Partial domain adaptive fault diagnosis method and system based on PDA-BLS model

The invention discloses a partial domain adaptive fault diagnosis method and system based on a PDA-BLS model. Firstly, a label of a target domain sample is predicted through a probability labeling method, and the weight of a source domain sample is obtained; then, in the BLS framework, fusing the maximum mean value difference of source domain sample weighting and the classification loss of source domain sample weighting, correcting the selection deviation of source domain shared samples through a weighting strategy, filtering out source domain samples irrelevant to the target domain fault category, and only migrating the knowledge of the source domain samples relevant to the target domain fault category; therefore, the negative migration of the private class sample of the source domain in the domain adaptation process is inhibited. In addition, a target domain sample classification loss method is adopted to further promote positive migration so as to learn a BLS classifier, the learned BLS classifier is utili