Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD33

3 - The Locomotive Assignment Problem with Distributed Power Camilo Ortiz-Astorquiza, Université de Montréal, Montreal, QC, Canada, Emma Frejinger, Jean-Fran ois Cordeau The Locomotive Assignment Problem consists of determining the optimal assignment of locomotive types to scheduled trains while ensuring power requirements and flow balance. We introduce a variant of the LAP where the operation mode of the train is also part of the decision process. Two modes of operation are considered: Distributed Power and Conventional. We also include additional constraints and preferences to capture the requirements of the partner railway company for this project, Canadian National Railway (CN). We model the problem as a network design problem, present an IP formulation and discuss a solution method with some preliminary computational results. n TD33 North Bldg 222C Design and Modeling of Advanced Manufacturing Emerging Topic: OR and Advanced Manufacturing Emerging Topic Session Chair: Linkan Bian, Mississippi State University, MS, 39762, United States 1 - Life Cycle Cost Trade-off Model: Comparison of Life Expectancy and Quality on Overall Cost Janis Terpenny, Penn State University, Industrial & Manufacturing Engineering, 310 Leonard Building, University Park, PA, 16802, United States, Connor Jennings The classic problem in business is the trade-off between quality and cost. Models have been created to predict the life expectancy of products based on usage and manufacturing specifications and other models have been created to estimate the total life cycle cost of the model. However, few models look at both life cycle length and life cycle cost together. This paper presents a framework that combines these two types of models to determine an optimal decision. A Genetic optimization algorithm is applied to find the minimal cost, given a desired life cycle. An alternative model is also presented, that determines the maximum life cycle, given a set cost. 2 - A Design and Process Improvement Framework for Additive Manufacturing Gul Kremer, Iowa State University, 2529 Union Drive, 3004 Black, Ames, IA, 50011, United States Although additive manufacturing (AM) offers unprecedented opportunities to design and manufacturing professionals to improve product designs and benchmark manufacturing processes for capability comparisons, the cost of training and technology for AM adoption is high. Specifically, AM adoption as a viable alternative requires part designs that take into account technology capabilities and limitations. Moreover, what is possible with a specific technology using one material may not be possible with another material. Finally, reproducibility is a significant concern across many AM technologies that requires attention. This talk will summarize our group’s work responding to these challenges. The design and process improvement framework synergistically uses Axiomatic Design, TRIZ, Design of Experiments and Process Capability Analysis. Several case studies will be summarized that show tangible benefits of using the developed framework. 3 - A Spare Parts Sourcing Policy with Additive Manufacturing Taner Cokyasar, University of Tennessee-Knoxville, Knoxville, TN, United States, Mingzhou Jin, Sean Willems Spare part demand is usually met through two suppliers: fast and slow. The trade- off in between these channels was extensively studied in the literature. In recent studies, additive manufacturing (AM), was proposed as an alternative option to minimize spare parts acquisition cost. Yet, these studies did not quantitatively measure the effectiveness of the technology. In this study, an optimization model was developed to assess the viability of acquiring AM technology and determine the sourcing policy of the parts. M/G/k queueing theory and (S-1, S) inventory policy were used to calculate the average long-term cost of using AM and inventory options, respectively. 4 - Layer-wise Quality Modeling of Laser-Based Additive Manufacturing Using High-Dimensional Thermal Images Linkan Bian, Mississippi State University, Industrial and Systems Engineering Department, MS, 39762, United States, Wenmeng Tian, Seyyed Hadi Seifi Shishavan, Hongyue Sun One of the main challenges in Additive Manufacturing (AM) process in concerned with functional quality of fabricated parts. A novel layer-based approach is introduced to evaluate the quality of the deposited layer based on analyzing the process properties (e.g. thermal history). High dimensional process data is mapped to a lower dimensional domain that is correlated to the porosity distribution of a layer. A real-world case study shows that the accuracy of the proposed methodology in predicting defected layers is close to 98%.

n TD34 North Bldg 223 2:00 - 2:45 Simio/

2:45 - 3:30 Provalis Research Emerging Topic: Technology Tutorials Emerging Topic Session 1 - Simulation and Scheduling Software All in One! Renee Thiesing, Simio LLC, 3616 Wynbrooke Circle, Lousville, KY, 40241, United States, Katie Prochaska Simio is a premier simulation and scheduling software that allows you to expand traditional benefits of simulation to improve daily operations. In this tutorial, we will demonstrate Simio’s 3D rapid modeling capability to effectively solve real problems. Explore how a single tool can be used to not only optimize your system design, but also provide effective planning and scheduling. Come explore the Simio difference and see why so many professional and novice simulationists are changing to Simio. 2 - The Different Text Analytics Approaches used for Business Analytics Normand Peladeau, Provalis Research, Montreal, QC, Canada Text analytics can provide you with real value by helping you quickly extract meaningful information from your text data such as incident reports, corporate reports, social media, customer reviews and much more. However, Text Analytics doesn’t work the same way for everyone. To make text analytics work for you, you need to know some of the pitfalls to avoid the pratfalls. We will show you techniques and methods you can deploy, what’s behind them and what to watch out for. n TD35 North Bldg 224A Traffic Flow Management Concepts Sponsored: Aviation Applications Sponsored Session Chair: James Calvin Jones, University of Maryland-College Park, Cambridge, MA, 02140, United States 1 - Autonomous Air Traffic Control for Sequencing and Separation with Nested Deep Reinforcement Learning Marc Brittain, Iowa State University, Ames, IA, United States, Peng Wei With the increasing air traffic density and complexity in the traditionally controlled airspace, an autonomous air traffic control system is needed as the ultimate solution to handle dense, complex air traffic. In this work, we design and build an artificial intelligence agent to perform air traffic control sequencing and separation. The approach is to formulate this problem as a reinforcement learning model and solve it using deep reinforcement learning. For demonstration, the NASA Sector 33 app has been used as our learning environment for the agent. Results show that this agent can guide aircraft safely and efficiently through Sector 33 and achieve required separation at the metering fix. 2 - Saturation Technique for Optimizing Planned Acceptance Rates in Traffic Management Initiatives Peng Wei, Assistant Professor, Iowa State University, Aerospace Engineering Department, Iowa State University, Ames, IA, 50011, United States In this paper, we have addressed a fundamental question in TMI PAR planning: do there exist optimal PARs which only depend on the physical airport or airspace capacity but not the demand? We show that this conjecture holds true in the deterministic capacity case but not in the general stochastic case. We propose a new heuristic saturation technique. We demonstrate that this technique can not only reveal the properties and limiting behaviors of GDP models but also could potentially be used as a robust PAR policy when facing demand uncertainty. The findings of this paper provide valuable insights in understanding the TMI rate planning problem and a robust algorithm for GDP optimization. 3 - Reinforcement Learning Methods for AFP selection James Calvin Jones, University of Maryland-College Park, Cambridge, MA, 02140, United States In this talk we discuss the use of a set of reinforcement learning algorithms for setting the planned flow rates on a set of airspace flow programs. The work is validated through the use of an agent-based air traffic management simulation.

358