Small Sample Fault Diagnosis Method of Three-phase Inverter Based on Transfer Learning

To address the problem of poor fault diagnosis for three-phase inverter faults when the effective data samples are insufficient under variable operating conditions. An inverter fault diagnosis method based on convolutional neural network(CNN) and transfer learning(TL) is proposed to migrate the fault diagnosis knowledge learned by the model on the source domain to the target domain. It is used in a small sample of three- phase inverter fault diagnosis research. First, the acquired faultsensitive signal is continuously wavelet transformed to obtain colorful two-dimensional time-frequency images conducive to CNN training. Secondly, a pre-training-fine-tuning transfer learning method is used to train the network using a sufficient number of source domain samples to avoid the overfitting phenomenon caused by insufficient data. After migrating the network structure and parameters to the target domain, the deeper network parameters are fine-tuned to make the network adapt to the data distribution of the target domain samples. Finally, TL experiments and fault classification diagnosis were performed on the dataset. The case analysis proves that combining continuous wavelet transform(CWT) and CNN can achieve automatic feature extraction and highly effective use of samples. The introduction of TL enables the accurate classification of small samples under other working conditions. It has a certain value for the research and application of TL learning theory in inverter fault diagnosis.