Deep learning-based remaining useful life (RUL) methods are widely used in the reliability assessment of power semiconductor devices due to the powerful nonlinear mapping capability of their models. However, traditional deep learning (DL) models oftentimes lack the incorporation of physical constraints and laws, leading to less reliable predictions. Moreover, owing to the heterogeneity, DL models achieve accurate lifetime prediction remains challenging in the unknown domain. To address these limitations, taking the insulated gate bipolar transistors (IGBTs) as the sample, a physics-informed unsupervised domain adaptation regression network (PI-UDARN) is developed for their RUL prediction. In brief, we analyze the degradation properties of IGBTs from the perspective of empirical degradation equations. degradation properties of IGBTs are modeled and the corresponding loss function constrains the training process of the network. In addition, the deep CORAL domain adaptation method is used to align the source and target domain features, which achieves cross-domain migration of degradation knowledge and improves the generalization of the model. Experiments on a laboratory dataset to validate the effectiveness of the proposed PI- UDARN.

RUL prediction,domain adaptation,transfer learning,unsupervised learning,physics-informed machine learning