return lines, a method for assessing the degradation state of the switch machine based on Multivariate state estimation technique (MSET) and PELT algorithm is proposed. Firstly, the time domain, frequency domain and wavelet packet energy entropy feature vectors of ZDJ9 switch machine power data of metro return line are extracted, and the redundancy features are eliminated by the Recursive feature elimination (RFE) method to construct the optimal feature set. Secondly, the optimal characteristic parameters of the switch machine in the healthy state are used to construct the history memory matrix of MSET, and thus establish the standard space of the equipment health state. Then, MSET was used to reconstruct the feature vectors of 6124 sets of switch machines’ power data collected in real time in the field to obtain the estimation vectors, and the Degradation indicator (DI) was constructed by calculating the Cosine similarity (CS) between the two vectors before and after the reconstruction to obtain the degradation curve of the switch machine. Finally, the PELT algorithm is used to identify the state change points of the switch machine degradation curve, clarify the stage division thresholds between the degradation states, and complete the evaluation of the degradation state of the switch machine. The results show that: The MSET algorithm can capture the state changes during the switch machine degradation process and accurately calculate the DI. The DI constructed are 6.37% more monotonic and 1.78% more trend than those using Euclidean, Frechet distances, and the DI constructed based on conventional MSET-ED, which can better reflect the actual health state of the switch machine. Combined with the PELT algorithm, the change point of the degradation stage of the switch machine is objectively and reasonably identified, which provides a feasible scheme for preventive maintenance of the switch machine.