Informs Annual Meeting 2017

WC28

INFORMS Houston – 2017

WC28

supply chain under trade credit with random yield. A Stackelberg game model is formulated where the supplier plays the dominated role and determines the wholesale price. And then the retailer announces the order quantity. Via theoretical results, random yield is verified to be negatively associated with the expected profit of the supply chain under trade credit. 2 - Equilibrium Financing in a Supply Chain under Uncertainty Reduction Efforts Qiang Li, The University of Hong Kong, Hong Kong, liqiang@connect.hku.hk, Lap Keung Chu We examine a supply chain consisting of one manufacturer and one retailer, in which the retailer is capital-constrained and can be funded either by a competitive bank, or by the supplier in the form of trade credit. We investigate how the financing equilibrium is affected by the demand variability and the cost associated with reducing demand uncertainty. 3 - Equilibrium Decision of Credit Guarantee and Capital Constraint Supply Chain under Different Ordering Patterns Xiuli He, Associate Professor, UNC-Charlotte, 9201 University City Blvd., Charlotte, NC, 28277, United States, xhe8@uncc.edu We consider a supply chain with a retailer and a capital-constrained manufacturer. The retailer uses credit guarantee to support manufacturer. We derive optimal production and ordering decisions for make to order and make to stock systems. 4 - Data-driven Raw Material Purchasing under Price Volatility Christian Mandl, Technical University-Munich, Munich, Germany, Christian.Mandl@tum.de, Stefan Minner Increasing price volatility at commodity markets requires efficient risk management for commodity-processing firms. We propose a new approach that integrates price prediction and procurement optimization (i.e., contracting decisions) considering real-time causal data like weather and economic indicators. We show that a MILP-based data-driven hedging strategy can lead to significant cost savings for risk-neutral and risk-averse decision makers. However, we observe that the benefit of data-driven optimization strongly depends on model selection. Hence, we additionally combine data-driven optimization with methods from machine learning to increase the out-of-sample performance. 350F Politics/Voting Contributed Session Chair: Joachim Arts, Luxembourg University, Luxembourg, j.j.arts@tue.nl 1 - Optimal Selective Maintenance of Networked Systems with Multiple Missions using Approximate Dynamic Programming Liuquan Li, Tsinghua University, Department of Industrial Engineering, Tsinghua University, Beijing, 100084, China, lq-li16@mails.tsinghua.edu.cn, Chi Zhang Many industrial and military systems have breaks during performing a sequence of missions, which can be utilized for maintenance to ensure the reliability of accomplishing the following mission. However, only a limited number of system components can be maintained during each break due to the limitation of resources, such as time and manpower. We propose a stochastic dynamic programming to study the maintenance of a networked system with general structures, while performing multiple missions. To deal with its complexity, the approximate dynamic programming (ADP) approach is employed. 2 - Determining Aerospace Electronic System Reliability using Regression Methods and MVO Lida Haghnegahdar, PhD Student, State University of New York, Schiller, Binghamton, NY, 13905, United States, lhaghne1@binghamton.edu, Yong Wang, Daniel Trembley The aerospace industry has a long-standing history of embracing new technologies built on predecessor systems of the same function. This research focuses and studies the reliability data of aerospace systems. In this research, three different data sets are considered for analysis by using statistical methods of ridge regression, LASSO, and MVO (Multi-verse optimizer). Ridge regression and LASSO are used to determine trends within the data to enhance prediction of design and component faults. In this paper, we propose the combination of feature selection methods of regression and a meta-heuristic algorithm (MVO) for intelligent optimization. WC30

350D Supply Chain, Optimization Contributed Session Chair: Chaitanya Kaul, Pennsylvania State University-University Park, State College, PA, United States, csk19@psu.edu 1 - How Cost Learning Effect Affects the Channel Decisions Fen Lu, Huazhong University of Science and Technology, 1037 Luoyu Road?Hongshan Distri, Wuhan, China, Wuhan, 430074, China, lufen@hust.edu.cn Based on a supply chain with a manufacturer and a retailer, we construct two- stage Stackelberg game models with a cost-learning effect in different modes (traditional channel mode and dual-channel mode). Backward induction is used to identify the equilibriums of the games. We find that the manufacturer, customers and the whole supply chain always benefit from the cost-learning effect, but the retailer is hurt in the dual-channel mode. Contrast with the traditional channel mode, we find that the optimal production quantity increases and price decreases in the second period in the dual-channel mode. 2 - Aligning the Design of Chemical Multiproduct Batch Plants with the Supply Chain Strategy of the Company Trijntje Cornelissens, Professor, University of Antwerp, Prinsstraat, 13,, Antwerpen, 2000, Belgium, trijntje.cornelissens@uantwerpen.be For chemical batch plants, strategic plant design primarily concerns the number, size and connectivity of raw material, production and storage tanks, and is mostly driven by the capital costs of this equipment. However, the production strategy defines how such a plant is operated, e.g. by dedicating tanks to product families and defining campaign modes. According to SCOR, this production strategy must align with the company’s supply chain strategy and meet the predefined objectives for asset efficiency, cost effectiveness, reliability, responsiveness and flexibility. The aim of this research is to introduce these objectives into plant design models, and study the impact on the plant layout. 3 - An Improved Mathematical Model for Truck Scheduling in Cross Dock Problems with Dock Repeat Truck Holding Pattern Maryam Keshtzari, Graduate Part Time Instructor, Texas Tech University, Lubbock, TX, United States, maryam.keshtzari@ttu.edu, Sasan Khorasani Cross-docking is a material handling and distribution concept.In cross-docking systems, items are transferred directly from receiving dock to shipping dock, i.e. the items will not be stored in a warehouse. In this study, an improved mixed- integer programming model (MILP) is introduced to minimize the makespan of the operations. In fact, the model gives us the best sequence of both inbound and outbound trucks that minimizes total completion time. Furthermore, this model determines the assignment of the items in inbound trucks to outbound trucks. This study endeavors to present a model which is able to solve large scale problems with less computational time compared to the current existing model. 4 - An Agent-based Modelling Approach for Optimizing the Last-mile Logistics Problem Chaitanya Kaul, Pennsylvania State University-University Park, 1926 Waddle Road, Apartment 3, State College, PA, 16803, United States, csk19@psu.edu, Dan Finke The last mile logistics refers to the last leg of delivery in a supply-chain transportation network. More transporters and delivery runs required to fulfil the demand and hence make the last-mile logistic a highly inefficient and time- consuming segment of a supply chain. In our work, we present the formulation of the last mile logistics problem, model it as an agent-based optimization problem and solve it using a heuristic approach. In the problem critical entities like transporters, customers, orders etc. are represented and modelled as a community of agents.

WC29

350E Operations/Finance Interface Contributed Session Chair: Christian Mandl, Technical University-Munich, Munich, Germany, Christian.Mandl@tum.de 1 - Performances of the Supply Chain under Trade Credit with Random Yield

Feng Lin, Xi’an Jiaotong University, Xi’an, China, lfxddz@stu.xjtu.edu.cn, Tao Jia, Y.K., Richard Fung

Previous models always adopt stochastic demand to investigate the optimal solutions of the supply chain under trade credit. However, random yield may lead to the contradiction between supply and demand, and obstruct the integration of material flow and cash flow. Thus, this paper studies the performance of the

514