A spatiotemporal feature learning-based RUL estimation method for predictive maintenance

Studies on deep learning (DL) methods in maintenance support systems has achieved great success since degradation patterns and remaining useful life (RUL) of critical equipment can be learned and predicted by DL techniques. However, mining spatial and temporal dependencies from multivariate sensor signals and fusing spatiotemporal features sufficiently is a daunting task. In this proposal, a novel signal-level DL framework containing three layers called STRUL is proposed for end-to-end RUL estimation. The first data segmentation layer is designed based on sliding window manner makes STRUL work directly on raw signals. Then in information extraction layer, two feature extractors based on convolutional neural network are synchronously used to learn spatial and temporal features from each time series. The last information aggregation layer is designed for fusing features so that the holistic spatiotemporal features can be learned and further contribute to RUL prediction. The proposed STRUL model achieves better comprehensive performance on RUL estimation task than state-of-the-art models which has been verified by two case studies.