2016 INFORMS Annual Meeting Program

SB68

INFORMS Nashville – 2016

3 - Efficient Multi-fidelity Decision Making For Dynamic Data Driven Application Systems Jie Xu, George Mason University, jxu13@gmu.edu Chun-Hung Chen, Edward Huang Dynamic data driven application systems enable real-time simulation-based decision making. However, existing simulation optimization algorithms lack the computational efficiency required for real-time decision making. In this talk, we present a Bayesian framework that makes use of data and models of multiple fidelity levels to achieve the computational efficiency necessary for decision support in the context of dynamic data driven application systems. 4 - Dynamic Data Driven Modeling Of Nanoparticle Self-assembly Processes Xin Li, Florida State University, 2525 Pottsdamer St, Building A, We present a dynamic data-driven modeling strategy, capable of tracking and predicting the transient dynamics of nanoparticle production processes. The proposed methodology is built upon two emerging multi-resolution instruments. The methodology regularly triggers cheap low resolution measurements while triggering expensive high resolution measurements when model predictions fail. The proposed strategy would provide crucial clues to understand nanoparticle productions as well as powerful insights to control the production of nanoparticles for yielding desirable morphology. Advanced Maintenance Modeling Sponsored: Quality, Statistics and Reliability Sponsored Session Chair: Yisha Xiang, Lamar University, Beaumont, TX, United States, yxiang@lamar.edu Co-Chair: David W Coit, Rutgers University, Piscataway, NJ, United States, coit@rci.rutgers.edu 1 - Reordering Of Spare Parts Experiencing Two Phase Onshelf Deterioration Haitao Liao, University of Arkansas, liao@uark.edu We study maintenance and inventory policies for a system carrying spare parts that experience two-phase on-shelf deterioration. Based on the parts’ degradation states, we introduce two different replacement strategies for the spare consumption, i.e., the Degraded-First strategy and the New-First strategy. 2 - A Model Of System Limiting Availability Under Imperfect Maintenance Suzan Alaswad, Zayed University, suzan.alaswad@zu.ac.ae Charles Richard Cassady, Edward A Pohl In this paper, we explore the impact of Kijima Type II imperfect repair model on equipment availability. Our specific interest is in the system steady-state availability. Since the derivation of a closed form expression for the limiting availability is extremely difficult, we use simulation modeling and analysis to estimate the system limiting availability. Next, we develop a meta-model to convert the system reliability and maintainability parameters into the coefficients of the limiting availability estimate without the simulation effort. Lastly, we identify an optimal age-based preventive maintenance policy that maximizes the system’s steady-state availability. Suite A231, Tallahassee, FL, 32310, United States, xl12d@my.fsu.edu, Chiwoo Park, Yu Ding, Tao Liu SB68 Mockingbird 4- Omni

deterioration. In this new model, two failure processes within each component are dependent due to simultaneous shared exposure to shock process. Furthermore, degradation paths among certain components are considered to be dependent. Components sharing dependent degradation can be determined by the MLE of model parameters.

SB69 Old Hickory- Omni Panel Discussion: Internet of Things (IoT) Data Analytics Sponsored: CPMS, The Practice Section Sponsored Session

Moderator: Robin Lougee, IBM Research, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, rlougee@us.ibm.com 1 - Panel Discussion: Internet Of Things (IoT) Data Analytics Robin Lougee, IBM Research, IBM TJ Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY, 10598, United States, rlougee@us.ibm.com What are the opportunities and challenges for analytics and operations research when virtually every machine that operates in every market and sector can be connected to the internet? Thought leaders who create IoT technologies, systems, and application solutions will share their experiences, delineate the substance from the hype, and engage in a lively discussion of the most needed areas of future research. 2 - Panelist Doug Meiser, The Kroger Co., 11450 Grooms Road, Cincinnati, OH, 45242, United States, doug.meiser@kroger.com 3 - Panelist Srinivas Bollapragrada, GE Global Research Center, 1 Research Circle, K1-5a50a, Niskayuna, NY, 12309, United States, bollapragada@research.ge.com 4 - Panelist Ihsan Sehgal, IBM, 3039 E Cornwallis Road, Research Triangle Park, NY, 27709, United States, rlougee@us.ibm.com 5 - Panelist Joseph Byrum, Syngenta, 913 31st Street, West Des Moines, IA, 50265, United States, joseph.byrum@syngenta.com SB70 Acoustic- Omni Transportation, Freight II Contributed Session Chair: Samaneh Shiri, Research Assistant, University of South Carolina, 101 pickens st. APt. G2, Columbia, SC, 29205, United States, shiri@email.sc.edu 1 - Commodity-based Econometric Empty Trip Models Carlos Alberto Gonzalez-Calderon, Research Associate, Rensselaer Polytechnic Institute, 4 25TH ST, APT 5, Troy, NY, 12180, United States, gonzac8@rpi.edu, Jose Holguin-Veras, Ivan Dario Sanchez- Diaz, Ivan Sarmiento, Johanna Amaya This paper estimates econometric models of empty trips of different commodities and vehicle types. In doing this, panel models with time-dependent parameters and fixed effects are used to assess how parameters change over time considering different commodities, and to detect the presence of time effects not captured by the other parameters. The performing of the formulation for the different commodities is tested in Colombia. 1 - On The Unique Features Of On-demand Peer-to-Peer Logistics Systems Jennifer A Pazour, Assistant Professor, Rensselaer Polytechnic Institute, 110 8th street, CII 5217, Troy, NY, 12180, United States, pazouj@rpi.edu On-demand peer-to-peer logistics systems use a business model for the movement and storage of goods that matches independent supply resources (warehouse space, truck space, delivery services) to demand requests on demand. These systems are part of the growing sharing economy and gig economy. We identify the unique features of these systems, comparing and contrasting them with traditional logistics systems. By mapping the characteristics to supply chain principles, we identify challenges with designing and operating, as well as using

3 - Predictive Maintenance For A Multi-unit System Yisha Xiang, Lamar University, yxiang@lamar.edu Zhicheng Zhu, David W Coit

Preventive maintenance has been extensively studied. Time-based PM and condition-based maintenance (CBM) are two major approaches for PM. However, time-based PM is often associated with high occurrence of system breakdowns, and CBM might incur more-than-necessary inspections. Recently, predictive maintenance has become popular since it aims to pinpoint when a failure is about to occur and prolong the operational time. However, only a few predictive models consider a multi-unit system. In this paper, we develop an opportunistic predictive maintenance structure for a multi-unit system. Numerical examples are

provided to illustrate the proposed predictive maintenance policy. 4 - Reliability Of System With Clusters Of Dependent Degrading Components Sanling Song, Rutgers University, Busch Campus,

Core Building Room 201, Piscataway, NJ, 08854, United States, sanling@scarletmail.rutgers.edu, David W Coit, Qianmei Feng, Yisha Xiang A reliability model is developed for complex multi-component system with each component subject to multiple failure processes. Degradation paths for certain components are stochastically dependent with clusters of dependent components. Gamma process is used to model the stochastic process of component

66