Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MA69

2 - A Wavelet-based Penalized Mixed-effects Model for Multichannel Profile Detection of In-line Raman Spectroscopy Xiaowei Yue, Georgia Institute of Technology, 755 Ferst Drive NW, ISYE, Atlanta, GA, 30332, United States, Hao Yan, Jin Gyu Park, Richard Liang, Jianjun Shi Modeling and analysis of high-dimensional (HD) profiles, is an important and challenging topic. We proposed a wavelet-based penalized mixed-effects decomposition (PMD) method for multichannel profile detection of in-line Raman spectra. The PMD exploits a regularized HD regression with linear constraints to decompose profiles into four parts: fixed effects, normal effects, defective effects, and noise. An APG optimization algorithm is developed for parameter estimation. Finally, extracted coefficients are associated with consistency, uniformity, and defects. In case study, we evaluated the performance of the PMD and demonstrated a better detection power with less computational time. 3 - Optimize the Signal Quality of the Composite Health Index via Data Fusion for Degradation Modeling and Prognostic Analysis Abdallah A. Chehade, Assistant Professor, University of Michigan- Dearborn, 4901 Evergreen Road, 2280 HPEC, Dearborn, MI, 48128, United States The rapid development of sensing and computing technologies enabled to simultaneously monitor different components of complex systems. This provided unprecedented opportunity to understand the degradation behavior of complex systems. In this paper, a signal-to-noise ratio (SNR) metric that is tailored to the needs of degradation signals is proposed. Then, based on the SNR metric, we develop a data-level fusion model to construct a health index via fusion of multiple degradation-based sensor data. Our goal is that the health index provides better estimates for the health condition and the remaining lifetime. A case study that involves the degradation dataset of aircraft gas turbine engines. 4 - Publication at IEEE T-ASE Shiyu Zhou, University of Wisconsin-Madison, 3254 Mech Eng Bldg, 1513 University Avenue, Madison, WI, 53706-1572, United States Abstract not available. 5 - Publication at IEEE T-ASE Part 1 Kaibo Liu, UW-Madison, 1513 University Avenue, Madison, WI, 53706, United States Panel discussion of paper publication in IEEE T-ase journal Degradation Test and Data Analysis Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Rong Pan, Arizona State University, Tempe, AZ, 85287-8809, United States Co-Chair: Guanqi Fang, Arizona State University, Tempe, AZ 1 - Bivariate Accelerated Degradation Modeling Based on Heterogeneous Marginal Stochastic Processes Guanqi Fang, Arizona State University, 745 N. Dobson Rd, Unit 115, Mesa, AZ, 85201, United States This presentation discusses a flexible modeling approach when dealing with bivariate degradation data. An example about application on polymer material is given to demonstrate the proposed modeling and data analysis flow. 2 - Predicting Lifetime by Degradation Tests: A Case Study of ISO 10995 Steve Rigdon, Saint Louis University, St. Louis, MO, 62026, United States ISO 109951 is the international standard for the reliability testing and archival lifetime prediction of optical media. The standard specifies the testing conditions in terms of the combinations of stress variables - temperature and relative humidity. The standard assumes that the projected failure time is the actual failure time and these projected failure times are then analyzed. Since true failure times are not directly observed, the uncertainties in the failure time must be taken into account. In this paper, we present a hierarchical model for degradation that can directly infer failure time at the use condition and compare this model with the ISO standard through a simulation study. Not accounting for the uncertainty in the projected failure times leads to unjustified confidence in the estimation for the median lifetime at both the stress conditions used in the experiments and at the use condition. n MA69 West Bldg 106A

3 - On Degradation Modeling Under Dynamic Environment Xiao Liu, University of Arkansas, 4172 Bell Engineering Center, Fayetteville, AR, 72701, United States, Mohammad Mahdi Hajiha This presentation discusses the modeling of degradation under dynamic environment. Application examples are presented to demonstrate the proposed modeling approach. n MA70 West Bldg 106B Joint Session QSR/DM: Machine Learning for Manufacturing Informatics Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Hao Yan, Arizona State University, Tempe, AZ 1 - Robust Change Detection via Affine and Quadratic Detectors Yao Xie, Georgia Institute of Technology, GA, United States, Yang Cao, Vincent Guigues, Anatoli Juditsky, Arkadi Nemirovski Change-point detection is a fundamental problem in statistics and signal processing with applications in manufacture systems. Given a sequence of data, the goal is to detect any change in the underlying distribution as quickly as possible. However, classic approaches usually require exact specification of the pre- and post-change distributions forms, which may not perform well with real data. We present a set of computationally efficient methods with certain near optimality properties, which allow uncertainties about the pre-and post-change distribution, by building a connection of change-point detection with robust convex optimization. 2 - Online Anomaly Detection with Adaptive Sampling for Multivariate Streaming Data Chen Zhang, National University of Singapore, E1-07-26, 3 Engineering Drive, Singapore, 117576, Singapore This paper addresses online anomaly detection of a system with multivariate variables, when only a subset of variables can be observed at each sensing epoch. We first propose a weighted Bayesian estimation framework for system state estimation which can deal with missing variables. Then based on the confidence region of the estimated system state, an anomaly detection scheme is constructed by considering correlations of different variables. We formulate its detection power as a multi-armed bandit problem, and propose an adaptive sampling strategy using the upper confidence bound algorithm. With the proposed strategy, we further derive the asymptotic detection power of our scheme. 3 - Understanding the Effects of Predictor Variables in Black Box Supervised Learning Models Jingyu Zhu, Northwestern University, Evanston, IL, United States, Daniel Apley A shortcoming of many supervised learning models is their lack of interpretability and transparency, in terms of understanding how the predictor variables affect the response. Partial dependence plots are used for visualization but involve unreliable extrapolation when the predictors are highly correlated. We present a new visualization approach termed accumulated local effects (ALE) plots that avoid extrapolation like functional ANOVA but are far more computationally efficient. They also yield convenient variable importance measures for quantifying and ranking the impact of each predictor. 4 - Real-time Anomaly Detection for Spatial-temporal Correlated Profile Hao Yan, 699 S. Mill Ave, Tempe, AZ, 85281, United States, Kamran Paynabar, Massimo Pacella Advanced 3D metrology technologies such as Coordinate Measuring Machine (CMM) and laser 3D scanners have facilitated the collection of massive point cloud data, beneficial for process monitoring, control and optimization. However, due to their high dimensionality and structure complexity, modeling and analysis of point clouds are still a challenge. In this paper, we utilize multilinear algebra techniques and propose a set of tensor regression approaches to model the variational patterns of point clouds and to link them to process variables. The performance of the proposed methods is evaluated through simulations and a real case study of turning process optimization.

144