An Aeroengine Fault Diagnosis Method Based on UTCA and DBN

Because the aeroengine measurement samples at different operating points are unbalanced, the diagnosis accuracy of the fault diagnosis model changes in the whole flight envelope. To solve the problem, a fault diagnosis method UDBN based on unsupervised transfer component analysis (UTCA) and deep belief network (DBN) was proposed in this paper. UTCA preprocessing maps the source domain dataset and target domain dataset to high-dimension space. Through optimizing the kernel function, the distance between the two datasets becomes closer enough. Then the mapped source domain dataset is used to train the DBN model, and the mapped target domain dataset is used to test the model performance. The UDBN model was validated through simulation experiments, and the results show that the model has high diagnosis accuracy at different operating points.