Fault diagnosis and predictive maintenance system of new energy vehicle powertrain based on digital twin

Abstract

Aiming at the problems of frequent failure of powertrain of new energy vehicles under complex working conditions, delayed early warning and high cost of traditional operation and maintenance methods, a set of fault diagnosis and predictive maintenance system driven by digital twin (DT) is constructed in this paper. Firstly, a multi-physical field coupling model of powertrain integrating electromagnetic, temperature and vibration fields is established, and a synchronization and fusion mechanism of virtual and real data is designed to realize high-precision virtual and real mapping. Secondly, a CNN-LSTM hybrid model is proposed to complete the fault feature extraction and pattern recognition, and combined with the degradation model to achieve accurate prediction of remaining service life (RUL). Finally, a decision-making model aiming at minimizing maintenance cost is constructed to optimize maintenance strategy. The accuracy of the proposed CNN-LSTM model is over 97.83%, and the MAPE predicted by RUL is as low as 2.35%.The maintenance cost is reduced by 30.6% and the failure rate is reduced by 56.2%. The research results provide an efficient technical scheme for intelligent operation and maintenance of powertrain of new energy vehicles.