Residual Life Prediction of Aeroengine Based on 1D-CNN and Bi-LSTM

Residual life prediction plays an important role in the preventive maintenance of aeroengine, and it is an important means to ensure the safe operation of aircraft and improve the efficiency of maintenance support. A residual life prediction model of aeroengine based on 1-dimensional convolution neural network (1D-CNN) and bidirectional long short memory neural network (Bi-LSTM) is proposed. Firstly, according to the engineering experience, on the basis of the principal component analysis of multi-state parameters, the degradation process is randomly distributed and fitted to obtain the comprehensive performance degradation amount; then, the multi variable time series samples and the corresponding performance degradation amount are brought into the 1D-CNN model for regression analysis to obtain the performance degradation analysis model; then, the performance degradation amount is predicted by the Bi-LSTM time series, The future trend of performance degradation is obtained. Finally, by setting the performance degradation threshold, the residual life prediction results are obtained, and the real-time dynamic perception model from multi state parameters performance degradation analysis performance degradation prediction residual life prediction is obtained. The results show that the proposed hybrid model has lower regression analysis error and degradation prediction error compared with other single deep learning and traditional models, and can get more accurate and reliable residual life prediction results. © 2021 Journal of Mechanical Engineering.