Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD68

3 - Human Behavior in Physical Space and Virtual Space Lei Dong, MIT, Cambridge, MA, United States With the widespread of smart phone and location-based services, human ‘movements’ in the physical space and virtual space are recorded simultaneously. As far as we know, many human mobility patterns in physical space have been studied before, while there are a lot of open questions not be reached combing two spaces together: Does the virtual world have similar patterns of movement with physical space? Are there any interplay patterns between online and offline behavior? How do demographic properties matter in these two spaces? With a mobile phone dataset documenting detailed online and offline behaviors, we are seeking a deep understanding of humans behaviors in online and offline spaces.

n TD69 West Bldg 106A Reliability Data Modeling and Analysis Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Kangwon Seo, University of Missouri, Columbia, MO, 65211, United States 1 - Comparing Two Kaplan-meier Curves with the Probability of Agreement Lu Lu, Assistant Professor, University of South Florida, 4202 E. Fowler Ave., Tampa, FL, 33544-8861, United States, Nathaniel Stevens The probability of agreement has been used as an effective strategy for quantifying the similarity between the reliability of two populations by accounting for a practically important difference. This talk discusses two approaches for assessing the probability of agreement and its associated uncertainty for comparing the Kaplan-Meier curves for estimating the reliability of two populations. The first approach provides an easy assessment based on large sample approximations. The second approach offers more precise estimation by using the fractional random-weight bootstrap approach. The methods are illustrated with examples for comparing the reliability curves of related populations. 2 - Nonparametric Bayesian Network Approach System Reliability Analysis Rong Pan, Arizona State University, School of Computing Informatics & Decison Sys, P.O. Box 878809, Tempe, AZ, 85287- 8809, United States, Dongjin Lee In this talk we present the a nonparametric Bayesian network method for evaluating a complex system’s reliability while considering the uncertainty in system reliability structure and in component reliability assessment. 3 - Realtime Flight Anomaly Detection and Health Management Ashwin Rai, Arizona State University, Tempe, AZ, United States A statistical model for in-flight aircraft health management has been developed, with the capability of automated feature extraction, dimension reduction, and anomaly detection. To efficiently handle the high-dimensional flight data, information is de-noised, fused, and encoded following which a multivariate Gaussian mixture model is developed to train the flight data with limited amount of fault cases and form a global probability distribution of the aircraft performance. A robust cross-validation method is then implemented to obtain the optimal safety threshold of the aircraft performance. Results indicate that the developed model can detect faulty operations with short time-lags. 4 - Time-to-failure Analysis of Hollow Shafts in High Speed Trains with Fatigue Crack Ruochong Liu, Tsinghua University, Yanfu Li Reliability of the hollow shaft, the critical component in high speed trains, has been a significant issue concerned by the researchers and administrators in the railway domain. Fatigue crack is one of the main failure modes in hollow shafts. Therefore, a stochastic model combining the physics-of-failure mechanism has been established to describe the randomness of fatigue crack growth, which is partially incurred by the uncertainty of the axle force during the operation of the train. A Monte Carlo method has been conducted to simulate the degradation path of crack growth and estimate the distribution of time-to-failure of hollow shafts. 5 - A Hybrid Network Model for Information Fusion in National Airspace System Yuhao Wang, Arizona State University, Tempe, AZ, United States, Yongming Liu, Yongming Liu This paper present a hybrid network model combining the classical Bayesian method and the Maximum Entropy method called the Bayesian-Entropy Network (BEN). With the proposed method, the information fusion from various sources can be achieved. The BEN method can take in information other than point data/observation as constraints. It is demonstrated in Air Traffic Control (ATC) system to model human knowledge and empirical information to help the risk assessment and accident prediction in the National Airspace System.

n TD68 West Bldg 105C

Joint Session QSR/Practice Curated: Spatio- Temporal Data Analysis and Applications I Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Jian Liu, University of Arizona, Tucson, AZ, 85721, United States Co-Chair: Shyam Ranganathan, Virginia Polytechnic Institute and State University, Blacksburg, VA, United States Co-Chair: Xiao Liu, University of Arkansas, University of Arkansas, Fayetteville, AR, 5, United States 1 - Spatial Variable Selection and an Application to Virginia Lyme Disease Emergence Yili Hong, Virginia Polytechnic Institute, 213 Hutcheson Hall, Department of Statistics, Blacksburg, VA, 24061, United States Lyme disease is an infectious disease that is caused by a bacterium called Borrelia burgdorferi sensu stricto. One of the research objectives for the infectious disease community is to identify environmental and economic variables that are associated with the emergence of Lyme disease. In this paper, we use a spatial Poisson regression model to link the spatial disease counts and environmental and economic variables, and develop a spatial variable selection procedure to effectively identify important factors by using an adaptive elastic net penalty. The proposed methods can automatically select important covariates, while adjusting for possible spatial correlations of disease counts. 2 - Penalized Spatio-temporal Regression for Video Monitoring of Additive Manufacturing Process Hao Yan, 699 S. Mill Ave, Tempe, AZ, 85281, United States In this work, we focused on detecting the local occurrence of the “hot-spot event” in the metal additive manufacturing (AM) process. Here we propose to model the temporal behavior of the hot-spot event by the autoregressive model and propose a spatial-temporal decomposition approach to isolate the hot-spot event from the background. We also propose a control chart scheme to tell when the event actually happens. The proposed methodology is validated using both simulation and case study in metal AM. 3 - Spatio-temporal Modeling of Topic Flows with an Information Diffusion Monitoring Application on Social Networks Shyam Ranganathan, Virginia Polytechnic Institute and State University, VT, Blacksburg, VA, United States, Scotland Leman, Peter Hauck, James Hawdon Spatio-temporal topic flow modeling can be used to understand how information diffusion takes place on networks. We present an extension of the popular Latent Dirichlet Allocation (LDA) model to include spatial and temporal diffusion using ideas from the Gaussian process (GP) and the timeseries literature. We show an application where such a model can be used to monitor how information diffuses on social networks such as Twitter. We build a ‘polarization barometer’ based on the spatio-temporal model and use it to evaluate the amount of polarization generated by the information diffusion. 4 - A Generic Framework for Multisensor Degradation Modeling Based on Supervised Classification and Failure Surface Changyue Song, University of Wisconsin, Rm 3221, Mechanical Engineering Building, 1513 University Ave, Madison, WI, 53715, United States We propose a generic framework for multisensor degradation modeling. Specifically, we model each sensor signal based on random-effect models and characterize failure events by a multi-dimensional failure surface. To overcome the challenges in estimating the failure surface, we transform the degradation modeling problem into a supervised classification problem, where classifiers can be incorporated to estimate the degradation status of the unit based on the underlying signal paths, i.e., the collected sensor signals after removing the noises. The proposed framework is also capable for sensor selection, able to handle asynchronous sensor signals, and easy to implement in practice.

372