Integrated testability modeling method of complex systems for fault feature selection and diagnosis strategy optimization

The efficacy of system fault diagnosis is intricately linked with the testability design of the system, particularly in intricate systems encompassing numerous components exhibiting diverse failure modes. However, current research shows a dearth of effective testability models and algorithms for generating diagnostic strategies in multi-valued attribute systems (MVAS) with uncertainty. To address this, the present paper introduces a novel method for testability modeling for complex systems. This approach incorporates signal features in lieu of raw sensor signals within the testability modeling process and accounts for the uncertainties surrounding test outcomes. The context of complex MVAS, characterized by inherent uncertainty, the paper proposes a novel method for constructing a four-value dependency matrix (D-matrix). Furthermore, the paper presents a novel sequential diagnosis strategy optimization approach based on a heuristic evaluation function for the multivalued D-matrix with uncertainty. The proposed methodology has been rigorously validated using a real-world case study involving an aero-engine fuel metering device system. Comparative experiments show that the proposed method can achieve better diagnostic performance in the shortest time.