Research on data augmentation with generative adversarial networks in the fault diagnosis of reciprocating compressor bearings

Reciprocating compressors are widely used in industrial fields, and the stable operation of their bearing components is crucial to the overall performance of the machine. However, as bearings are one of the components most prone to failure, their fault diagnosis is particularly important. The challenge in accurate diagnosis arises due to the fact that bearings typically operate in a stable state, resulting in a scarcity of abnormal data samples. This study focuses on the fault diagnosis of bearings in reciprocating compressors and proposes a method based on Generative Adversarial Networks (GAN). By simulating real fault data, GAN can generate a large number of synthetic fault samples, addressing the issue of data imbalance. These synthetic samples are combined with real normal operating data samples to form a more balanced dataset for training a neural network classifier. Experimental results validate the effectiveness of this method in enhancing the fault diagnosis performance of reciprocating compressor bearings, demonstrating its immense potential in industrial applications.