2015 Informs Annual Meeting

SA74

INFORMS Philadelphia – 2015

SA72 72-Room 203A, CC Predictive Modeling and Control for Additive Manufacturing Sponsor: Quality, Statistics and Reliability Sponsored Session

3 - A Discrete Semi-markov Model to Determine Optimal Repair Decisions for Trend-renewal Process Ernie Love, Professor Emeritus, Simon Fraser University, 8888 University Drive, Burnaby, BC, v5a1r5, Canada, love@sfu.ca, Qingyu Yang, Wujun Si The failure and repair process of a repairable machine (system) is modeled as a trend-renewal process permitting the modeling of imperfect repairs. The state of such a system can be characterized by the real age of the system and the failure count permitting the use of a two-state semi-Markov model to determine optimal repair/replacement decisions. Threshold type policies are established. Failure data from a cement kiln is used to demonstrate the approach. 4 - A Mixed Effect Kijima Model and Application in Optimal Maintenance Analysis Wujun Si, PhD Student, Wayne State University, Detroit, MI, 48202, United States of America, wujun.si@wayne.edu, Qingyu Yang The Kijima model has been widely applied to analyzing repairable systems with general repair efficiency. Most existing studies treat the repair efficiency as a fixed value while it can vary among a series of repair actions. In this paper, we propose a mixed effect Kijima model to characterize the variation of repair efficiency. An SAEM algorithm is developed for model parameter estimation. Based on the proposed model an optimal maintenance analysis is developed as a case study. SA74 74-Room 204A, CC IEEE T-ASE Invited Session: Manufacturing Systems Automation Sponsor: Quality, Statistics and Reliability Sponsored Session Chair: Jingshan Li, Professor, 1513 University Ave, Madison, WI, 53706, United States of America, jli252@wisc.edu 1 - A Quality Flow Model in Battery Manufacturing Systems for Electric Vehicles Feng Ju, Assistant Professor, Arizona State University, Tempe, AZ, 53705, United States of America, jeffrey0930@gmail.com In this paper, we present a flow model to analyze product quality in battery assembly lines with 100% inspections and repairs for defective parts. A Markov chain based model is introduced to analyze quality propagations along the battery production line. Analytical expressions of final product quality are derived and structural properties are investigated. A case study is presented to illustrate the applicability of the method. 2 - Energy-efficient Production Systems through Schedule-based Operations Liang Zhang, University of Connecticut, 371 Fairfield Way UNIT Control of production operations is considered as one of the most economical methods to improve energy efficiency in manufacturing systems. This paper investigates energy consumption reduction in production systems through effective scheduling of machine startup and shutdown. The theoretical methods are applied through a case study in automotive paint shop operations. 3 - Adaptive Sensor Allocation Strategy for Process Monitoring and Diagnosis in a Bayesian Network Kaibo Liu, Assitant Professor, UW-Madison, 1513 University Avenue, Madison, 53706, United States of America, kliu8@wisc.edu, Xi Zhang, Jianjun Shi This talk proposes a novel approach to adaptively reallocate sensor resources based on online observations in a Bayesian Network model, which can enhance both monitoring and diagnosis capabilities. The proposed method addresses two fundamental issues in an integrated manner: when to reallocate sensors and how to update sensor layout. Case studies are performed on a hot forming and a cap alignment process to illustrate the performance of the proposed method under different fault scenarios. 4 - Online Steady-state Detection for Process Control using Multiple Change-point Models Shiyu Zhou, Professor, University of Wisconsin-Madison, Department of Industrial and Systems Eng, 1513 University Avenue, Madison, WI, 53706, United States of America, shiyuzhou@wisc.edu, Jianguo Wu, Yong Chen, Xiaochun Li Steady-state detection is critical in process performance assessment, fault detection and process automation and control. We proposed a robust on-line steady-state detection algorithm using multiple change-point model and particle filtering techniques. Extensive numerical analysis shows that the proposed new method is more accurate and robust than the other existing methods. 4157, Storrs, CT, 06269-4157, United States of America, liang@engr.uconn.edu, Jorge Arinez, Stephan Biller, Guorong Chen

Chair: Qiang Huang, Associate Professor, University of Southern California, GER 240, USC, Los Angeles, CA, United States of America, qiang.huang@usc.edu Co-Chair: Arman Sabbaghi, Assistant Professor Of Statistics, Purdue University, Department of Statistics, 150 N. University Street, West Lafayette, IN, 47907, United States of America, sabbaghi@purdue.edu 1 - Bayesian Additive Modeling for Quality Control of 3D Printed Products Arman Sabbaghi, Assistant Professor Of Statistics, Purdue University, Department of Statistics, 150 N. University Street, West Lafayette, IN, 47907, United States of America, sabbaghi@purdue.edu, Tirthankar Dasgupta, Qiang Huang Three-dimensional (3D) printing is a disruptive technology with the potential to revolutionize manufacturing. However, control of product deformation remains a major issue. Quality control requires a generic methodology that can predict deformations for a wide range of designs based on data available for a few previously manufactured products. We develop a Bayesian methodology to update prior conceptions of deformation for a new design based on printed products of different shapes. 2 - Predictive Modeling of in-plane Geometric Deviation for 3d Printed Freeform Products He Luan, University of Southern California, GER 236, USC, Los Angeles, CA, United States of America, hluan@usc.edu, Qiang Huang Although additive manufacturing holds great promise, dimensional geometric accuracy remains a critical issue and lacks of generic solve method. Our work fills the gap by establishing a general model predicting in-plane deviations of AM built freeform products. Built upon our previous model for cylinder and polyhedron, this work directly predicts freeform shape deviations from CAD design. SLA experiments validated this method, indicating the prospect of optimal compensation for freeform products. Chair: Qingyu Yang, Assistant Professor, Wayne State University, 4815 4th street, Room 2167, Detroit, Mi, 48202, United States of America, qyang@wayne.edu Co-Chair: Eunshin Byon, Assistant Professor, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States of America, ebyon@umich.edu 1 - Modeling of Degradation Data from Disjoint Time Intervals Xiao Liu, liuxiao@sg.ibm.com Motivated by real-life problems, this paper presents a statistical model for degrdation data collected from disjoint time intervals (blocks). Within each block, high-frequency degradation measurements are available. Of interest is the extreme (i.e., maximum or minimum) of the degradation level within each interval. 2 - A Generic Method for Analyzing Complex Data with Covariates Haitao Liao, Associate Professor, The University of Arizona, The University of Arizona, Tucson, AZ, 85716, United States of America, hliao@email.arizona.edu, Yiwen Xu, Neng Fan In this research, we study an automated modeling approach to constructing phase-type (PH) distributions via mathematical optimization and develop PH- based models to analyze complex data with covariates. . SA73 73-Room 203B, CC Data Analytics for Reliability Evaluation and Maintenance Optimization I Sponsor: Quality, Statistics and Reliability Sponsored Session

63