要闻动态
当前位置:  首页 要闻动态 学术活动 研究院论坛:“云上”NO.4 Robust Capacity Planning with Product Substitution
研究院论坛:“云上”NO.4 Robust Capacity Planning with Product Substitution
2020年05月29日

报告题目:Robust Capacity Planning with Product Substitution

报 告 人:郝兆伟

报告时间:2020年05月29日(周五),13:30-13:00

报告地点:线上直播

主办单位:现代供应链管理研究院

【报告人简介】

郝兆伟,北京大学管理学博士。曾在加州大学伯克利分校工业工程与运筹学系进行联合培养,在新加坡国立大学商学院从事博后研究。主要研究领域包括智慧城市运营管理,稳健性优化,库存和供应链管理,产品创新管理,供应链金融等。其研究成果已发表在Production and Operations ManagementOmega等期刊。

【摘要】

Product substitution is broadly employed in a firm’s practice to match supply with demand. When a low- quality product is out of stock, the firm may substitute it with a high-quality product for the customer. In this paper, we develop a distributionally robust optimization model to determine the initial capacity planning decision considering product substitution. For the single-period model, the key challenge lies in how to characterize the extreme flows of a capacitated network. We propose an algorithm that can cut the capacitated network into a series of uncapacitated subnetworks whose extreme flows can be effectively characterized. The robust model can then be reformulated as a second-order cone program (SOCP). Although in general there can be an exponential number of constraints of the SOCP, we show it is polynomially solvable for some important cases. For the case where the unit upgrade revenue is equal to the selling price of the downward product, we find that the profit margin plays an important role in determining how the product substitution impacts the firm’s initial ordering decision. For the multi-period model, we derive a tractable approximation formulation using the linear decision rule (LDR) techniques. We then develop a theoretical performance guarantee for the LDR by deriving an upper bound on the expected profit of the exact robust optimization model. Our extensive numerical studies show that the LDR is more capable to address temporal correlation and adverse distribution in the out-of-sample demand as compared to the dynamic programming (DP) approach. Computationally, LDR is shown to be much more efficient than DP. 




联系我们
  • 地址:中国·辽宁·大连市沙河口区尖山街217号
  • 邮编:116025
  • 电话:(+86)0411-84713573
  • 电子邮箱:isca_2016@dufe.edu.cn
现代供应链管理研究院