Multi-level fault diagnosis of gearbox based on knowledge graph and deep learning

The fault diagnosis of gearbox plays an important role in ensuring the safe and stable operation of machinery. To address the problems caused by the imbalanced distribution of gearbox fault samples and complex noise components in signal, a fault diagnosis method that combines knowledge graph and attention and dilation based one-dimensional convolutional neural network (KG-AD-1DCNN) is proposed. Firstly, entity extraction from equipment information and preprocessing such as data augmentation and denoising of fault signals are carried out based on knowledge graph. Subsequently, the preprocessed one-dimensional signal is input into the deep learning model for fault classification. The model utilizes several parallel dilated convolutional layers to extract multi-dimensional features, and adaptively weights important features through the attention module. Finally, the noisy PHM2009 gearbox dataset is used to evaluate the model. The experimental results show that the proposed method can effectively achieve multi-level fault diagnosis of gearbox, and the testing accuracy is higher than that of traditional deep learning methods.