Multiple item capacitated random yield systems

Two main contributors to the uncertainty of production systems are capacity and random yield; therefore, strategies are needed that incorporate both random yield and the increased effect due to capacity restrictions. This paper utilizes dynamic programming and linear programming transformation to provides a method to specify an optimal decision for a given inventory state. In doing so, it can be empirically shown that the structure of the optimal policy is not an order-up-to policy. Using this method, one only needs to solve a small subproblem, which using LP transformation can be done using commercially available LP solvers without concern for solution time. These results may improve the decision-making capabilities of real-time complex environments since the emphasis is on developing policy rules that are easy to implement in manufacturing applications, as well as service industries such as airlines, healthcare and education. Additionally, the results may be used as a means of evaluating (bounding) existing approximation methods.