Aggregate capacitated production planning in a stochastic demand environment
Most production planning models suffer from one or more of the following limitations: they either (i) ignore the stochasticity of demand, or (ii) ignore the relationship between load and lead time by treating capacity as an upper limit or using fixed lead times, or (iii) ignore the role of work in process (WIP) inventory in buffering against uncertainty; or (iv) make all the decisions at the beginning of the planning horizon with no simple way to adjust the decisions other than re-solving the problem on a rolling horizon basis. In this thesis, using: (i) clearing functions to capture the load and lead time dependency, (ii) chance constraints to capture both demand stochasticity and WIP contribution towards buffering, and (iii) linear decision rules to provide flexibility in the decision making mechanism, we propose a tractable model for single stage single product systems that addresses the limitations of the existing models. An extensive computational study of the performance of the proposed model under different non-stationary demand conditions and a multi-segment clearing function is performed. A comparison of the performance of the proposed model to some of the commonly used models under the different conditions is presented. The computational experiments indicate that the proposed model is capable of yielding high service levels with lesser cost than conventional approaches.^
Reha Uzsoy, Purdue University, Hong Wan, Purdue University.