A new bearing fault diagnosis method via simulation data driving transfer learning without target fault data

Abstract

Transfer learning exhibits exciting advantages in solving the data shortage in fault diagnosis, while most of the existing methods still need target domain fault data, which weakens the performance in some applications where the target fault data could not be provided. Focusing on the no fault data problems, this paper proposes a new transfer learning method based on simulation data. During the route of the proposed method, the theoretical fault characteristic frequencies are pre-evaluated for the monitored bearing, based on which the fault impulses are then constructed. The fault vibration signals are further simulated via mixing the constructed fault impulses with the measured normal baseline data. The envelope spectra of the simulation signals are used as the input to train a network with multi-head attention to identify fault types of the target bearing. The diagnosis performance of the proposed method has been validated via three groups of experimental data.