Test Optimization Selection for Fault Detection and Isolation Under Multivariable and Multifault Scenarios

Abstract

Test optimization selection (TOS) is a crucial technology in testability design, playing a key role in intelligent manufacturing by enhancing product maintainability and reliability while reducing life-cycle costs. As intelligent manufacturing systems demand higher reliability and efficiency, effective TOS methods are essential for ensuring real-time fault diagnosis and predictive maintenance. However, existing TOS methods inadequately account for correlations between test outcomes in metrics modeling and offer limited solutions to the low fault isolation rate (FIR) caused by multiple faults. An innovative TOS approach is developed by considering fault detection rate (FDR) and FIR metrics via the D-vine copula and Bhattacharyya coefficient method, along with an improved binary particle swarm optimization (DVBC-IBPSO) method to minimize the number of required test points. First, the D-vine copula method is introduced to model test metrics, effectively capturing strong correlations between test outcomes. Second, considering the ambiguity group problem induced by multiple faults, a DVBC combined method is developed to quantify the similarity between fault distributions and model the FIR metric. Third, leveraging the constructed test metrics models, an IBPSO algorithm is employed by incorporating a newly designed objective function that selects the most cost-effective test points while ensuring FDR and FIR remain within acceptable thresholds. The proposed method enhances the reliability and efficiency of intelligent manufacturing systems by optimizing fault diagnosis processes and improving overall system health management. Its validity is established through experimental studies on one commonly used critical circuit in industrial systems.