An Approach for Residuals Generation based on Structural Analysis and Genetic Algorithm

The world of industry has increasingly complex efficient machines and installations, the fault detection and isolation procedure is an essential module used to guarantee the continued functioning of the systems in the presence of failing component. The diagnostic system depends on the relevance of the information, which is retrieved from the industrial systems, the analytical redundancy relations (ARRs) provide symptoms needed to detect and isolate faults. This work contributes to the field of fault diagnosis and presents a structural approach with genetic algorithm for the design of Analytical Redundancy Relations (ARR) allowing to detect faults in the systems to be diagnosed. This approach aims to extend the possibilities of existing approaches and to reduce the order of complexity. This paper presents a hybrid model-based and data-based approach using structural analysis (bipartite graph) to model the system to be diagnosed. We use the "ranking" algorithm to generate the possible matching in the bipartite graph to generate the ARRs. Genetic algorithm is developed to optimize the generation of matchings in the bipartite graph to be able to analyse and diagnose our system.