Data-Driven Optimal Test Selection Design for Fault Detection and Isolation Based on CCVKL Method and PSO

Yang Li (1) , Hongtian Chen (2) , Ningyun Lu (1) , Bin Jiang (1) , Enrico Zio (3, 4)

1 NUAA - Nanjing University of Aeronautics and Astronautics [Nanjing]
2 University of Alberta
3 POLIMI - Politecnico di Milano [Milan]
4 CRC - Centre de recherche sur les Risques et les Crises

Accurate fault detection and isolation (FDI) relies on the information collection. This can be done by the optimal test selection which can also reduce the life cost of engineering systems. In recent years, some researchers have made lots of achievement on solving the test selection design (TSD) problem. However, few of them concerned a method to deal with the ambiguity problem caused by the multiple fault modes. In this article, a data-driven-based method for test selection is proposed to build an accurate TSD model. Then, we propose a copula function on cross validation-based Kullback–Leibler divergence (CCVKL) method to construct an accurate constraint model. An improved discrete binary particle swarm optimization (IBPSO) algorithm is used to obtain the optimal test design solution. The proposed method also in comparison to three other existing methods are performed in an electrical circuit.

Mots clés: Cross validation; fault detection and isolation (FDI); improved discrete binary particle swarm optimization (IBPSO); Kullback–Leibler (KL) divergence; multiple faults

Domaines: Sciences de l'ingénieur [physics]

Yang Li, Hongtian Chen, Ningyun Lu, Bin Jiang, Enrico Zio. Data-Driven Optimal Test Selection Design for Fault Detection and Isolation Based on CCVKL Method and PSO. IEEE Transactions on Instrumentation and Measurement, 2022, 71, pp.1-10. ⟨10.1109/TIM.2022.3168930⟩. ⟨hal-03908325⟩