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

•A novel signal-level end-to-end deep learning framework with three layers is proposed for automatically extracting spatiotemporal features.•A spatiotemporal feature learning method is designed by the double deep CNN architecture for mining spatial dependencies and temporal correlations.•A Remaining Useful Life estimation method is proposed for predictive maintenance.•RUL prognostics with C-MAPSS degradation data of turbofan engines and the tool wear data from the PHM Society.