Multi-Client Group-Level Collaborative Diagnosis with Adaptive Multi-Source Domain Transfer

Group-level diagnostic tasks are inherently complex due to significant individual differences and data circulation barriers, and how to leverage collaboration among diverse individuals within the group to achieve effective cooperative diagnosis remains a critical challenge. To address this issue, this article proposes a method named multi-client group-level collaborative diagnosis with adaptive multi-Source domain transfer. The proposed method initially derives the theoretical formulations of the generalization error for source clients and the adaptive multi-source error for target clients. Then, it constructs a unified joint optimization objective by combining them. A Monte Carlo-based optimization strategy is then introduced to solve this joint objective, enabling optimal partitioning of source and target clients, as well as the planning of optimal multi-source adaptation paths for each target client. Experiments on multiple real-world datasets under a multi-client group-level diagnostic setup show that the proposed method significantly outperforms existing methods in diagnostic performance on target clients. This provides effective guidance for collaborative fault diagnosis.