Nonlinear integer goal programming models for acceptance sampling

Two lexicographic goal programming models are developed for determining the optimal sample size and acceptance number for acceptance sampling plans in quality control. Both models address the conflicting criteria inherent in such sampling problems, namely the average lot inspection cost and the average outgoing quality. The first model assumes a known constant lot fraction defective, while the second relaxes this assumption and instead assumes knowledge of a prior distribution on the fraction of defectives. A three-phase algorithm is developed which exploits the problem structure in order to find optimal solutions after examining a small percentage of the feasible sampling plans. On a set of 64 test problems the algorithm always found the optimal solution, typically after evaluating only 3-5% (and never more than 9%) of the feasible points.