A wavelet packet transform-based deep feature transfer learning method for bearing fault diagnosis under different working conditions

Deep learning has achieved significant advances in the fault diagnosis of rotating machinery. However, it still suffers many challenges such as various working conditions, large environmental noise interference and insufficient effective data samples. Signal time–frequency analysis feature transfer methods can help solve these problems. Combining wavelet packet transform (WPT) multi-kernel maximum mean discrepancy (MK-MMD), this paper proposes a novel residual network (ResNet)-based deep model for bearing faults. Firstly, devises distinctive WPT map (WPT-TFFM) construction method using on nonlinear non-stationary vibration signals. Then, modified multi-group parallel ResNet is structured to extract depth features WPT-TFFM characteristics small size dispersion. MK-MMD further applied evaluate distribution difference between source target domain data. with classification loss sample set domain, extraction optimized achieve better cross-domain invariance state differentiation capability features. To proposed method, work conducts comparative experiments two test rigs under different loads speeds. The results reveal that offers excellent prevention condition tasks.