Cost optimization in supply management policies for assembly systems with random component yield times

This paper addresses the problem of cost minimization in multi-product assembly systems where component yield times are subjected to random uncertainties. At the start of each cycle time, two interrelated decisions have to be made: 1) how many products to assemble for each product type; and 2) how many components to order. The total cost associated with these decisions is the sum of the inventory-holding cost and the product-backloging cost. A fast and efficient gradient-descent algorithm is applied to minimize the total cost over a given time horizon. The algorithm is based on the infinitesimal perturbation analysis (IPA) gradient estimation technique, and on continuous-parameter modeling formulation of the system's dynamics. Numerical experiments corroborate the conceptual developments.