Aiming at resonance-state identification for hydraulic gates, firstly, finite element simulation was employed to analyze the vibration frequencies of the gate under various opening degrees. Subsequently, excitation experiments were conducted to compare the vibration frequencies of the gate at different opening degrees, obtaining the vibration state of the gate. Based on the finite element simulation and experimental data, a Generative Adversarial Network (GAN) data augmentation method was constructed to expand the dataset. The improved multi-scale convolutional neural network algorithm was utilized for training and recognition. Validated by real-world and historical dynamic data, the proposed model achieved an accuracy of 85.00% on the held-out test subset for identifying normal and resonance-induced abnormal vibration states, providing a data-driven basis for subsequent gate condition monitoring.