Informs Annual Meeting 2017

WC60

INFORMS Houston – 2017

5 - Joint Location and Inventory Analysis in IoT-based Service Parts Logistics Systems Murat Karatas, University of Texas-Austin, Austin, TX, 78703, United States, mkaratas@utexas.edu, Erhan Kutanoglu We investigate a joint location and inventory problem in service parts logistics (SPL) taking advantage of data on equipment condition via Internet of Things (IoT) technology. Continuous condition data may make replacing the parts before failure possible, potentially saving time and money, as opposed to the conventional approach of issuing replacement parts upon failures. We propose a model that captures the optimum balance between increased demand due to parts replaced more frequently and time/cost savings due to non-emergency shipments. Our preliminary results quantify the effects of this tradeoff on location and inventory decisions as well as time-based service levels.

We discuss parallel implementations of RLT2 formulation and branch-and-bound algorithm for solving the Quadratic Assignment Problem (QAP). Our parallel architecture consists of NVIDIA Graphics Processing Unit (GPU) clusters on the Blue Waters supercomputer at the University of Illinois at Urbana-Champaign. We propose a “distributed” Dual Ascent algorithm for the GPUs, which shows excellent parallel speedup, and can effectively solve some of the well-known QAPs from the literature. 2 - Comparing Two Goods-to-person Order Picking Systems for Online Retailing Francisco Jose Aldarondo, PhD Candidate, University of Michigan, 3512 Green Brier Boulevard, Apartment 497C, Ann Arbor, MI, 48105, United States, faldaron@umich.edu Using simulation modeling and an on-line retail setting, we compare the performance of two types of goods-to-person order picking (OP) systems, namely, the Kiva system and the Miniload-AS/RS with a conveyor loop (to connect the pick stations). The two systems are compared on the basis of quantitative factors such as expected throughput (line items picked per hour), expected picker and material handling equipment utilization, and order completion times. We also compare the two systems in terms of qualitative factors that are relevant for OP systems. 3 - An Integrated Supply Chain Design for Additively Manufactured Products Sudipta Chowdhury, Mississippi State University, 260 McCain Engineering Building, ISE Department, Starkville, MS, 39762, United States, sc2603@msstate.edu, Mohammad Marufuzzaman, Linkan Bian Despite the promising features of AM technologies, make-or-buy decisions are not straightforward. Moreover, decision regarding the optimal number of layers for the AM product is also an issue that has not been addressed before. Based on the budget or quality requirements, number of layers for the product may vary which affect the supply chain cost. A stochastic model is proposed to quantify both the process and supply-chain level costs associated with the production of AM parts. Two algorithms, i.e., hybrid Sample Average Approximation (SAA) and Progressive Hedging Algorithm (PHA) and hybrid SAA and Adaptive Large Neighborhood search (ALNS) based algorithm are developed to solve the problem. 4 - Solving the Student Assignment and School Bus Routing Problems Sunderesh S. Heragu, Oklahoma State University, 322 Engineering North, School of Industrial Engineering & Management, Stillwater, OK, 74078, United States, sunderesh.heragu@okstate.edu, Harshwardhan Rathod In this presentation, we address the assignment of students to bus stops as well as the school bus routing (SBRP) problems. The student assignment problem (SAP) is solved using a specialized branch-and-bound algorithm. The result is provided as input to the SBRP. The latter is solved using a meta heuristic. We then apply both the SAP and SBRP to the Stillwater middle student assignment and school bus routing problems. 370A Forecasting Contributed Session Chair: YoungJun Park, Seoul National University, Seoul, pyjoon90@snu.ac.kr 1 - Predicting Customer Activity by Cumulative Purchase Quantity and Consumption Rate Huan Yu, PhD candidate, University of Science and Technology of China, Hefei, China, huany@mail.ustc.edu.cn, Yugang Yu, Hanqin Zhang When analyzing customer activity, usually one assumes that the independence between inter-purchase time, purchase quantity and lifetime. However, many data sets we have collected demonstrate that this independence assumption may not hold. For example, inter-purchase time and purchase quantity may depend on each other through customer consumption rate, and lifetime and inter- purchase time may depend on each other through cumulative purchase quantity. Considering these dependences, we propose a new stochastic model of predicting customer activity. The proposed model can formulate our data sets very well, which cannot be fitted by the existing models that assume the independence assumption. WC60

WC57

362D Scheduling Contributed Session

Chair: Mansoor Shekarian, North Carolina A&T State University, Greensboro, NC, United States, mshekarian@aggies.ncat.edu 1 - Total Earliness and Tardiness in a Two-machine Flowshop Jeffrey Schaller, Eastern Connecticut State University, 83 Whindham Street, Dept of Business Administration, Willimantic, CT, 06226-2295, United States, schallerj@easternct.edu, Jorge Valente This paper considers the problem of scheduling jobs in a two-machine permutation flow shop with the objective of minimizing total earliness and tardiness. Unforced idle time is considered in order to reduce the earliness of jobs. 2 - Utility Based Priority Dispatching Rules in Complex Dynamic Job Shops: the Case of Customer Balking Kevin D.Sweeney, Assistant Professor, Sam Houston State University, Box 2056, Sam Houston State University, Huntsville, TX, 77341, United States, ksweeney@shsu.edu, Stanislaus Solomon, William A. Ellegood We investigate the performance of utility based priority dispatching rules in a reactive complex job shop environment. In particular, we focus on job shops defined by two characteristics: (1) customers that can potentially balk to an alternative job shop if a certain threshold for service or utility is not met, and (2) job shops characterized by heterogeneous classes of jobs which each receive different amounts of utility from the job shop. Within these job shops, we look at the impact of using different priority dispatching rules under different ratios of utility of heterogeneous classes of jobs, processing times of different classes of jobs, and a range of attractiveness levels of the alternative. 3 - A Decision Model for Two Objective Job Shop Scheduling Problem with Sequence Dependent Setup Times Mansoor Shekarian, Graduate Student, North Carolina A&T State University, 3100 N Elm St, Apt #22L, Greensboro, NC, 27408, United States, mshekarian@aggies.ncat.edu, Adel Aazami, Mahour Parast The job shop scheduling problem (JSP) is one of the most difficult problems in traditional scheduling problems. When an operation happens in a machine, it is necessary to consider a sequence-dependent setup time (SDST). This research develops a two-objective model, including minimizing the Makespan and maximum tardiness. The proposed mixed integer nonlinear programming (MINLP) model is converted into a MILP to achieve a globally optimal solution. The -constraint method is used to solve the model. A set of numerical data from a real case study is investigated to show the model’s efficiency and flexibility. These results have a great value for managers, especially in a manufacturing industry. 362E Applications of Facility Logistics - II Sponsored: TSL, Facility Logistics Sponsored Session Chair: Francisco Jose Aldarondo, University of Michigan, 3512 Green Brier Boulevard, Apartment 497C, Ann Arbor, MI, 48105, United States, faldaron@umich.edu 1 - Solving the Quadratic Assignment Problem using Graphics Processing Unit Clusters on the Blue Waters Supercomputer Rakesh Nagi, U. of Illinois at Urbana-Champaign, Department of Industrial & Enterprise Systems, 117 Transportation Building, MC- 238, Urbana, IL, 61801, United States, nagi@illinois.edu, Ketan Date WC58

523