Toward Unsupervised Domain Adaptation Fault Diagnosis: A Multisource Multitarget Method

Despite the remarkable results that can be achieved by data-driven intelligent methods developed for fault diagnosis, they typically presuppose the same distribution for the training and test data sets, as well as sufficient labeled data. Various operating states often exist in practical scenarios, leading to a domain shift that hinders fault diagnosis effectiveness. While recent unsupervised domain adaptation (UDA) methods have focused on cross-domain fault diagnosis, they can struggle to effectively utilize information from multiple source and target domains and achieve simultaneous fault diagnosis in multiple target domains. In this article, we introduce an approach termed weighted joint maximum mean discrepancy-enabled multisource multitarget unsupervised domain adaptation (WJMMD-MDA), which facilitates domain adaptation in complex multisource multitarget settings, a previously unexplored area within this field. The proposed method can capture pertinent information from multiple source and target domains and induce domain alignment between these multiple domains via an improved weighted distance loss mechanism. Consequently, the model acquires domain-invariant and discriminative features across multiple source and target domains, thereby enabling the implementation of cross-domain fault diagnosis. The performance of the proposed method is evaluated in comprehensive comparative experiments on two datasets, and the experimental results demonstrate the superiority of this method compared to the state-of-the-art methods.