Image classification method of bearing fault based on BOA optimization ResNet-18

Aiming at some inefficient and difficult classification problems caused by the difficulty of determining parameters in data-driven bearing fault image classification, a bearing fault image classification method based on BOA ResNet-18 is proposed. Firstly, the original vibration signal data of the bearing is added to the noise and converted into an image method by signal processing such as two-dimensional wavelet transform for preprocessing, and the corresponding target data set is collected. Then, the ResNet-18 network model that has optimized the corresponding parameters by the BOA group optimization algorithm is trained and fine-tuned on the generated target dataset. The final adjusted ResNet-18 model is applied to the fault image classification of bearings. The method was verified on the bearing dataset of Case Western Reserve University, and the result was that the surface method could obtain an average accuracy of 99.5% of the bearing fault image classification, which has certain engineering application prospects and is of great significance for solving the bearing fault image classification.