A higher-order moment matching based fine-grained adversarial domain adaptation method for intelligent bearing fault diagnosis

Due to the discrepancy in data distribution caused by the time-varying working conditions, intelligent diagnostic methods fail to achieve accurate fault classification in engineering scenarios. This paper presents a novel higher-order moment matching-based adversarial domain adaptation method (HMMADA) for intelligent bearing fault diagnosis. First, a deep one-dimensional convolution neural network is constructed as the feature extractor to learn the discriminative features of each category through different domains. Then, the distribution discrepancy across domains is significantly reduced by using joint higher-order moment statistics (HMS) and adversarial learning. In particular, HMS integrates the first-order and second-order statistics into a unified framework and achieves a fine-grained distribution adaptation between different domains. Finally, the feasibility and effectiveness of HMMADA are validated by several transfer experiments constructed on two different bearing datasets. The results demonstrate that HMS is more effective than lower-order statistics.