Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WB42

3 - The Impact of Product Recall on Advertising in a Duopoly Market Arka Mukherjee, PhD Candidate, Concordia University, Montreal, QC, H3H 0A1, Canada, Satyaveer S. Chauhan Product recalls negatively affect the brand image of a responsible firm. Brand image is positively influenced by advertising. We study the optimal advertising strategies of firms in a duopoly environment with an impending product recall. We find that the optimal advertisings for both the firms are dependent on several parameters like initial brand value, pre-crisis and post-crisis profit margin, advertising effectiveness, damage effect of the recall and the recall probability. 4 - Modeling the Value of Agents in Supply Chains of Malaria Rapid Diagnostic Test Kits with Decision Analysis Gilberto Montibeller, Professor of Management Science, Loughborough University, School of Business & Economics, Loughborough, LE11 3TU, United Kingdom Malaria poses one of the greatest global health challenges. Accurate malaria diagnosis, with rapid diagnostic tests (RDTs), is critical for treating the disease. In this project we developed and deployed a multi-attribute value model for agents along the RDT kit supply chain (manufacturers, first line buyers, and retailers) in Uganda. The model represents the different priorities, incentives, and preferences of each agent in the supply chain and supported the design of high value bundles for RDT kits. The decision analysis provided a systematic evaluation of alternatives and enabled reflection and learning among the stakeholder. Queueing Models Contributed Session Chair: Iqra Ejaz, Texas A&M University, Bryan, TX, 77801, United States 1 - Determining the Size of Oscillations in Queues with Customer Choice and Delayed Information Sophia Novitzky, Cornell University, 136 Hoy Road, Ithaca, NY, 14850, United States With the advancement of online technologies, it is common for the service systems to provide waiting time or queue length information to customers. This information allows the customers to determine whether to remain in line or in the case of multiple lines, which line to join. However, there is usually a delay associated with the waiting time information: either the information is not provided in real time or it takes the customers travel time to join the service after having received the information. Previous work shows that if these delays are large enough, unwanted oscillations in the queue lengths can occur. In this paper, we develop two methods for approximating the amplitude of these oscillations. 2 - Dynamically Scheduling and Maintaining a Flexible Server Jefferson Huang, Assistant Professor, Naval Postgraduate School, Monterey, CA, United States, Douglas Down, Mark E. Lewis, Cheng-Hung Wu Advances in sensing technologies have made condition-based maintenance (CBM) in flexible manufacturing systems more economically feasible. We consider the problem of jointly making CBM and scheduling decisions for a server in such a system. In the context of a queueing model of this problem, we show that a natural scheduling heuristic can be extremely suboptimal. We also provide conditions under which one can restrict the search for an optimal policy to policies that (1) schedule according to a static priority rule, and (2) exhibit a certain monotonicity property with respect to the maintenance decisions. 3 - Three-moment Approximation for the Mean Queue Time of a GI/G/1 Queue Kan Wu, Nanyang Technological University, School of MAE, 50 Nanyang Ave, Singapore, 639798, Singapore, Sandeep Srivathsan, Yichi Shen The approximation of a GI/G/1 queue plays a key role in the performance evaluation of queueing systems. To improve the conventional two-moment approximations, we propose a three-moment approximation for the mean queue time of a GI/G/1 queue based on the exact results of the H2/M/1 queue. The model is validated over a wide range of numerical experiments. Based on the paired t-tests, our three-moment approximation outperforms the two moment ones especially when both service time and inter-arrival time variabilities are greater than one. 4 - Optimization of Nursing Teams with Patient Assignment Using a Queueing Theory Approach Parisa Eimanzadeh, Wichita State University, Wichita, KS, 67220, United States, Ehsan Salari Inpatient units are typically staffed with nursing teams that consist of nurses at different skill levels providing care to patients. This research aims at developing staffing models to determine appropriate staff levels and skill mix for the nursing teams by explicitly incorporating the patient assignment decision. In particular, queueing theory and multi-criteria optimization are combined to find the optimal skill-mix configurations of the nursing teams and the corresponding patient n WB42 North Bldg 227A

assignments that minimize staffing costs and nurse burnout while ensuring timely delivery of nursing care. 5 - An Efficient Algorithm for Non Markovian Two-node Cyclic Network Muhammad El-Taha, Professor, University of Southern Maine, Department of Mathematics and Statistics, 96 Falmouth Street, Portland, ME, 04104-9300, United States, Bacel Maddah Computing the steady state probability distribution of a non-Markovian two-node closed queueing cyclic network is known to be computationally challenging. In this talk, we propose a new efficient convolution method that is significantly more efficient than existing algorithms. 6 - Condition-based Maintenance of Queues with Degrading Servers for Stochastic Service Times Iqra Ejaz, Texas A&M University, College Station, TX, 77843, United States, Michelle M. Alvarado, Natarajan Gautam, Nagi Gebraeel, Mark Lawley We derive an analytical model for condition monitoring of a single server queue with Markovian degradation, Poisson arrivals, and general service and repair times. Stability conditions and performance measures (e.g., average queue length, average degradation.) are derived through steady state analysis. An optimal repair decision model is presented that minimizes an objective function with four costs: repair, catastrophic failure, quality and holding. We develop and verify a simulation model, perform a sensitivity analysis, and show insights learned from relaxing underlying assumptions. Chair: Itir Z. Karaesmen, American University, Kogod School of Business, Room 212, Washington, DC, 20016, United States 1 - Two-stage Bond Portfolio Optimization under Different Scenarios Nasser Alreshidi, PhD Student, Florida Institute of Technology, 150 W. University Blvd, Melbourne, FL, 32901, United States, Ersoy Subasi, Munevver Mine Subasi, Andras Daniel Prekopa Following Hodges and Schaefer’s (1977) deterministic bond portfolio model, where the objective is to minimize the cost of bond portfolio, the bond portfolio optimization has become one of the attractive areas of research in finance literature. In this work we propose a new bond portfolio model as a two-stage stochastic programming problem, where a decision maker may optimize the cost of bond portfolio, while deciding which bonds to sell, which bonds to hold, and which bonds to purchase from the marketplace based upon present market conditions under different scenarios. We present the application of our model on real-world datasets. 2 - Production and Capacity Utilization Strategies in Supply Chains for Complex Engineered Products Ashesh Sinha, Kansas State University, Manhattan, KS, United States, Ananth Krishnamurthy We analyze a multi-product manufacturing systems where individual products can be made either at a shared in-house manufacturing facility or at dedicated facilities of external subcontractors. Using Markov decision process models, we determine the optimal policy and derive the set of conditions that partitions the state space into regions and analytically characterize optimal policies and value function in each region. 3 - A Game-theoretic Two-stage Stochastic Programing Model to Protect Cyber Physical Systems Against Attacks We present two-stage stochastic programming models that determine the check blocks to assign to measurement signals and controls to secure a cyber physical system (CPS). The models incorporate uncertainty on the number of signals to be protected over the horizon considered and on the effectiveness of the check blocks. We illustrate the superiority of the proposed approach by computing the value of the stochastic solution and by comparing to other assignment strategies. Results are validated with a simulation to emulate a component in the operation of an autonomous vehicle and with experiments on a mini-boiler testbed that operates on a closed loop and interacts with a Matlab-Simulink program. 4 - Integrating Logistics and Bidding Decisions in B2B Markets Itir Z. Karaesmen, American University, Kogod School of Business, Room 212, Washington, DC, 20016, United States We build an analytical model for brokers that procure and sell goods in a B2B market. The broker submits bids to procure goods from several suppliers. The outcome of the bids are uncertain. After the goods are procured, they are shipped to meet the customer demand. We show how the optimal bids can be determined and computed. Clara Novoa, Associate Professor, Texas State University, 601 University Dr, San Marcos, TX, 78666, United States, Khan Sidiqqe, Mina Guirguis, Alireza Tahsini n WB43 North Bldg 227B Stochastic Optimization I Contributed Session

456