Attribute Acceptance Sampling as a Tool for Verifying Requirements Using Monte Carlo Simulation

ABSTRACT

Developed by quality engineers as an efficient statistical approach to verifying the acceptability of procured items within production environments, acceptance sampling can be generalized to other verification problems which similarly rely on the outcomes of stochastic sampling experiments. This article illustrates how the techniques and perspectives of attribute acceptance sampling can be adapted to the verification of probabilistic design requirements using Monte Carlo simulation. We consider requirements expressed as limit standards, specifying the performance indicator for conforming Monte Carlo trials; the minimum limiting proportion of conforming trials; and the maximum risk of accepting a nonconforming design. For such requirements, an attribute sampling plan prescribes the number of simulation replications that must be run and the number of nonconforming replications that can not be exceeded. The derivation and analysis of single-sample attribute acceptance plans for the mass-delivery requirement for NASA's Constellation Program (CxP) is provided as an example. More demanding (but potentially more efficient) alternatives to attribute acceptance sampling are suggested for applications with limited budgets for simulation trials.

KEYWORDS:

• acceptance sampling
• discrete-event simulation
• limit standards
• Monte Carlo simulation
• requirements verification
• statistical quality control

ACKNOWLEDGEMENTS

Special thanks are due to Timothy C. Adams at NASA Kennedy Space Center and Chester J. Everline at the Jet Propulsion Laboratory for their many direct and indirect contributions to the work reported in this article. The authors also are deeply appreciative for the insights provided by the other members of the CxCEF Verification of Probabilistic Requirements Team, especially to Stephen J. Scotti at NASA Langley Research Center for his consistent and patient leadership.