E-Dominance Guided Multi-Objective Test Selection Under Test Uncertainty

Optimal test selection aims to reduce test cost and time while increasing fault coverage and diagnostic accuracy, making it a core task in design for testability. This work accounts for uncertainty in fault–test relationships and unreliable test outcomes by introducing a new testability metric defined in terms of Miss Detection Rate (MAR) and False Alarm Rate (FAR), by formulating test selection as a multi-objective optimization problem. An E-dominance mechanism is developed to embed predefined preference weights directly into the dominance relation, enabling solution quality to be judged in accordance with prioritized indices.We integrate E-dominance with a genetic particle swarm optimizer (GPSO) and maintain an external archive that is updated at each iteration under the E-dominance rules to preserve a diverse, high-quality candidate set. Experiments show that, across five systems, E-GPSO reduces the median MAR and FAR by 6.1% and 7.5% compared with NSGA-II, and by 7.8% and 9.5% compared with MPSO. The reduced solution multiplicity and closer alignment with stated preferences improve decision support, demonstrating the effectiveness of the method for test selection under uncertainty.