Dynamic quality control in assembly systems

We study an assembly system, in which each end product consists of two different components. The components are produced in batches, with possible defective units. The proportion of defective units in each batch is itself a random variable, known only in terms of its distribution. Defective units can be identified and corrected through an inspection-repair procedure, before final assembly taking place. We are concerned with developing an optimal inspection policy that minimizes the total expected cost due to inspection, repair, and warranty. Particularly, we focus on policies that have simple threshold structures and a near-optimal performance.