As every processor node is the basic component of the multiprocessor system,its failure may lead to the crash of the entire system.With the increasing size of the multiprocessor system,it is indispensable to locate the faulty unit effectively and replace it with the normal node in time for the normal operation of the system.The PMC model,a classical fault diagnosis model,its fault diagnosis problem has always been an essential direction of system reliability researches.Therefore,the fault diagnosis algorithms of large multiprocessor system under the PMC model are studied in the paper and the specific contents are as follows:The probabilistic matrix diagnosis algorithm under t-diagnosis system based on the PMC model is researched.Firstly,according to the analysis result of simulation experiments on the diagnostic,the higher fault alarm rate is presented.Therefore,the thought of absolute fault nodes aggregation and nodes grouping is introduced in the next.The former is calculated to identify some fault nodes,and the latter is used to replenish the non-fault sets.Thus the rigorous condition is impaired and the accurate diagnosis result is obtained with this diagnosis strategy.Finally,the modified probability matrix diagnosis algorithm is proposed to improve the diagnostic efficiency.S imulation experiments show the modified probability matrix diagnosis algorithm keeps the superiority of high detection accuracy,reduces fault alarm rate with the nodes increasing and improves the diagnostic efficiency,thus,the algorithm application gets extended.The adaptive sequential diagnosis algorithm in PMC model is studied,it would complete the whole diagnosis task through four rounds of diagnostic tests.The first testing round is to diagnose the part faults by testing each other with a pair of nodes.Then,the second round is to divide the remaining nodes into 01 sequence,the nodes in 01 sequence would be estimated with loop-back diagnosis procedure to find faults again.Next,every the definite fault node in before rounds is the diagnosis center,the unknown nodes in detection area which is the diagnosis center need tes t the fault nodes repeatedly and can be identified with the test results in the third round.Lastly,the remnant processors are judged with normal nodes to accomplish the system-level fault diagnosis in the final round.The simulation results show that the algorithm has effective detection and wide adaptability. |