Remaining Useful Life Prediction of Turbofan Engines based on CNN-LSTM Hybrid Model

Good working condition of mechanical facility is necessary to ensure orderly industrial production, and remaining useful life (RUL) estimation of mechanical facility via current monitoring parameters is an important issue for condition-based maintenance. Deep learning has good capacity of nonlinear feature extraction, high dimension time series modeling and complex pattern recognition, and has good application prospect in the field of RUL prediction of mechanical facility. A hybrid model with CNN-LSTM structure for RUL prediction of mechanical facility is proposed, and a multivariate time series model with optimal length based sliding window is established to extract inner features. The method is validated on NASA's public turbofan engines dataset, achieving accurate remaining useful life prediction of turbofan engines.