Remaining service life prediction method of multi-source domain transfer learning based on dynamic distribution self-adaption

The invention discloses a residual service life prediction method of multi-source domain transfer learning based on dynamic distribution self-adaption. The method comprises the following steps: 1) giving degradation data of an active domain and a target domain; 2) preprocessing the degradation data; 3) extracting degradation feature representations of the degradation data of the source domain and the target domain; 4) aligning the degradation characteristic distribution of each source domain and the target domain to obtain various degradation characteristic representations of the target domain; and 5) fusing the multiple degradation characteristics of the target domain to obtain an RUL label through each specific domain predictor, and taking the RUL label as a final RUL prediction label. The transfer learning can utilize the similarity among data, tasks or models to apply the models and knowledge learned in the old field to the new field. According to the transfer learning-based RUL prediction method, a predi