Probabilistic Remaining Useful Life Prediction Based on Deep Convolutional Neural Network

Remaining useful life (RUL) prediction plays a vital role in prognostics and health management (PHM) for improving the reliability and reducing the cycle cost of numerous mechanical systems. Deep learning (DL) models, especially deep convolutional neural networks (DCNNs), are becoming increasingly popular for RUL prediction, whereby state-of-the-art results have been achieved in recent studies. Most DL models only provide a point estimation of the target RUL, but it is highly desirable to have associated confidence intervals for any RUL estimate. To improve on existing methods, we construct a probabilistic RUL prediction framework to estimate the probability density of target outputs based on parametric and non-parametric approaches. The model output is an estimate of the probability density of the target RUL, rather than just a single point estimation. The main advantage of the proposed method is that the method can naturally provide a confidence interval (aleatoric uncertainty) of the target prediction. We verify the effectiveness of our constructed framework via a simple DCNN model on a publicly available degradation simulation dataset of turbine engines. The source codes will be released at https://github.com/ZhaoZhibin/Probabilistic_RUL_Prediction.