Aiming at the mechanical failures of vertical centrifugal pumps generated under the combined effects of centrifugal force and fluid power, which are difficult to be accurately recognized by traditional diagnostic methods, a vertical centrifugal pump fault diagnosis method based on the combination of convolutional neural network (CNN) and bidirectional long and short-term memory neural network (BiLSTM) is proposed. Firstly, a recurrent neural network is introduced on the basis of a convolutional neural network, and a feature extraction and fault classification module is established, which can automatically extract spatial and temporal features from the original input signals and identify the key fault modes. Subsequently, a vertical centrifugal pump fault simulation test rig is built, and the vibration signals of the centrifugal pumps under different faults are collected and used to train the proposed diagnostic model. Finally, the diagnostic models constructed by CNN and BiLSTM are compared with a single CNN diagnostic model for fault identification of vibration signals under different fault conditions. The results of the study show that the accuracy of fault diagnosis reaches 100% by using the technique of multi-channel information fusion, which verifies the advantages of multi-channel data fusion. Moreover, the addition of BiLSTM on the basis of CNN is able to better extract and capture useful information from the time series data. In summary, the method proposed in this study can effectively improve the fault diagnosis accuracy and reliability of vertical centrifugal pumps, providing a feasible technical solution for equipment health monitoring in engineering practice.