Efficient detection of faulty nodes with cuckoo search in t-diagnosable systems

• We adapted two versions of the recently proposed Cuckoo Search metaheuristic for use in the combinatorial optimization realm.
• We have applied the two adapted Cuckoo Search implementations to system-level fault diagnosis.
• We have experimentally validated the optimization capabilities of these schemes and compared them to particle swarm optimization, dynamic mesh optimization, and firefly algorithm.
• Tests were performed across three different distributed diagnosis models under systems of small, medium, and large sizes.
• After a rigorous statistical analysis, we conclude that cuckoo search algorithms offer superior performance, especially YD-CS.

System-level fault diagnosis (SLFD) deals with the detection of all faulty nodes in a set of networked units in a parallel or distributed system. By allowing nodes to test each other under well-defined conditions and carefully inspecting the collection of test outcomes, the true set of damaged nodes can be located. In this paper, we tackle the SLFD problem by using a recently proposed bio-inspired optimization method: the Cuckoo Search. Two formalizations of this scheme were independently put forth by Yang and Deb [16] and Rajabioun [17] for numerical optimization. In this work, we have adapted both methods to work in the combinatorial search space induced by SLFD and we have tested them against other nature-inspired metaheuristics in presence of a class of distributed systems called t-diagnosable systems, since they are easy to generate and eliminate ambiguity in the optimization-driven fault diagnosis. Three well-known test models (i.e. PMC, symmetric and asymmetric comparison) that define how the system units test one another were employed in the simulations. The empirical results reveal that the two cuckoo-based approaches outperformed their competitors in terms of solution quality and spatio-temporal requirements, with Yang and Deb's version achieving notable improvements over Rajabioun’.

Figure optionsDownload as PowerPoint slide