Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SA77

2 - Conic Reformulations of Production Planning Models with Clearing Functions Karthick Gopalswamy, North Carolina State University, 3000 Kings Court, Unit G, Raleigh, NC, 27606, United States, Reha Uzsoy We consider the problem of planning the releases into a production system to meet demand in a near-optimal manner. Clearing function models that capture the queuing behavior of production systems under congestion are considered. We provide a conic reformulation of the non-linear constraints in the multi-product problem and provide computational experiments on the solution quality of different model in comparison to traditional LP models and a linear approximation of the non-linear model. 3 - Approximations for Dynamic Multi-class Manufacturing Systems with Priorities and Finite Buffers Girish Jampani Hanumantha, Arizona State University, Tempe, AZ, 85281, United States, Ronald G. Askin Steady-state queueing models are commonly used as a performance evaluation component in aggregate planning. However, with shortened product introduction and life cycles and constantly evolving customer preferences the assumption of stationary demand processes is no longer reasonable. Nonstationary demand processes capture the dynamic nature of modern manufacturing systems well. We present computationally efficient numerical approximations for dynamic manufacturing systems with part priorities and finite buffers. The performance of these approximations is tested for practically sized instances. 4 - Q-learning Based Overhead Transport (OHT) Routing Algorithm for Massive Vehicle Control Young Jae Jang, KAIST, Department of Industrial and Systems Engineer, 291 Daehakro, Yuseong-Gu, Daejeon, 34141, Korea, Republic of, Illhoe Hwang The Overhead Hoist Transport (OHT) is a class of automated material handling system (AMHS) consisting of automated vehicles operating on a network of guideways or specially designed rails. This OHT is widely used in semiconductor fabs and modern high-tech manufacturing production lines. One of the challenges in operating OHT based AMHS for a large scale production line is the development of a routing algorithm controlling massive number of vehicles. We introduce AI based algorithm, Q-learning routing, which effectively control OHT system which consists of more than 500 vehicles. The performance of the algorithm is compared to conventional optimization algorithms. Humanitarian Operations Sponsored: Public Sector OR Sponsored Session Chair: Seyma Guven-Kocak, Georgia Institute of Technology, Atlanta, GA, 30340, United States Co-Chair: Pinar Keskinocak, Georgia Institute of Technology, Atlanta, GA, 30332, United States 1 - Nonprofit Operations: Challenges and Strategies Gemma Berenguer, Purdue University, 1020 Happy Hollow Rd, West Lafayette, IN, 47906, United States In this talk I will provide a definition of nonprofit operations and I will give an overview of this field of Operations Management (OM). I will examine the objectives, major actors and major activities of nonprofit operations and I will identify the major challenges and strategies to address the challenges. A specific analysis of the most relevant applied areas and their unique challenges will also be provided. 2 - Competition Over Funding Resources in Humanitarian Operations Arian Aflaki, University of Pittsburgh, Pittsburgh, PA, United States, Alfonso J. Pedraza-Martinez We study the optimal funding strategy of Humanitarian Organizations (HOs) in the presence of control seeking donors. We find that monopolist HOs can ask for control over donations to improve their operational efficiency; however, competition between HOs limits their ability to collect non-earmarked donations and contributes to lower operational efficiency as well as higher fundraising costs. n SA77 West Bldg 213A Challenges and Strategies in

n SA75 West Bldg 212B Joint Session MAS/Practice Curated: Cybersecurity Analysis and Applications I Joint Session Chair: Paul L. Goethals, United States Military Academy, West Point, NY, 10996, United States 1 - Quantum Correlation Sets and the Limits of Quantum Communication Travis Russell, PhD, Army Cyber Institute, West Point, NY, 10996, United States A quantum correlation is a kind of probability distribution that can be achieved with quantum technology but cannot be achieved by classical probabilistic means. The study of quantum correlations goes back to debates between Albert Einstein and his contemporaries during the formative years of quantum mechanics, yet there are many questions about quantum correlations that remain unanswered to this day. In this talk, we introduce the idea of quantum correlations and discuss some connections with positive semidefinite programming, operator algebras and cyber security. 2 - Cyber Attacks Models Based on Weighted Rooted Trees Elie Alhajjar, Army Cyber Institute, West Point, NY, 10996, United States In recent years, Agnarsson et al. defined a rooted tree model for cyber-security systems based on defense-in-depth and layered-security approaches, as many systems do. In the model, the concepts of penetration cost and acquisition gain are introduced and decision and optimization problems are formulated. In this talk, we look at the problem of storing targets via the intuition that high-value targets should be deeper in the system and outer layers should have higher penetration costs. Moreover, we address the possibility of repositioning targets periodically and how it affects the dynamics of the network. 3 - Locating Arrays: A New Experimental Design for Complex Engineered Systems Violet Syrotiuk, PhD, Arizona State University, Tempe, AZ, United States Screening experiments are used to identify significant factors and interactions on a response. Traditional experimental designs for screening in complex engineered systems requires either restricting the factors considered, which automatically restricts the interactions, or restricting interest to main effects, which fails to consider interactions. To address this problem we propose a locating array (LA) as a screening design. LAs exhibit logarithmic growth in the number of factors allowing an order of magnitude more factors in experimentation than traditional approaches. We apply an LA for screening responses in an experiment on w- iLab.t, a wireless network testbed in Belgium. 4 - Modeling in Cyberspace: Accounting for the Human Dimension Paul L. Goethals, United States Military Academy, 66-B Schofield Place, West Point, NY, 10996, United States, Natalie Scala It is now widely recognized that a large number of our cybersecurity vulnerabilities are attributed to human error of some kind. The models that account for the human dimension, however, are somewhat sparse in the cyberspace literature. This talk will outline some of the existing research lines of effort and discuss potential opportunities for future research in this area. Design and Control of Manufacturing Systems Emerging Topic: Design and Control of Manufacturing Systems Emerging Topic Session Chair: Yunyi Kang, Arizona State University, Tempe, AZ, 85281, United States 1 - Real-time Manufacturing System Performance Evaluation using Machine Degradation Signals Yunyi Kang, Arizona State University, 1850 E. Hayden Lane, Apt 207, Tempe, AZ, 85281, United States, Hao Yan, Feng Ju Evaluating and predicting real-time system performance is important while challenging for modern manufacturing practitioners. With the help of sensor technology, real-time system conditions can be measured and collected. In this work, we build up a Markovian model to evaluate the instant production performance, given the machine degradation signals, in two-machine-and-one- buffer systems. Phase-type distributions are used to mimic the machine operational time distributions based on signal information. A case study is provided to show that such a model can be applied effectively in practice. n SA76 West Bldg 212C

29