Remaining useful life predictions for turbofan engine using semi-supervised DBN-LSTM joint training model

In the industrial field, the health of the machine may decline in the process of working, so it is necessary to regularly maintain the health of the machine. However, traditional health maintenance methods have problems of insufficient or redundant health maintenance. Predictions of remaining useful life (RUL) play a vital role in realizing more accurate system monitoring and health management. In recent years, with the rapid development of industrial big data and deep learning, the use of multi-sensor equipment information and deep learning neural network to predict RUL has made significant progress. However, the current RUL prediction faces the following challenges: (1) Separate Data fusion and RUL prediction steps usually result in a lack of internal connection between the two models; (2) The end-to-end prediction method using a single deep learning neural network does not provide health indicator information about the degradation. (3) The industrial data available for model training is still insufficient. To overcome these shortcomings, a new RUL joint prediction model is proposed in this work. The framework combines data fusion and RUL prediction models in series for simultaneous training, which not only provides continuous visualization of the health degradation of the system, but also ensures the efficiency of RUL prediction. Based on this framework, a semi-supervised joint training model composed of deep belief network (DBN) and long short-term memory (LSTM) is designed. The effectiveness of the proposed method is verified through the C-MAPSS dataset. The application results demonstrated that the proposed method is superior to other state-of-the-art methods.