The paper proposes a fault diagnosis method for five-phase voltage source inverter diagnosis using one-dimensional convolutional neural network (1D-CNN). Firstly, 1D-CNN architecture with three convolutional layers is designed. The switching function is defined, and the phase current with faulty feature is determined as the input by analyzing the switching function. Then, the phase current with faulty feature is processed using Z-Score normalization, and the faulty data are obtained. The faulty data are processed by sliding overlap sampling, and the dataset for training and testing the model are formed. Thirdly, the datasets are input to 1DCNN model to extract fault features layer by layer. In the end, the open-circuit faults of the power switch with the inverter are identified at the output layer. The research results show that the model can not only diagnose inverter open-circuit faults effectively and accurately, but also has an excellent recognition performance for fault data with mixed noise.