Improving analog circuit fault diagnosis by testing points selection

Abstract

This paper describes two different methods of improving the ranking list algorithm for catastrophic faults localization method. First one, based on the simulated annealing and the genetic algorithm approach, enables us to search in very large set of possible testing points. Second one is established on the invented fitness function closely related to the structure of modified GRA algorithm. Both methods enable us to investigate circuits with parameter's tolerances included. Presented simple filter-wrapper approach for finding the set of analyzed circuit features may be also used for improving any other fault function defined for the proposed ranking list method.