Two Efficient Algorithms About System-level Fault Diagnose

With the growing size of the system,the probability of system failure increases,system-level fault diagnosis become an effective means to solve the problem.Since the current system-level fault diagnosis algorithm can not take the high-speed and precision into consideration at the same time,so some networks systems focus rapid diagnose,in contrast,some network systems focus high precision diagnosis,this article designs two different efficient algorithms to meet different needs of the network systems.In order to diagnose the fault units in the system quickly,we firstly use the Mussels Wandering Optimization algorithm to solve the system-level fault diagnosis problem,Propose an efficient fault diagnosis algorithm---the Mussels Wandering Optimization Fault Diagnosis(MWOFD).Considering the Mussels Wandering Optimization algorithm can not be used into system-level fault diagnosis directly we propose the Mussels Wandering encoding and Initialization,and design the new fitness function according to equation constraint conditions that the diagnostic model has to met,at the same time we optimize the existing binary mapping algorithm.Finally,the new algorithm is compared with AD-FAFD algorithm,FAFD algorithm and EAFD algorithm experimentally,experimental results show that MWOFD algorithm improves the diagnostic accuracy slightly and reduces the number of iterations to improve the efficiency of diagnosis.In order to diagnose the fault units in the system effectively,we design a fault diagnosis algorithm,CS-BPFD algorithm,which is based on Chwa & Hakimi model.By introducing adaptive dynamic step size factor,this thesis improves a novel swarm intelligence algorithm,the cuckoo search algorithm;then the improved algorithm is used to optimize BP neural network,improving the activation function of BP neural network,and a new diagnosis algorithm based on the optimized BP is designed.Finally,the new algorithm is compared with BPFD algorithm experimentally,experimental results show that the new algorithm not only reduces the number of iterations and the training time effectively,but also improves the diagnosis accuracy.