2016 INFORMS Annual Meeting Program

MA67

INFORMS Nashville – 2016

2 - Stabilizing Gradient Enhanced Kriging With Sparsity Constraints Peter Qian, University of Wisconsin,re thepeter.qian@wisc.edu Gaussian processes are widely used for emulating computer simulations. It is known that the use of partial derivative information can dramatically improve function estimation. However, the use of partial derivative information comes at the cost of high numerical instability. We investigate an approach to mitigate this instability by exploiting the possibility that some partial derivatives may introduce enough error due to numerical instability to significantly degrade predictive accuracy. Experimental results indicate this procedure can dramatically reduce numerical error in interpolation. Applications to model calibration will also be discussed. 3 - Model Calibration With Censored Data Fang Cao, Georgia Institute of Technology, Atlanta, SGA, United States, fcao6@gatech.edu, Shan Ba, William A Brenneman, Roshan Joseph The purpose of model calibration is to make inference about the unknown parameters of a computer model. The Kennedy-O’Hagan approach is widely used for calibration which accounts for the inadequacy of the computer model while simultaneously estimating the calibration parameters. In many applications censorship occurs when exact outcome of the physical experiment is not observed but is known to fall within a certain region. In such cases KO approach cannot be used directly and we propose a method to incorporate the censoring information when performing calibration. The method is applied to study the stability of liquid and the results show significant improvements over traditional methods. Chair: Murat Kurt, Merck & Co, Inc, 351 N. Sumneytown Pike, North Wales, PA, 19454, United States, murat.kurt7@gmail.com Co-Chair: Anahita Khojandi, University of Tennessee, everykhojandi@utk.edu 1 - Optimal Design Of Hybrid Sequential Testing For A System With Mixtures Of One-shot Units Yao Cheng, Rutgers University, Department of Industrial & Systems Engineering, Piscataway, NJ, 08809, United States, yao.cheng.ise@gmail.com, Elsayed Elsayed Non Destructive Testing is conducted to determine the functionality of the units without permanent damage in order to estimate the units’ reliability. In this presentation, we investigate a system composed of non-identical units with different characteristics and subjected to hybrid reliability testing (Destructive and NDT). It is of interest to optimally design the hybrid sequential reliability testing. After conducting a number of hybrid testing, we decrease the sample size of the destructive testing as the accuracy of reliability metrics estimation improves. Eventually, we only need to conduct NDT only. The efficiency and accuracy of the proposed methods are validated. 2 - Wind Farm Replacement In A Markov Modulated Environment David Abdul-Malak, University of Pittsburgh, dta10@pitt.edu, Jeffrey P. Kharoufeh In this talk we will present a model for jointly replacing wind turbine components in a wind farm setting. Components are assumed to degrade in a shared, exogenous, Markov modulated environment. Continuous state variables and a high dimensional state space cause the problem to be computationally intractable. To overcome these complications, structural results are proven and a reinforcement learning (RL) approach is employed. 3 - An Enhanced Copula-based Prognosis For Proactive Maintenance Of Lithium-ion Batteries Zhimin Xi, University of Michigan-Dearborn, zxi@umich.edu Data-driven prognostics typically requires sufficient offline training data sets for accurate remaining useful life (RUL) prediction for the purpose of proactive maintenance of engineering products. We investigate performances of typical data-driven methodologies when the amount of training data sets is insufficient to better understand the methodology limitation. An enhanced copula-based approach is specifically developed for the scenarios with insufficient run-to-failure training data sets. RUL prediction of lithium-ion batteries in terms of the capacity degradation is employed for the demonstration. MA67 Mockingbird 3- Omni Maintenance and Reliability Planning Sponsored: Quality, Statistics and Reliability Sponsored Session

4 - Optimizing Periodic Inspection Frequencies For a Class Of Stochastically Degrading Systems David Kaufman, University of Michigan- Dearborn, Dearborn, MI, United States, davidlk@umich.edu, Mahboubeh Madadi, Murat Kurt We consider existing models that optimize repair-replacement decisions for systems the degradation status of which follow a discrete time Markov chain over a set of finite states and can be revealed only by costly inspections. Given worse conditions imply higher operation costs, we utilize first-order stochastic dominance relationship among the powers of IFR-structured degradation matrices to propose approximately-optimal periodic inspection decisions that minimize the total expected discounted cost due to operation, repair and inspection. We illustrate our approach through numerical examples. MA68 Mockingbird 4- Omni Panel: IOT-enabled Data Analytics: Opportunities, Challenges and Applications Sponsored: Quality, Statistics and Reliability Sponsored Session Moderator: Kaibo Liu, kliu8@wisc.edu 1 - LoT-enabled Data Analytics: Opportunities, Challenges And Applications Kaibo Liu, University of Wisconsin - Madison, kliu8@wisc.edu The goal of this session is to push the frontier in IoT application and the enabled data analytics research. The session provides a forum where participants can describe current opportunities, identify important problems and areas of application, explore emerging challenges, and formulate future research directions. 2 - LoT And Data Analytics Tobin Jansenberger, American Family Insurance, tjansenb@amfam.com 3 - LoT Analytics Rong Duan, AT&T, rongduan@research.att.com 4 - LoT Data Analytics Subrat Sahu, Caterpillar Inc, sahu_subrat@cat.com 5 - LoT Data Analytics Gul Ege, SAS, Gul.Ege@SAS.com MA69 Old Hickory- Omni Game Theory and Competitive Applications Sponsored: Military Applications Sponsored Session Chair: Brian J Lunday, Assistant Professor, Air Force Institute of Technology, 2950 Hobson Way, WPAFB, OH, 45433, United States, brian.lunday@afit.edu 1 - 1 Vs. (n-1) Modeling For Project Scheduling Interdiction Zachary Little, The Perduco Group, 3610 Pentagon Boulevard A bilevel programming problem is developed for a one-to-many game involving project scheduling interdiction. As a coalition, the many (n-1) adversaries aim to minimize the total cost of a set of project schedules given a time/cost trade-off. The single interdictor aims to maximize this same total cost for the coalition’s project schedules. The modeling framework and use of duality are discussed, with emphasis placed on coalition interaction for this study. Initial results examine the impact of player perceptions on interdictor and coalition decisions. 2 - Approximate Dynamic Programming For Missile Defense Interceptor Fire Control Matthew J Robbins, Air Force Institute of Technology, Wright- Patterson AFB, OH, United States, matthew.robbins@afit.edu, Michael T Davis, Brian J Lunday A missile defense system must protect assets against multiple offensive missile salvos over time. The defender must determine how many interceptors to fire at each incoming missile. We develop a Markov decision process (MDP) model to determine optimal fire control policies. Approximate dynamic programming (ADP) is utilized to explore the efficacy of applying approximate methods to the problem. We obtain policy insights by analyzing subsets of the state space that #110, Beavercreek, OH, 45431, United States, zach.little@theperducogroup.com

144