2015 Informs Annual Meeting

TA73

INFORMS Philadelphia – 2015

TA73 73-Room 203B, CC Functional Data Analysis Sponsor: Quality, Statistics and Reliability Sponsored Session

2 - Concurrent Process Plan Optimization for Additive Manufacturing Bahir Khoda, Professor, North Dakota State University, Room # 202F Civil and Industrial Enginee, 1410 14th Avenue North, Fargo, ND, 58102, United States of America, akm.khoda@ndsu.edu, Amm Nazmul Ahsan, Md Habib Implementing additive manufacturing processes effectively requires addressing issues of process proficiencies and resource utilization, both of which have a strong environmental impact. In this paper, both part build orientation and material deposition direction are concurrently optimized by analyzing part geometry to minimize the resource requirement. A concurrent multicriteria process plan optimization framework is developed using Genetic Algorithms (GA) technique. 3 - Online Sensor-based Monitoring in Aerosol Jet Printing Process Prahalad Rao, SUNY Binghamton, 4400 Vestal Pkwy. E, Binghamton, NY, United States of America, prao@binghamton.edu, Roozbeh Salary, Jack Lombardi, Mark Poliks Aerosol Jet Printing (AJP) is an additive manufacturing process (AM) is emerging as a viable method for printing conformal electronics. However, teething quality related problems in AJP remain unresolved. We propose approaches based on image processing and sensor data analytics to achieve online quality monitoring in the AJP process. The effectiveness of the proposed approach is assessed and evaluated with several real case studies implemented on an aerosol jet printer setup. TA75 75-Room 204B, CC IBM Research Best Student Paper Award I Sponsor: Service Science Sponsored Session Chair: Ming-Hui Huang, National Taiwan University, Taiwan - ROC, huangmh@ntu.edu.tw 1 - Best Student Paper Competitive Presentation Ming-Hui Huang, National Taiwan University, Taiwan - ROC, huangmh@ntu.edu.tw Finalists of the IBM Research Best Student Paper Award present their research findings in front of a panel of judges. The judging panel will decide the order of winners, which will be announced during the business meeting of the Service Science Section at the Annual Conference. 1- Service Innovation and the Role of Collaboration Cong Feng, Syracuse University, 721 University Avenue, Syracuse NY, United States of America, feng@congfeng.net, K. Sivakumar Results show that (1) the effect of service innovation on firm performance is greater for service firms than manufacturing firms; (2) the relationship between the propensity for service innovation and three types of collaboration is signifi- cant; and (3) vertical and third-party collaborations are more beneficial than hor- izontal collaboration for service firms. 2 - Brand Equity and Extended Service Contract Purchase Decisions Moein Khanlari Larimi,University of Alberta, Canada, khanlari@ualberta.ca, Paul Messinger In this paper, we explore the role of brand equity on consumers’ extended serv- ice contract (ESC) purchase decisions. We draw from past findings to show that higher brand equity has an overall positive impact on ESC purchase decisions. We also explore the positive impact of stores on ESC purchase decisions. 3 - Regulating Greed over Time Stefano Traca, Massachusetts Institute of Technology, Cambridge, MA, United States of America, stet@mit.edu, Cynthia Rudin In retail, there are predictable yet dramatic time-dependent patterns in customer behavior, such as periodic changes in the number of visitors, or increases in visi- tors just before major holidays (e.g., Christmas). The current paradigm of multi- armed bandit analysis does not take these known patterns into account, which means that despite the firm theoretical foundation of these methods, they are fundamentally flawed when it comes to real applications. This work provides a remedy that takes the time-dependent patterns into account, and we show how this remedy is implemented in the UCB and e-greedy methods. In the corrected methods, exploitation (greed) is regulated over time, so that more exploitation occurs during higher reward periods, and more exploration occurs in periods of low reward. In order to understand why regret is reduced with the corrected methods, we present a set of bounds that provide insight into why we would want to exploit during periods of high reward, and discuss the impact on regret. Our proposed methods have excellent performance in experiments, and were inspired by a high-scoring entry in the Exploration and Exploitation 3 contest using data from Yahoo! Front Page. That entry heavily used time-series methods to regulate greed over time, which was substantially more effective than other contextual bandit methods.

Chair: Moein Saleh, Discover Financial Services/ Arizona State University, 699 S Mill Ave, Tempe, AZ, 85281, United States of America, Moein.Saleh@asu.edu 1 - On the Use of Gaussian Processes for Surface and Profile Data Enrique Del Castillo, Penn State University, Industrial Eng. and Statistics Depts., State College, United States of America, exd13@psu.edu Standard applications of Gaussian Processes in manufacturing data have traditionally been based on models of the form z(x,y) where x,y,z are coordinates acquired with some sensor, so correlation is assumed to occur on euclidean space external to the surface. We show new methodology that assumes instead correlation exists on the intrinsic surface points along geodesic distances, and show how this leads to better surface reconstruction in both simulated and real datasets. 2 - Functional Clustering with Applications in Single Molecule Experiments Ying Hung, yhung@stat.rutgers.edu Cell adhesion experiments refer to biomechanical experiments that study protein, DNA, and RNA at the level of single molecules. Motivated by analyzing a single molecule experiment, a new statistical framework is proposed based on functional clustering approaches. Simulations and applications to real experiments are conducted to demonstrate the performance of the proposed method. 3 - Design of Experiments for Functional Response Moein Saleh, Discover Financial Services/ Arizona State University, 699 S Mill Ave, Tempe, AZ, 85281, United States of America, Moein.Saleh@asu.edu, Rong Pan Applications of DOE for single response variable can be seen in nearly every disciplines in science and engineering. However, there are very few publications that discussed optimal design for the experiments with multiple responses taken over different points of a continuum variable. This continuum can be any other continuous variable for functional data analysis such as time in longitudinal study. My study focuses on developing a framework for designing the experiments for functional response. 4 - Monitoring and Diagnostics of High Dimensional Multi-stream Data Samaneh Ebrahimi, Research Assistant, Georgia Institute of Technology, 755 Ferst Drive, Atlanta, GA, 30332, United States of America, samaneh.ebrahimi@gatech.edu, Kamran Paynabar, Chitta Ranjan Correlated high-dimensional data streams (HDDS) pose significant challenges in Statistical Process Monitoring. In this research, we integrate PCA and Adaptive Lasso, and propose a novel approach for effective process monitoring and diagnosis of HDDS. The effectiveness of the proposed approach is validated through simulation and a case study.

TA74 74-Room 204A, CC System and Process Informatics in Additive Manufacturing (I) Sponsor: Quality, Statistics and Reliability Sponsored Session

Chair: Linkan Bian, Assistant Professor, Mississippi State University, 260 McCain Building, Mississippi State, Starkville, MS, 39762, United States of America, bian@ise.msstate.edu 1 - Accelerated Process Optimization for Laser-based Additive Manufacturing (LBAM) Amir M. Aboutaleb, Mississippi State University, 260 McCain Building, Mississippi State, MI, 39762, United States of America, aa1869@msstate.edu, Linkan Bian, Alaa Elwany, Nima Shamsaei, Scott M. Thompson A novel Design-of-Experiment methodology is proposed to efficiently optimize process control parameters for LBAM by leveraging data obtained from prior related but non-identical studies. Our method accounts for unavoidable difference between the experimental conditions of the current and prior studies and quantify the associated uncertainty, which is further updated using real-world data generated in the current study.

282