Abstract: Helium leak detection equipment plays a crucial role in industrial production by identifying minor leaks, thereby ensuring the safety and reliability of systems. However, the complexity of the equipment and the variability of the operating environment pose significant challenges for fault diagnosis. This paper presents a novel fault diagnosis method for helium leak detection equipment based on neural networks, with the aim of enhancing the accuracy and efficiency of fault detection. The methodology involves several key steps. First, operational data from the equipment is collected and preprocessed. The data is then fed into a convolutional neural network (CNN) to extract meaningful feature vectors and reduce dimensionality. Subsequently, the extracted features are input into a recurrent neural network (RNN) to capture temporal dependencies. Finally, the fault classification results are generated using an appropriate activation function. The experimental results show that the proposed CNN-RNN model achieves excellent diagnostic performance across various fault types, with classification accuracy and recall exceeding 99% on the test set, which is significantly outperforming traditional methods. Furthermore, the optimization strategies were explored, such as hyperparameter tuning and model architecture selection, to further improve the model's accuracy and generalization capability.

Key words: helium leak detection equipment, fault diagnosis, neural network, hyperparameter tuning