A Genetic Algorithm Approach for Diagnosability Analysis

Ruben Leal (1) , Jose Aguilar (2) , Louise Travé-Massuyès (3) , Edgar Camargo (1) , Addison Ríos (2)

1 PDVSA - Petroleos de Venezuela
2 ULA - Universidad de Los Andes [Mérida, Venezuela]
3 LAAS-DISCO - Équipe DIagnostic, Supervision et COnduite

In this work we propose to use an approach based on genetic algorithms to obtain analytical redundancy relations to study the diagnosability property on a given continuous production system. Diagnosability analysis for production systems examines the detectability property (the faults are discriminable from the normal behavior of the system) and the isolability property (the faults are discriminable between them). The redundancy relations are based on the minimal test equation support and in a structural analysis over a bipartite graph. The faults analysis is studied using a multi-objective fitness function in a genetic algorithm, which describes the different constraints to be covered in order to reach the diagnosability property on the system. Our approach is tested in a theoretical example and in a real continuous system, a process of extraction of oil by gas injection.

Mots clés: Analytical Redundancy Relations; Gas Lift Well; Index Terms -Genetic Algorithm; Diagnosability; Structural Analysis

Domaines: Intelligence artificielle [cs.AI]; Automatique

Ruben Leal, Jose Aguilar, Louise Travé-Massuyès, Edgar Camargo, Addison Ríos. A Genetic Algorithm Approach for Diagnosability Analysis. International Journal of Engineering Development and Research, 2014, 2 (4), pp.3786-3799. ⟨hal-01701572⟩