Multi-objective Decisions on Capacity Planning and Production-Inventory Control under Uncertainty
详细信息 查看全文
- 作者：Lifei Cheng ; Eswaran Subrahmanian ; and Arthur W. Westerberg
- 刊名：Industrial & Engineering Chemistry Research
- 出版年：2004
- 出版时间：April 28, 2004
- 年：2004
- 卷：43
- 期：9
- 页码：2192 - 2208
- 全文大小：232K
- 年卷期：v.43,no.9(April 28, 2004)
- ISSN：1520-5045
文摘
In the presence of a changing and stochastic environment, firms must appropriately choose thetiming and sizes of capacity acquisition as well as the production decisions to replenish inventory,such that various conflicting goals are satisfied. This paper considers multicriteria decisionmaking on joint capacity planning and inventory control under uncertainty. We formulate thisclass of problems into a multi-objective Markov decision process, which simultaneously searchesfor both capacity and inventory policies that optimize multiple objectives. In a previous work,we developed a multi-objective dynamic programming algorithm to solve small problems bypropagating Pareto optimal solutions recursively backward in time. The numerical intractabilityof the rigorous algorithm for large problems necessitates approximation approaches. However,and importantly, on the basis of the optimality equations constructed in the dynamicprogramming framework, we are able to obtain analytical insights into the structure of optimalsolutions for certain simplified problems. For example, through convexity analysis, we showthat such structural policies, as the target interval capacity policy and base-stock inventorypolicy, are optimal for a certain class of problems. We propose using these structural results toform a basis for parametric representation of suboptimal policies for more general problems.This approach greatly simplifies the optimization to that of finding a parameter vector thatdefines these optimal or suboptimal policies. We propose a simulation-based optimizationframework to find the parameters that optimize multiple performance measures. Next we showhow to exploit a multi-objective evolutionary algorithm to find a diverse set of Pareto optimalsolutions in one single run, and finally we present numerical results from an example problemthat we carry from formulation to solution.