Test sequencing for sequential system diagnosis with precedence constraints and imperfect tests

• Published in: Decision Support Systems Vol. 103; pp. 104 - 116
• Main Authors: Wei, Wenchao, Li, Hongbo, Leus, Roel
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.11.2017; Elsevier Sequoia S.A
• Subjects: Algorithms; Computer simulation; Diagnostic systems; Diagnostic tests; Imperfect tests; Importance sampling; Job shops; Monte Carlo simulation; Precedence constraints; Sampling; Search algorithms; Sequences; Sequencing and scheduling; Studies; System testing; Tabu search; System testing; Sequencing and scheduling; Imperfect tests; Tabu search; Precedence constraints; Monte Carlo simulation; Importance sampling
• ISSN: 0167-9236; 1873-5797
• DOI: 10.1016/j.dss.2017.09.009

We study sequential system testing with the objective of minimizing the total expected testing costs. The goal is to discover the state of a system that consists of a set of independent components. The state of the system depends on the states of the individual components and is classified as working if at least a pre-specified number of components are working, otherwise it is said to be down. During the diagnostic testing procedure, components are tested one by one, in a pre-specified order. The resulting test sequencing problem is NP-hard with general precedence constraints even when the tests are perfect, in which case a component test always reports the correct state of the component. In this work, we will also consider the additional complication that tests can be imperfect, meaning that a test can report a component to be working when it is actually down, and vice versa. We develop a tabu search algorithm together with a simulation-based evaluation technique that incorporates importance sampling to find high-quality solutions within limited runtimes. •We study sequential system testing to minimize expected testing costs.•We examine the generalization where tests can be imperfect.•We develop a tabu search algorithm.•We incorporate simulation with importance sampling.•We find high-quality solutions within limited runtimes.