Remaining Useful Life Prediction of Turbofan Engine Based on Probabilistic Principal Components Analysis and Gated Recurrent Unit

Abstract

As the core of the aircraft, the prediction of the remaining useful life(RUL) of the engine is of great significance to the stable operation of the aircraft. A remaining useful life prediction model based on a combination of Probabilistic Principal Components Analysis(PPCA) and Gated Recurrent Unit(GRU) is proposed for complex system monitoring data with large data volume and high dimensionality. The model uses PPCA to reduce the dimensionality of the original data, combines the Gaussian model with the Principal Components Analysis (PCA) and uses the Expectation Maximum (EM) algorithm to solve the best probability model in order to maximize the retention of useful information from the original data. The processed data is fed into the GRU model to mine the time series information, and finally, the model is further modified by the difference between the predicted and actual values. Simulation experiments using the C-MAPSS turbofan engine dataset and compared with the results of related studies such as support vector machines and Long-Short Term Memory(LSTM) neural network. The prediction results prove effectiveness of the method and provide support for predicting the remaining useful life.