A novel fault diagnostic algorithm with multiple characteristics for multiprocessor systems

As cloud computing, Internet of Things, and parallel distributed system undergo rapid development, accompanied by increasing complexity, the scale of multiprocessor system has grown considerably. Tremendous amounts of processors by interconnected have made it possible to yield faulty units through physical factors or hacking. Such failures are inevitably bound to hinder the reliability and robustness of the system. Fault diagnosis techniques incorporating multiple conditions can greatly enhance the recognition accuracy and efficiency of models in tasks with limited samples. In this article, we first provide a theoretical derivation to explore the non-inclusive 1-good-neighbor conditional diagnosability of the n-dimensional star graphs under PMC and MM* diagnosis models. Then we develop a non-inclusive fault diagnostic algorithm (NFD-PMC) based on 1-good-neighbor fault pattern to detect faulty nodes and faulty-free nodes. Simulation experiments are executed in star networks and email networks for random faults, which verify the correctness and efficiency of our proposed algorithm in terms of ACCR, TPR, FPR and TNR.