A hybrid fault detection algorithm with the g-good-neighbor pattern and its applications

Fault diagnosis has been a key learning paradigm, supporting a wide range of tasks such as network reliability, wafer test, and data center network. However, the accuracy of fault detection depends on the underlying topology of interconnection networks. While existing diagnostic schemes have made significant progress in node failures, addressing the challenges imposed by distinct fault patterns. In many real network scenarios, the presence of faults usually exhibits characteristics such as complexity and heterogeneity, where the communication links between processors may be faulty. To tackle these challenges, we develop a hybrid g-good-neighbor fault diagnosis scheme. First, we establish the g-good-neighbor diagnosability of the hypercube network with missing edges and broken-down nodes. Besides, we present an intelligent hybrid fault perception algorithm (for short IHFP) to identify all faulty nodes and faulty edges. Subsequently, we apply this algorithm to hypercube as well as real flight network. Finally, we verify the efficiency and correctness of the proposed algorithm in terms of Precision, Recall, F1 Score and Accuracy.