2016 INFORMS Annual Meeting Program

WB70

INFORMS Nashville – 2016

WB68 Mockingbird 4- Omni Spatiotemporal Data based Quality Control Methods in Manufacturing Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Mohammed Saeed Shafae, Virginia Tech, 112 Durham Hall, 1145 Perry Street, Blacksburg, VA, 24061, United States, shafae1@vt.edu Co-Chair: Lee Wells, Assistant Professor, Western Michigan University, E-208 Floyd Hall, 1903 W Michigan Ave, Kalamazoo, MI, 49008, United States, lee.wells@wmich.edu 1 - Spatial Discrete Model For Clustered Defects On Wafer Maps Hao Wang, Tsinghua University, w-h14@mails.tsinghua.edu.cn Yield, which contributes to increase the process quality, is a key concern in the fabrication of semiconductor wafers. In this paper, we propose a novel discrete spatial model based on discrete spatial defects data on wafer maps for analyzing and predicting the yields at different chip locations. Based on the Bayesian framework, we adopt a generalized linear model that considers both the spatial coordinates and random spatial error, thereby significantly improving the performance of the model. The experimental results show that the generalized linear Poisson model that considers both the spatial coordinates and random spatial error offers an improved fit to spatially correlated wafer map data. 2 - Spatial Models In Metal Additive Manufacturing Bianca Maria Colosimo, Politecnico di Milano, Milano, Italy, biancamaria.colosimo@polimi.it, Marco Grasso Additive manufacturing is more and more often moving from prototyping to production and this is why new methods and tools for quality inspection and monitoring are needed. In order to model shapes and internal defects, spatial models can be used. This talk shows how spatial modeling can be used to model complex shapes and internal defects (e.g., porosity) in metal additive manufactured products. 3 - Reduced-dimension Mcusum Chart For Spatio-temporal Surveillance Junzhuo Chen, Georgia Institute of Technology, Atlanta, GA, United States, jz.chen@gatech.edu, Seong-Hee Kim, Yao Xie In spatiotemporal surveillance, control chart with scan statistics is a powerful method. Usually calculating monitoring statistic requires observations from the whole monitoring area and the full covariance matrix inversion. However, if the dimensionality is high, implementation can be challenging. First, it is hard to estimate the full covariance matrix. Second, the computation of matrix inversion is expensive. Finally, collecting data from all the sensors may be costly. To address such issues, we propose the Reduced-Dimension MCUSUM chart that constructs the monitoring statistic using measurements from a small group of locations. We conduct simulations to study performance of the method. WB69 Old Hickory- Omni Network Design and Optimization Sponsored: Telecommunications Sponsored Session Chair: Richard Li-Yang Chen, Sandia National Laboratories, 7011 East Ave, Livermore, CA, 94550, United States, rlchen@sandia.gov 1 - Modulation Design For MIMO HARQ Channel Hans Mittelmann, Arizona State University, mittelma@asu.edu, Wenhao Wu, Zhi Ding Modulation diversity (MoDiv) is a simple and practical transmission enhancement technique that utilizes different modulation mappings to reduce packet loss rate and achieve higher link throughput. MoDiv is particularly meaningful and effective in hybrid-ARQ systems. We study the deployment and optimization of MoDiv for HARQ in a MIMO-coordinated multi-point (MIMO-CoMP) scenario to mitigate packet loss. We formulate the design optimization of MoDiv into a quadratic three-dimensional assignment problem (Q3AP), then solve it using a modified iterated local search (ILS) method. Numerical results demonstrate clear performance gain over simple retransmissions and over a heuristic design.

2 - A Robust Optimization Approach For Network Interdiction Amelia Musselman, Georgia Institute of Technology, Atlanta, GA, United States, amusselman@gatech.edu, Richard Li-Yang Chen, Janson Wu Networks arise in many systems that play an integral role in our daily lives, from the internet to the national electric grid. It is important to protect these networks from potential attacks. However, oftentimes the budget for network security is limited, and the effectiveness of various defense mechanisms may be uncertain. In this research we develop a robust optimization algorithm for selecting defense options to strengthen network security with a limited budget. We use a bi-level programming model, where the adversary’s goal is to maximize damage to the system while the defender’s goal is to minimize this damage. We test our method on a large synthetic problem of the U.S. supply chain network.

WB70 Acoustic- Omni Transportation, Ops II Contributed Session

Chair: Antoine Petit, U of Illinois at Urbana-Champaign, 205 N Mathews Avenue, # B156, Urbana, IL, 61801, United States, apetit@illinois.edu 1 - Travel Time Estimation Based On Complex Networks Within A DTA Framework Rui Chen, Tsinghua University, Shunde building of Tsinghua University, Beijing, China, chenruiest@163.com, Satish Ukkusuri Data-driven estimation based on complex networks is a most important one of new methods which could deal with DTA for both associated mathmatical properties and computational ability. Therefore, we proposed a data-driven travel time estimation model based on Complex Networks within a DTA framework. 2 - System Optimal Traffic Signal Control Under Dynamic User Equilibrium Dynamic user equilibrium (UE) introduces nonlinear constraints with time- varying state-dependent delay terms. An optimal traffic signal control framework is proposed to find the signal control settings that minimize the total travel time in a road network, as well as maintaining the UE condition and other realistic spillback constraints. In this study, a heuristic solution method is proposed to solve the nonlinear problem with time-varying, state-dependent delays. The resulting solution satisfies the desired properties, which suggests the proposed solution algorithm is better than the previous introduced method with approximation by constant time delays. 3 - Managing The Daily Operations Of A Bike Sharing System Using Portable Stations Rahul Swamy, PhD Student, University of Illinois Urbana- Champaign, Champaign, IL, 61820, United States, rahulswa@illinois.com, Jose Luis Walteros This research aims to provide an integrated mathematical framework for operating a bike-sharing system using portable stations. We propose solving a sequence of MILPs to optimally determine the locations of portable stations across a time period in order to minimize redistribution logistics. A Bender’s decomposition based solution strategy is presented. Existing work in this area treat this problem separate from the bike station location problem. We propose an integrated approach. 4 - The Impact Of Demand Uncertainty On Subsidy Design In BOT Road Projects Zhuo Feng, Dalian University of Technology, No. 2 Linggong Road, Dalian, 116024, China, zhfeng@dlut.edu.cn, Yiwen Zhang The private investor faces substantial demand uncertainty in BOT (Build-Operate- Transfer) road projects, which depresses her participation. To attract private investors, the government often offers subsidies in some BOT road projects. In this paper, we will mainly consider usage-based subsidy. We first examine the impact of demand uncertainty on the government’s subsidy design by considering the private investor’s response in designing toll price and road capacity. We further make two extensions by locating the project in a road network and by considering the private investor’s information advantage of her operating cost, respectively. Rui Ma, Postdoc Scholar, University of California, Davis, 1001 Ghausi Hall, One, Davis, CA, 95616, United States, drma@ucdavis.edu, Hao Yu, Michael Zhang

419