Application of MSET and SPRT model for high pressure nitrogen system PHM

To improve the reliability and availability, prognostics and health management technology was applied to the high pressure nitrogen system. This paper presents a multivariate state estimation techniques and sequential probability ratio test model to predict equipment health. In the approach, correlation model among monitoring parameters in normal work condition is constructed firstly. Then, according to the similarities between the current observed feature vector and each history feature vector contained in process memory matrix, estimation of the current feature vector is calculated by using MSET, and residual signal between the current feature vector and its estimation is obtained in turn. Finally, mean test and variance test for the residual signal is executed by using SPRT, and work condition of the system is pronounced. Use the realtime data to test the model, the result showed that using this method can obtain good effect.