Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WC72

2 - Condition-based Maintenance (CBM) Optimization for a Two-component System through State Discretization and Proportional Hazard Model (PHM) Mengkai Xu, Northeastern University, Boston, MA, 02115, United States, Noor E. Alam, Sagar S. Kamarthi, Xiaoning Jin CBM of mechanical systems with stochastic dependence among their components has drawn much attention recently. Researchers commonly assume the independence of degradation and failure of multiple components to keep the models simple. However, the existence of degradation and failure interactions diminishes the accuracy of the models. To address this issue, the state-rate dependence denoting interaction between degradation states and hazard rates is proposed. A state discretization technique integrated with PHM aimed at capturing the effect of state-rate dependence for maintenance optimization is presented. 3 - Stochastic Modeling of Corrosion Growth Changxi Wang, Ph.D Student, Rutgers University, Piscataway, NJ, 08854, United States Corrosion growth modeling is important in industry. Existing methods model corrosion pits depth growth and estimate reliability accordingly. However, volume loss may also lead to failures such as rupture even if corrosion pits depth is small, which usually cannot be captured by such models. We develop a degradation model that captures corrosion volume growth, as well as corrosion depth growth. The influence of stresses on corrosion growth is considered. The distribution of volume loss increments and failure probability in the next time increment is obtained. 4 - Study on Travel Reliability of Urban Rail Transit Network Using Automatic Fare Collection Data Yong Yin, Southwest Jiaotong University, Chengdu, China Yong Yin, National United Engineering Laboratory of Integrated and Intelligent Transportation, Chengdu, China, Jie Liu, Qiyuan Peng, Xu Yan, Anjun Li It is important to study the travel reliability of Urban Rail Transit network to ensure people travel through Urban Rail Transit. The OD matrix and travel time between stations are obtained through Automatic Fare Collection data. Three reliability indicators that considering tolerable index are constructed from the aspects of network connectivity, travel time and transport capability. Taking Chengdu Urban Rail Transit as an example, five workday data of Urban Rail Transit are analyzed. The results show that the multi-state model is more effective than single model to fit the travel time distribution and tolerable index has a huge influence of the travel reliability of Urban Rail Transit network. 5 - Process Monitoring Of Three-dimensional Topographic Surfaces Mejdal A. Alqahtni, Rutgers University, New Brunswick, NJ, 08901-8554, United States, Elsayed A. Elsayed, Myong K. Jeong This paper develops a real-time monitoring approach for assessing the quality of 3D topographic surfaces. The approach initially improves the representation of 3D surface features by slicing the 3D surface topography into several layers. The functional spatial randomness (FSR) profile is then suggested for surface characterization in which the spatial randomness of topographic values is computed at each layer. By utilizing the functional principal component analysis (FPCA), an anomaly detection approach based on FSR profile is proposed. The developed approach reveals outstanding performance compared to the existing approaches in identifying various forms of surface defects. n WC71 West Bldg 106C Decision Diagrams and Optimization Sponsored: Computing Sponsored Session Chair: Willem van Hoeve, Carnegie Mellon University, Pittsburgh, PA, 15213, United States 1 - An MDD-based Formulation for Scheduling Clinical Rotations Tallys Yunes, University of Miami, Coral Gables, FL, United States, Andre Augusto Cire, Adam Diamant We study the scheduling problem of a medical school that must assign a cohort of students to a set of clinical rotations, subject to operational and QoS constraints. Students become eligible for clerkship throughout the year and complete rotations at hospitals in different regions. Each hospital offers a subset of the rotations, with different start dates, capacities, and cost rates. We propose a network-flow formulation based on MDDs that is computationally superior to an MILP model and provides useful managerial insights. Using a real-life dataset, we run a discrete-event simulation of the assignment process. Compared to current practice, we achieve an average cost reduction of 25%.

2 - Integer Programming Postoptimality Analysis Using Decision Diagrams

John Hooker, Carnegie Mellon University, Tepper School of Business, Pittsburgh, PA, 15213, United States, Thiago Serra We show how a decision diagram can compactly represent all solutions of an integer programming problem that are within a given tolerance of the optimal value. The structure of the diagram facilitates rapid processing of a wide range of queries about the near-optimal solution space. “Sound decision diagrams can more compactly represent near-optimal solutions by innocuously admitting some spurious solutions. In fact, repeated application of a simple “sound reduction operation yields a smallest possible sound diagram for a given problem instance. Computational tests show that it is typically far smaller than a tree representing the same set of solutions. 3 - Solving Multi-follower Bilevel Mixed-integer Programs via a Generalized Value Function Onur Tavaslioglu, University of Pittsburgh, Houston, TX, 77025, United States, Oleg A. Prokopyev, Andrew J. Schaefer We introduce a generalized value function of a mixed-integer program, which is simultaneously parameterized by its objective and right-hand side. We describe its fundamental properties, which we exploit through three algorithms to calculate it. We then show how this generalized value function can be used to reformulate multi-follower bilevel mixed-integer programming problems. We present the solution of instances that are significantly larger than those solved in the literature in terms of the total number of variables, and the number of followers. We compare our approach with an open source solver, MibS. 4 - Exact Multiple Sequence Alignment by Synchronized Multi-valued Decision Diagrams Willem-Jan Van Hoeve, Carnegie Mellon University, Tepper School of Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United States, Amin Hosseininasab Sequence alignment problems appear in the context of bioinformatics, natural language processing, and financial applications, and aim to identify regions of similarity in sequences of data. While pairwise sequence alignment is tractable, optimal multiple sequence alignment is challenging in theory as well as practice. We present a new exact method that first represents all pairwise sequence alignments using multi-valued decision diagrams. These are then synchronized with an integer programming model, which we solve using a logic-based Benders decomposition. Our approach can improve state-of-the-art heuristic solvers, and outperforms the best exact methods from the literature. n WC72 West Bldg 211A Practice- Operations Management IV Contributed Session Chair: Fang Fang, California State University, Los Angeles, CA, 91801, United States 1 - Forward-reserve Warehouses: Why a Dynamic Selection of Items is Necessary Rakesh Venkitasubramony, Assistant Professor, Indian Institute of Management Lucknow, IIM Road, Prabandh Nagar, Lucknow, 226013, India Forward-Reserve warehouses are popular when small number of Stock Keeping Units (SKU) account for large number of picks. Most studies that approach the design and allocation of SKUs to forward area have static decisions. In this simulation based article, I argue for dynamic selection of SKUs to the Forward area of the warehouse, when individual demands are seasonal. 2 - Competition in Two-sided Market with Quality Control Strategies Lyu Gaoyan, Peking University, No.5 Yiheyuan, Beijing, 100871, China, Lihua Chen There are many two-sided markets with similar products. We design mechanism to explore how to cooperate and compete with competitors at the same time. Meanwhile, how to improve their product quality to attract consumers is another important factor in this mechanism. We set cooperate parameter and compete parameter to measure the degree of market competition. The result shows that, with low degree competition, both markets get less profit.

493