Research Publications

Recent Submissions

Now showing 1 - 10 of 19689

• 

  Item

  Aviation-BERT: A Preliminary Aviation-Specific Natural Language Model

  ( 2023-06) Chandra, Chetan ; Jing, Xiao ; Bendarkar, Mayank ; Sawant, Kshitij ; Elias, Lidya R. ; Kirby, Michelle ; Mavris, Dimitri N. ; Georgia Institute of Technology. Aerospace Systems Design Laboratory

  Data-driven methods form the frontier of reactive aviation safety analysis. While analysis of quantitative data from flight operations is common, text narratives of accidents and incidents have not been sufficiently mined. Among the many use cases of aviation text-data mining, automatically extracting safety concepts is probably the most important. Bidirectional EncoderRepresentations from Transformers (BERT) is a transformer-based large language model that is openly available and has been adapted to numerous domain-specific tasks. The present work provides a comprehensive methodology to develop domain-specific BERT model starting from the base model. A preliminary aviation domain-specific BERT model is developed in this work. This Aviation-BERT model is pre-trained from the BERT-Base model using accident and incident text narratives from the National Transportation Safety Board (NTSB) and AviationSafety Reporting System (ASRS) using mixed-domain pre-training. Aviation-BERT is shown to outperform BERT when it comes to text-mining tasks on aviation text datasets. It is also expected to be of tremendous value in numerous downstream tasks in the analysis of aviation text corpora.
• 

  Item

  BERT for Aviation Text Classification

  ( 2023-06) Jing, Xiao ; Chennakesavan, Akul ; Chandra, Chetan ; Bendarkar, Mayank ; Kirby, Michelle ; Mavris, Dimitri N. ; Georgia Institute of Technology. Aerospace Systems Design Laboratory

  The advent of transformer-based models pre-trained on large-scale text corpora has revolutionized Natural Language Processing (NLP) in recent years. Models such as BERT (Bidirectional Encoder Representations from Transformers) offer powerful tools for understanding contextual information and have achieved impressive results in numerous language understanding tasks. However, their application in the aviation domain remains relatively unexplored. This study discusses the challenges of applying multi-label classification problems on aviation text data. A custom aviation domain specific BERT model (Aviation-BERT) is compared against BERT-base-uncased for anomaly event classification in the Aviation Safety Reporting System (ASRS) data. Aviation-BERT is shown to have superior performance based on multiple metrics. By focusing on the potential of NLP in advancing complex aviation safety report analysis, the present work offers a comprehensive evaluation of BERT on aviation domain datasets and discusses its strengths and weaknesses. This research highlights the significance of domain-specific NLP models in improving the accuracy and efficiency of safety report classification and analysis in the aviation industry.
• 

  Item

  Publishing without perishing: Alma publishing profiles for fun and profit

  (Georgia Institute of Technology, 2023-05) Patrick, Martin ; Traill, Stacie ; Georgia Institute of Techology. Library ; University of Minnesota. Library

  Alma publishing profiles are a powerful and flexible way to export metadata from Alma. In this presentation, we will demonstrate how publishing profiles play a key role in customizing and normalizing data to meet project needs, and discuss when they are a better choice than Alma Analytics or the Export Bibliographic Records job. Publishing profiles can easily incorporate physical or electronic inventory information into MARC bibliographic records for further analysis and transformation through various MARC editing and processing tools. Publishing also allows powerful filtering rules and normalization routines to customize data outputs and can work incrementally via OAI-PMH for data harvest and processing by external programs. Applications covered in the presentation include: using published data for analysis to support enrichment and remediation projects, exporting transformed metadata for ingest into a local digital repository, incremental OAI publishing for external partners, and using general publishing for OCLC Datasync.
• 

  Item

  Thinking about Design Thinking During a Migration

  (Georgia Institute of Technology, 2023-05) Patrick, Martin ; Georgia Institute of Technology. Library

  The systems librarian's goal should be to optimize and enhance the systems in use by centering the needs of the system's users. However, much of the literature and the focus on centering users is about, in academic libraries, our students and faculty. For the systems librarian, though, there is a group of users whose needs are equally valid: the library’s staff, faculty, and student workers (that is, the internal users). A recent survey of 55 of the most recent articles indexed in LISTA about user experience in libraries revealed one that focused on staff users. In this presentation, I outline some ideas for discovering and designing for the internal user’s needs, particularly given our upcoming migration, based on ideas found in IBM’s Enterprise Design Thinking Framework (EDT). I will cover a very brief history of Design Thinking in general, and then discuss some of the ways I think EDT could prove useful during a system migration. I will also share a few real-world examples of how I have approached problems and issues in the past, and how EDT might change that approach. The goal of this presentation is not to propose a dogma around EDT but to challenge those of us on the systems side to do user experience work with and for our colleagues, in addition to our student and faculty communities.
• 

  Item

  Tokyo Smart City Studio at Nihonbashi – Spring 2023

  (Georgia Institute of Technology, 2023-05) Beattie, Aaron ; Brock, Cooper ; Farooq, Umar ; Khorashahi, Yasamin ; Mase, Heather ; Zhao, Yuxiang ; Xie, Yan (Lucy) ; Rawlins, Miles ; Sivakumar, Siddharth ; Aceto, Steven ; Knight-Scott, Ethan ; Dean, Emily ; Yan, Peirui ; Chen, Yining (Annie) ; Shetty, Jayita ; Lin, Yizhou ; Yang, Perry Pei-Ju ; Lejeune, Dillon ; University of Tokyo. Department of Urban Engineering ; Mitsui Fudosan UTokyo Laboratory ; Keio University. Graduate School of System Design and Management ; Dean, Emily ; Khorashahi, Yasamin ; Mase, Heather

  The Tokyo Smart City Studio explores a method of data-driven urban design, and how digital urban technologies enable architects and planners to comprehend cities, urban spaces and architecture from data visualization, mapping, modeling, performance evaluation to architecture and urban form making. The project aims to design a smart urban district that is carbon neutral, climate resilient and post-covid-19 conscious.
• 

  Item

  Augmenting the Impact of Community Organizations

  (Georgia Institute of Technology, 2022-05) Al-Khayyal, Sarah ; Allgaier, Orion ; Amahrir, Safae ; Mers, Bianca ; Wagliardo, Nathan ; Webber, Noel ; Wright, Janelle ; Yandell, Izzy

  Based on a philosophy of community-based planning that centers community needs which are supported by external partners, the 2022 studio worked to leverage a planning skillset to bolster the organizational capacity of this year’s partner organization, Impacto Juventud. Building off the work of the 2019 and 2020 Georgia Tech studios in Puerto Rico, the 2022 studio recognizes that global studios can be mutually beneficial for students to broader than perspectives and for partners to increase access to additional resources. In the context of this studio, this was best accomplished by building the capacity of community organizations to support their outreach and activism. The studio’s sub-focuses included community asset mapping, renewable energy, and non-profit communications.
• 

  Item

  Tokyo Smart City Studio at Nihonbashi – Spring 2022

  (Georgia Institute of Technology, 2022-05) Yang, Perry Pei-Ju ; Arsam, Muhammad ; Buchanan, Regan ; Chen, Lu ; Conschafter, Stephen ; Clowse, Maddy ; Foxley, Sebastian ; Franco-Pinilla, Rossana ; Garcia, Mirna ; Nicolson, Maggie ; Manitius, Natalie ; Snedaker, Tasha ; Wineski, Olivia ; Manitius, Natalie

  The studio's mission is to enhance the Nihonbashi neighborhood through carbon neutrality, climate resiliency, and post-Covid-19 consciousness. The studio focused on: 1. Celebrating the progress and history of the neighborhood 2. Engaging stakeholders across social, cultural, and geographic distances 3. Ensuring that future development supports climate resiliency and livable- and people-focused communities 4. Adding open spaces that support synergy between blue and green systems 5. Designing streetscapes and transit that makes movement enjoyable and accessible 6. Helping the neighborhood become more resilient to shocks such as Covid-19 or natural disasters 7. Anticipating trends and needs of population changes with land use 8. Harnessing smart technologies to enhance quality of life and economic opportunity, as well as our designs and processes 9. Catalyzing Tokyo's pursuit of carbon neutrality by using Nihonbashi as an example
• 

  Item

  NOx Production from Premixed Hydrogen/Methane Fuel Blends

  (Georgia Institute of Technology, 2023-09-01) Breer, Benjamin R. ; Godbold, Conner W. ; Emerson, Benjamin L. ; Sun, W. ; Noble, Bobby ; Lieuwen, Timothy C. ; Georgia Institute of Technology. Strategic Energy Institute

  Hydrogen (H2) fuel is a promising means for long duration energy storage and dispatchable utilization of intermittent renewable power, which can be combusted without CO2 emissions. However, combustion of any fuel in air can still lead to NOX production. This whitepaper summarizes recent analyses of NO emissions of premixed H2/CH4 blends, demonstrating how fundamental drivers of NOX production change with hydrogen addition. Three major findings are presented: (1) At constant temperature, NO emissions decrease with the addition of H2 for typical gas turbine conditions; (2) Although NOX emissions are typically quoted as parts per million (ppm), it is not appropriate to use ppm as a comparison between different hydrogen blended compositions; one must use mass based comparisons (e.g., ng/J or lbm/MMBTU); (3) atmospheric pressure fuel sensitivity NOX studies will not capture the controlling NO production physics that are present in practical applications, such as gas turbines. These results provide important context for several experimental studies that have been reported. First, they are consistent with several recent demonstrations of fielded gas turbines with hydrogen blending, that show constant to declining NOX levels with hydrogen addition. Second, some lab studies have noted that hydrogen blended systems have elevated NOX emissions relative to natural gas, but these appear to be for nonpremixed systems and it is not entirely clear what is being held constant for these comparisons (temperature, power, etc.). Given the strong temperature sensitivity of NOX production, these results cannot be applied more generally to understand NOX emissions tendencies. Taken together, we conclude that utilization of modern premixing combustion technologies with hydrogen blending should lead to constant or decreasing NOX emissions, but use in older, diffusion type burners can lead to elevated NOX. item_description: This whitepaper summarizes recent analyses of NO emissions of premixed H2/CH4 blends, demonstrating how fundamental drivers of NOX production change with hydrogen addition. Three major findings are presented: (1) At constant temperature, NO emissions decrease with the addition of H2 for typical gas turbine conditions; (2) Although NOX emissions are typically quoted as parts per million (ppm), it is not appropriate to use ppm as a comparison between different hydrogen blended compositions; one must use mass based comparisons (e.g., ng/J or lbm/MMBTU); (3) atmospheric pressure fuel sensitivity NOX studies will not capture the controlling NO production physics that are present in practical applications, such as gas turbines.
• No Thumbnail Available

  Item

  Video: A platoon of differential-drive robots maintaining uniform inter-robot spacing while driving in formation

  (Georgia Institute of Technology, 2019) Wardi, Yorai ; Seatzu, Carla ; Cortés, Jorge ; Egerstedt, Magnus ; Shivam, Shashwat ; Buckley, Ian ; Georgia Institute of Technology. School of Electrical and Computational Engineering ; University of Cagliari, Italy. Department of Electrical and Electronic Engineering ; University of California, San Diego. Department of Mechanical and Aerospace Engineering ; University of California, Irvine. Samueli School of Engineering ; Agtonomy, San Francisco, CA

  Video Referenced by the article "Tracking Control by the Newton-Raphson Method with Output Prediction and Controller Speedup", International Journal of Robust and Nonlinear Control, DOI: 10.1002/rnc.6976, Wiley, 2023.
• 

  Item

  Propeller Generating User-defined Primitive (UDP) in Engineering Sketch Pad

  ( 2021-04) Oluwalana, Daniel ; Georgia Institute of Technology. School of Aerospace Engineering

  Engineering Sketch Pad (ESP) is a web-based system used to create and manipulate geometry for the aim of designing and analyzing aerospace vehicles. There are User-defined Primitives that are pre-packaged with ESP; however, the system also allows users to create their own single body primitives written in C, C++ or FORTRAN and have them coupled with ESP and compiled in real time. The purpose of this paper is to detail the process used in creating a User Defined Primitive (UDP) within Engineering Sketch Pad that generates a propeller using a well-established design process. Prior to this propeller scheme implementation in the software, a user would generate a propeller by manually arranging a series of airfoils at certain angles and applying a covering or skin over them, an inefficient method as users would have to permute the airfoil arrangements to achieve the design shape and power. UDP Propeller is the first power/thrust auto-derived propeller primitive to be implemented in any CAD software as it creates optimum propeller blades for an aircraft’s engine based on specifications from a user such as power coefficient and advance ratio.