接單生產環境下的離散動態訂單允收模型

Discrete Dynamic Order Admission Model under MTO B2B Environment

武為棣(Wei-Di Wu)

指導教授 ： 蔣明晃

國立臺灣大學/管理學院/商學研究所/博士(2010年)

https://doi.org/10.6342/NTU.2010.00279

摘要

對於接單生產的製造業者，季節性需求等因素常造成產能緊迫供不應求的問題，接單允收控制乃是一項解決該問題的作業策略，相較於超時生產、產能擴充、或是委外加工，此一策略在施行面上具有低成本的優點。本研究的貢獻乃是針對該環境下的訂單允收問題進行模式化分析，模式考量到潛在訂單集合的有限性、訂單產能需求的批量性、以及需求分配間的異質性，本訂單允收問題可化約為離散時間離散狀態的馬可夫決策模型。
本文第三節至第五節為本研究第一部份之分析內容，在問題的解析上，首先論證確定型馬可夫策略的存在性與最適性，最適策略DSKP型態為一動態利潤門檻值策略；透過實驗設計，本研究比對DSKP策略、先進先出策略策略、以及動態產能配額策略的績效，在多種需求結構因子的影響之下，統計結果顯著地呈現出DSKP策略對於提升獲利能力所具備的良好潛力。
本文第六節分析訂單大小估計偏誤所造成的影響。第一部份的研究已然證明DSKP策略在理論上具備提升獲利的功效，然而，需求預測的準確性可能影響到此一策略的落實。在實務操作上，可以觀察到企業的計畫人員對於需求預測的估計偏誤存有戒心，此般保守心態有可能影響經理人對於DSKP策略的接受度。有鑑於此，第二部份的研究採用實驗模擬對無偏誤DSKP策略、有偏誤DSKP策略、以及先進先出策略的績效進行交叉分析。統計結果揭示DSKP策略具備相當程度的容錯性，此一結果進一步強化該策略的操作成效。

關鍵字

訂單允收控制 ； 馬可夫決策 ； 隨機動態背包問題 ； 收益管理 ； 估計偏誤

並列摘要

Swarming demands or demand seasonality may cause a stringent capacity issue to MTO (make to order) B2B (Business to Business) manufacturers. The finite capacity of production lines may not fulfill all the key-account orders. Although manufacturers are reluctant to loss any business, some of the order requests must be turned away. Desirable order admission control is a tactic of minimal installation cost to handle this stringent capacity issue. This research studies how a dynamic admission policy should be designed to maximize the manufacturer’s profit.
This admission problem faced by a MTO B2B manufacturer is characterized by the finite potential order set and heterogeneous batch-size demands. This research uses a Dynamic and Stochastic Knapsack Problem (DSKP) to describe this order admission issue. To our knowledge, there is no mathematical model before discussing the order finiteness and heterogeneity. In the following content, the existence and optimality of the Markov deterministic policy is verified. It is shown that the optimal policy, named as the DSKP policy, is a kind of reward-threshold policy. Furthermore, the performances of three types of policies (DSKP and the other two baseline policies) are compared to each other in numerical experiments. The results reveal the great improvement brought by the DSKP policy.
Furthermore, we continue to study the steadiness and effectiveness of the DSKP policy. Although the DSKP policy is theoretically optimal to the admission problem, its “real” performance in practice may be sensitive to the unavoidable estimation error of order forecasts. For this reason, the impact of order estimation error is investigated thoroughly. A post analysis is performed which shows significant statistical difference among the optimal unbiased DSKP policy, biased DSKP policy, and FCFS (first come first serve) policy. The results reveal the fault tolerance of the DSKP policy in most scenarios. On account of this analysis, we specify the factors that should be watch for carefully because, under the high-level impact of these factors, applying DSKP policy may be risky.
In sum, this research contributes on modeling the dynamics of order structure and the corresponding decision process for a MTO B2B manufacturer. The proposed order admission policy – the DSKP policy - is the optimal solution maximizing the manufacturer’s expected total profit. The properties and algorithm of the DSKP policy are verified and presented in this research. The impact of estimation error to this policy is reviewed and the statistical tests validate its steadiness and effectiveness.

並列關鍵字

Order Admission ； Markov Decision Problem ； Dynamic & Stochastic Knapsack Problem ； Revenue Management ； Estimation Error

參考文獻

3.Balakrishnan, N., Patterson, W., and Sridharan, V., 1999, Robustness of capacity rationing policies, European Journal of Operations Research, 115(2), 328–338.

4.Ball, M. O., Chen, C. Y., and Zhao, Z. Y., 2004, Available to promise, In: D. Simchi-Levi, S. D. Wu and Z. M. Shen, (Eds.). Handbook of Quantitative Supply Chain Analysis: Modeling in E-Business Era, Kluwer, Boston, 447–481.

1.Altman, E., Jimenez, G., and Koole, T., 2001, On optimal call admission control in a resource-sharing system, IEEE Transactions on Communications, 49(9), 1659–1668.

2.Balakrishnan, N., Sridharan, V., and Patterson, W., 1996, Rationing capacity between two product classes, Decision Sciences, 27(2), 185–213.

5.Barut, M. and Sridharan, V., 2004, Design and evaluation of a dynamic capacity apportionment procedure, European Journal of Operational Research, 155(1), 112–133.