Remaining Useful Life Prediction Method Based on Convolutional Neural Network and Long Short-Term Memory Neural Network

Remaining Useful Life (RUL) estimation plays a critical role in the health management system, and its accuracy is essential for maintenance decisions and safe operations. Traditional empirical and physical models are limited in some cases as equipment grows increasingly complicated and intelligent. This paper proposes a data-driven method combining Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network, which is more effective and generalizable by modeling the mapping relationship between collected data and RUL with deep learning algorithms. To handle the multiple sensor data and minimize the noises of operations, the K-means clustering approach is used to pre-identify operation conditions. The CNN extracts deep features from the input samples which are sent to the LSTM network to model the dynamics and learn potential degradation representation. Multi-layer perceptron is applied to obtain the regression prediction output of the corresponding RUL. The main advantage of the CNN-LSTM method is the integration of data processing and modeling analysis in a framework, which enables end-to-end RUL prediction, enhances prediction accuracy, and simplifies the prediction process. We verified the performance of the proposed approach on a public aircraft turbofan dataset under different operation conditions and compared it with deep neural networks such as the standard CNN, LSTM, etc. The experiments show promising and effective results.