Reliability evaluation of mechanical and electrical equipment based on probability statistical model

Abstract To solve the limitation of traditional reliability evaluation methods for electromechanical equipment in dealing with the uncertainty and dynamic evolution characteristics of operating data, a dynamic reliability evaluation model integrating physical degradation mechanism and data-driven is constructed in this study. Based on the Wiener process, this model constructs a physical model that can describe the random degradation of equipment performance with time to represent the internal random failure process of equipment. Furthermore, Bayesian inference framework is introduced, and Markov Chain Monte Carlo (MCMC) algorithm is used to integrate the real-time monitoring data into the model to realize the dynamic online updating and uncertainty quantification of the parameters of the degraded model. By modifying the prior knowledge with real-time data, the probability distribution of the Remaining Useful Life (RUL) of the equipment at any time can be obtained, and its reliability can be dynamically evaluated. To verify the effectiveness and superiority of the proposed model, this study uses the degradation dataset of turbofan engine for example analysis. The experimental results show that the proposed model is superior in predicting RUL. Its Root Mean Square Error (RMSE) is as low as 12.88, which is significantly better than the benchmark models such as Long Short-Term Memory network (LSTM). The model can also effectively quantify the prediction uncertainty, and its Prediction Interval Coverage Probability (PICP) reaches 94.1%. The research proves that the fusion model can effectively integrate the physical prior information with the actual monitoring data, and realize a more accurate and dynamic reliability evaluation of electromechanical equipment.