Automatic multi-differential deep learning and its application to machine remaining useful life prediction

•Two active level division units are designed for classifying the raw input data.•A multi-branches CNN (MBCNN) is first proposed to extract the health features.•A multi-cellular BLSTM (MCBLSTM) is first developed to learn the degradation trend.•ADLDNN is first constructed based on MBCNN and MCBLSTM for machine RUL prediction.