Research On System-level Fault Diagnosis Based On Improved Genetic Algorithm

With the rapid development of information technology,multiprocessor system is widely applied in the field of high-performance computing.However,with the gradual increase of the number of processors in the system,the system becomes more and more unreliable.After the system breaks down,an effective diagnosis technology is urgently needed to diagnose the fault processor set quickly.System-level fault diagnosis is a powerful diagnostic tool.The basic idea is to use the processor's own communication capabilities to test or compare each other,and then locate faulty processors based on symptoms and effective diagnostic algorithm.System-level fault diagnosis not only realizes the self-diagnosis of the system,but also adds no additional cost,so it has high research value.Designing efficient diagnostic algorithms is the key to system-level fault diagnosis.Based on the advantages of genetic algorithms in global search capability and robustness,this paper proposes efficient and reasonable system-level fault diagnosis algorithms based on genetic algorithms.In this paper,the major research contents and task embrace following two aspects:(1)As a centralized diagnosis model,the diagnosis results based on Malek model are more accurate.Therefore,a system-level fault diagnosis algorithm under the Malek model is proposed.The algorithm constructs the constraint equation according to the definition of the Malek model,and designs the fitness function to evaluate the diagnostic accuracy,and then searches the faulty processor set of the system under the action of the genetic operator.Finally,theoretical analysis and simulation results show that the algorithm is effective,efficient and accurate.(2)The traditional diagnostic logic needs to discuss the test results between all the processors in the system,and then draw the diagnostic conclusion.However,in a large processor system,the logic has a lot of redundancy,and most of the processors are fault-free under the premise of the t-diagnostic system.Therefore,the idea of the group is introduced,and the system-level fault diagnosis algorithm under the Ex-test PMC model is proposed.The algorithm first divides all the processors into a small number of groups,and proposes a designated fault-free group method to generate the initial population.Secondly,it designs the fitness function to guide the search direction of the population,and improves the genetic operator to improve the local search ability of the population.In the framework of the genetic algorithm,the faulty processor set in the system is finally diagnosed.The simulation results indicate that the algorithm has good performance in terms of diagnosis accuracy and efficiency.