GA-Based fault diagnosis algorithms for distributed systems

Abstract

Distributed Systems are becoming very popular day-by-day due to their applications in various fields such as electronic automotives, remote environment control like underwater sensor network, K-connected networks. Faults may aect the nodes of the system at any time. So diagnosing the faulty nodes in the distributed system is an worst necessity to make the system more reliable and ecient. This thesis describes about dierent types of faults, system and fault model, those are already in literature. As the evolutionary approaches give optimum outcome than probabilistic approaches, we have developed Genetic algorithm based fault diagnosis algorithm which provides better result than other fault diagnosis algorithms. The GA-based fault diagnosis algorithm has worked upon dierent types of faults like permanent as well as intermittent faults in a K-connected system. Simulation results demonstrate that the proposed Genetic Algorithm Based Permanent Fault Diagnosis Algorithm(GAPFDA) and Genetic Algorithm Based Intermittent Fault Diagnosis Algorithm (GAIFDA) decreases the number of messages transferred and the time needed to diagnose the faulty nodes in a K-connected distributed system. The decrease in CPU time and number of steps are due to the application of supervised mutation in the fault diagnosis algorithms. The
time complexity and message complexity of GAPFDA are analyzed as O(n*P*K*ng) and O(n*K) respectively. The time complexity and message complexity of GAIFDA are O(r*n*P*K*ng) and O(r*n*K) respectively, where ’n’ is the number of nodes, ’P’ is the population size, ’K’ is the connectivity of the network, ’ng’ is the number of generations (steps), ’r’ is the number of rounds. Along with the design of fault diagnosis algorithm of O(r*k) for diagnosing the transient-leading-to-permanent faults in the actuators of a k-fault tolerant Fly-by-wire(FBW) system, an ecient scheduling algorithm has been developed to schedule dierent tasks of a FBW system, here ’r’ denotes the number of rounds. The proposed algorithm for scheduling the task graphs of a multi-rate FBW system demonstrates that, maximization in microcontroller’s execution period reduces
the number of microcontrollers needed for performing diagnosis.