Research on fault diagnosis method of transfer learning based on OTASMM

The fact that the labeled fault data is missing in practical engineering affects the accuracy of intelligent fault diagnosis methods. In this paper, a transfer learning method based on OTASMM, OTASMM-1DCNN-TL, is proposed to solve this problem. The method is divided into two parts:1) This paper proposes an oversampling method suitable for highdimensional vibration signals and multi-classifications problem-Oversampling Technique for Adaptive Synthesis of Multiple Minority Classes - OTASMM. 2) OTASMM-1DCNN-TL uses the data synthesized by OTASMM as the source domain data, uses the one-dimensional convolution(1DCNN) as the basic model, and uses transfer learning to diagnose faults. The results on datasets with different imbalance rates show that OTASMM-1DCNN-TL improves the diagnostic accuracy of 1DCNN by 1.25%,3.25%,5.25%,7% respectively. This shows that the method improves the diagnostic accuracy of 1DCNN under unbalanced data sets, and provides an effective solution to the problem of mechanical fault diagnosis under unbalanced data.