2016 INFORMS Annual Meeting Program

WD71

INFORMS Nashville – 2016

WD68 Mockingbird 4- Omni Complex Process Modeling, Monitoring and Optimization Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Youngseon Jeong, Chonnam National University, Korea, Republic of, youngseonjeong@gmail.com Co-Chair: Myong K Jeong, Rutgers University, 96 Frelinghuysen Road, Piscataway, NJ, 08854, United States, mjeong@rci.rutgers.edu 1 - A Predictive Modeling Of Pump Breakdown For Semiconductor Fabrication Based On Sensor Data Kil Soo Kim, Principal Engeneer, Samsung Electronics, SamsungJeonja-ro, Hwasung, Gyeonggi-do, 18448, Korea, Republic of, ks1.kim@samsung.com, Chanhwi Jung, Solmi Park, Hyung-Seok Kang, Dongkyu Jeon, Seung Hoon Tong We present a method for predictive monitoring of pump breakdown failure which may result in wafer scrap or unnecessary main chamber stop loss based on secondly collected sensor data. The sensor data set that were collected massively during several months from pumps contains useful information to monitor and predict the condition of pump connected with main chamber. We developed the monitoring statistics and predictive model which can estimate the remaining useful life. The real-life case study is presented to illustrate our proposed procedures. 2 - Application For Modeling And Optimization Of A Crucial Parameter Identification Shreya Gupta, University of Texas at Austin, Austin, TX, shreya.gupta@utexas.edu We are working with Samsung Semiconductor on building a statistical optimization model that identifies equipment and parameters to rank best and worst routes for the purpose of scheduling and testing modified recipes. We are also employing data mining and statistical techniques to identify control spec to improve process capability and yield for semiconductor manufacturing. Finally, we are developing a prototype software to demonstrate the use and value of the proposed analysis and algorithms. 3 - Phase I Analysis Of Nonlinear Profiles Via Gaussian Process Models Yang Zhang, Tianjin University of Commerce, No. 409, Guangrong Rd., Tianjin, China, yzhang@tjcu.edu.cn, Nini Gao, Qing Wang In profile monitoring, process monitoring and diagnosis remains and important and challenging problem. Although the analysis of nonlinear profile data have been extensively studied in the literature, the challenges associated with diagnosis of nonlinear profiles with within-profile correlation are yet to be well addressed. In consideration of within-profile correlation, a Gaussian process model is applied to model the nonlinear profiles. A practical diagnosis scheme based on Schwarz’s Bayesian information criterion is proposed to identify the outliers in Phase I. Simulation results show that the proposed method could effectively find outlying profiles in a historical dataset. 4 - Adaptive Variability Monitoring Procedure For High-dimensional Processes Jinho Kim, Qatar University, jhkim04@gmail.com Monitoring process variability of a multivariate process is crucial to ensure high quality of product. However, monitoring process variability in high-dimensional processes is considerably difficult due to the large number of variables and the limited number of samples. In this talk, we present a procedure based on an adaptive LASSO-thresholding for monitoring changes in the covariance matrix. The performance of the proposed chart, is evaluated for various shift patterns and compared with one of the existing penalized likelihood based methods. The results show the effectiveness of the proposed chart.

This paper studies how to subsidize a monopolistic public transit operator with unknown production cost parameters. An incentive-compatibility regulatory mechanism which induces the operator to report its true parameter are design: government seeks to maximize social welfare by determining the transit service parameters and the subsidy to operator to induce its participation, and the operator implements the operation schedule to meet financial constraint. Comparison between complete and asymmetric information in terms of fixed cost and marginal cost are proposed. It will provide an effective tool for designing policies and evaluating practices regarding public transit subsidization. 2 - Effects Of Multiple Capacity Changes on Congestion Pricing Model To Handle “Day Of Operations” Airport Capacity Reduction Abdul Qadar Kara, Asst. Professor, King Fahd University of Petroleum and Minerals, KFUPM Box 5067, Dhahran, 31261, Saudi Arabia, aqkara@kfupm.edu.sa In an earlier work, a model was built on basic econometric principle of congestion pricing embedded within an optimization model. The model provided a mechanism to manage airport runway capacity reduction on “day of operations”. The current work reports further analysis of the model, mainly its response towards the effect of multiple changes in runway access to arriving flights on both the schedule and the pricing. 3 - Price-compatible Matching Mechanisms For Carrier Collaboration Su Xiu Xu, The University of Hong Kong, LG108, Composite Building, HKU, Pokfulam Road, Hong Kong, 999077, China, xusuxiu@gmail.com This study is the first extending the existing market design theory to the field of supply chain and logistics management. It is known that money flow is not allowed in the matching markets like stable marriage, house allocation, and kidney exchange. In this study, we explore the potential of lane exchange among a number of self-interested truckload carriers in a collaboration network. We propose the (price-compatible) top trading cycles and deals (TTCD) mechanism and the price-compatible top trading cycles and chains (PC-TTCC) mechanism. Both mechanisms are effective in terms of the compatibility with money flow, strategy-proofness, the realized welfare of carriers, and budget balance. 4 - Dynamic Team Orienteering Problem Emre Kirac, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701, United States, ekirac@uark.edu, Ashlea Bennett Milburn This study introduces the dynamic team orienteering problem (DTOP), which is a combinatorial optimization problem with many practical applications such as humanitarian relief logistics and tourist trip organizations. In DTOP, some locations are known at planning time while others are dynamic and each associated with a profit. The goal is to maximize collected profits by visiting a set of static and dynamic locations throughout a planning horizon within a specified timeframe. The multiple plan approach (MPA) is adapted to solve DTOP. Competitive ratio analysis using an offline algorithm is performed to assess the performance of MPA.

WD71 Electric- Omni

Game Theory IV Contributed Session

Chair: Manxi Wu, Massachusetts Institute of Technology, 235 Albany Street, 3112B, Cambridge, MA, 2139, United States, manxiwu@mit.edu 1 - On Learning How Players Learn: A Mechanical Turk Experiment Walid Krichene, University of California, Berkeley, Berkeley, CA, 94720, United States, walid@eecs.berkeley.edu We consider a noncooperative game in which players compete for resources. In the online model, the game is played repeatedly and players update their strategies using an online learning algorithm. We study whether learning dynamics are descriptive of human behavior. We developed a web application to simulate the game, and used the Mechanical Turk platform to collect data on decision dynamics of human players. Using this data, we pose and solve a dynamics estimation problem, and show that a parameterized online model (based on the mirror descent method) can be descriptive of players’ decision dynamics. We give qualitative insights, evaluate the predictive ability of this model, and discuss its limits.

WD70 Acoustic- Omni Transportation, Ops IV Contributed Session

Chair: Emre Kirac, University of Arkansas, 4207 Bell Engineering Center, Fayetteville, AR, 72701, United States, ekirac@uark.edu 1 - Public Transit Regulation And Subsidization Under Asymmetric Information Yanshuo Sun, PhD Candidate, the University of Maryland, College Park, MD, 20740, United States, yssun@umd.edu, Qianwen Guo, Zhongfei Li

477