Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WB72

2 - Optimization Modeling in Matlab Aurele Turnes, MathWorks, Natick, MA, 01760, United States, Steve Grikschat, Paul Kerr-Delworth, Adam Hug The problem-based workflow for optimization has made it much easier to model and solve an optimization problem in MATLAB. First, optimization variables are identified and defined. They can be N dimensional and can be indexed similar to standard MATLAB arrays. Next, familiar MATLAB operators can be used to define the objective and constraints as expressions of the optimization variables. Finally, the optimization solver is automatically selected based on the type of objective and constraints. Large and complex optimization problems expressed in this way are more compact and readable. We will review the problem-based workflow and demonstrate new capabilities with examples. 3 - Constraint-based Sequential Pattern Mining Using Multi-valued Decision Diagrams Amin Hosseininasab, Tepper School of Business, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, Willem van Hoeve, Andre Augusto Cire Constraint-based sequential pattern mining (SPM) has gained recent attention as it aims to find patterns interesting to the user, rather than an arbitrary set of frequent patterns. We propose a general framework for imposing non-monotone constraints into SPM. Such constraints impose restrictions on different item attributes, and are challenging or impossible for current mining algorithms. We use multi-valued decision diagrams to compactly model the database, and use it to store critical information required to impose all constraints. We compare our algorithm to specialized SPM algorithms in the literature, and present a number of novel insights from a real-world database. n WB72 West Bldg 211A Practice- Operations Management III Contributed Session Chair: Yiwei Huang, Pennsylvania State University, Mars, PA, 16046, United States 1 - Impact of Shorttermism on Capacity Investment in Competitive Markets Xiaoyan Liu, PhD Student, Cornell University, Sage Hall, Ithaca, NY, 14853, United States, William Schmidt We study how a firm’s short-term interest in its capital market valuation will affect its capacity decision and long-term profits in the presence of a competitive entrant. We show that the firm with private demand information prefers to signal a high (low) demand when short-termism is high (low). Moreover, pooling equilibria can occur with either low or high levels of short-termism, and both firm types will under-invest in a pooling outcome when short-termism is sufficiently low. We find that the impact of short-termism on long-term profits changes with the firm type and the equilibrium type. 2 - A Wholesale Supply Contract with a Secondary Price Commitment Ayhan Aydin, George Mason University School of Business, 4400 University Drive, MS 5F4, Fairfax, VA, 22030, United States, John R. Birge, Izak Duenyas In procurement from low-cost but less flexible supply options, especially of products with highly variable demand and of seasonal nature, advanced auctions and non-linear compensation schedules can be used to correct inefficiencies due to double marginalization, information asymmetry, and demand uncertainty. We introduce an alternative simple-to-administer contracting scheme that requires minimal effort to implement, which can be used especially when the suppliers do not offer revenue sharing, buy-back, or real option contracts. We study how it allows the buyer to benefit further from low-cost distant supply options. We show that a truly low-cost supplier can also benefit from such a scheme. 3 - Quality Control Strategies in Two-sided Market Lyu Gaoyan, Peking University, No.5 Yiheyuan Street, Beijing, 100871, China, Lihua Chen Consumers focus on product quality increasingly. We design two quality control strategies to help two-sided platform improve its product quality. Using Stankburg method, we compare the optimal profits with two quality control methods. The result shows that taking any strategy is better than taking no action. Moreover, taking both strategies will bring higher social welfare.

n WB70 West Bldg 106B Reliability II Contributed Session Chair: Oluseye Fafiolu, Lamar University, Beaumont, TX, Texas, United States 1 - Statistical Inference on Remaining Useful Life in a Two-phase Degradation Model Under Gamma Process Hon Keung Tony Ng, Professor, Southern Methodist University, 3225 Daniel Avenue, Department of Statistical Science, Dallas, TX, 75275-0332, United States, Man Ho Ling, Kwok-Leung Tsui Due to physical and chemical changes that take place with usage and with age, a significant degradation rate change of a product usually exists. To accurately predict the remaining useful life prediction for products with two-phase degradation, we incorporate a probability distribution of the time of rate change into the gamma degradation model. A Bayesian approach and a frequentist approach are proposed for statistical inference of the remaining useful life. A simulation study is used to evaluate the performance of the developed methodologies and a real data set on light emitting diodes is presented to illustrate the application of the proposed model. 2 - Condition-based Maintenance and Production for Systems with Overcapacity Michiel uit het Broek, MSc, University of Groningen, Groningen, Netherlands, Ruud Teunter, Bram de Jonge, Jasper Veldman Many multi-unit systems face significant economic dependencies for performing maintenance, e.g., expensive vessels are needed to perform maintenance at offshore wind farms. Therefore, it is often cost efficient to cluster maintenance for several units in the system. However, when maintenance for units with different degradations levels is clustered, then maintenance is performed too early for the low deteriorated ones or too late for the highly deteriorated ones. In such situations, an interesting question is whether it can be profitable to adjust the production rates in order to synchronize the deterioration processes for the different units. 3 - Joint Optimization of Job Scheduling and Condition-based Maintenance Bram de Jonge, University of Groningen, P.O. Box 800, Groningen, 9700 AV, Netherlands We consider a production facility that processes incoming jobs of various lengths and that requires maintenance due to deterioration and ultimately failure. We present an approach to determine whether the system has a steady state. Based on maintenance costs and total weighted flow time costs we formulate a Markov decision process to determine optimal maintenance and job scheduling policies. We provide insights on how job scheduling and maintenance planning interact. 4 - Physics-of-failure Based Reliability Prediction Model for Rail Track Geometry Oluseye Fafiolu, Lamar University, 4400 Mlk Pkwy, Beaumont, TX, 77701, United States, Jaeyoung Cho, Maryam Hamidi We propose a rail track geometry reliability prediction model based on physics-of- failure approach. The failure mechanisms that causes rail track geometry degradation are incorporated into the model. We expect that the reliability model predicts accurately the optimum working hours under normal operating conditions of the rail track geometry prior to maintenance actions being required. n WB71 West Bldg 106C Advances in Computational Optimization Sponsored: Computing Sponsored Session Chair: John W Chinneck, Carleton University, Ottawa, ON, K1S 5B6, Canada 1 - Linear Programming Based Sparse Solutions Revisited John W. Chinneck, Carleton University, Systems and Computer Engineering, 1125 Colonel By Drive, Ottawa, ON, K1S 5B6, Canada Finding sparse solutions for underdetermined linear systems is the key step in compressive sensing and a number of other important problems. One category of solutions uses linear programming, e.g. Basis Pursuit. However there are other ways to formulate the LP which have not been explored and which lead to different algorithms with different characteristics. The talk presents a variety of new LP-based formulations and associated algorithms, along with numerical evaluation and comparisons.

467