Person:

Bendarkar, Mayank

Permanent Link

https://hdl.handle.net/1853/71092

Associated Organization(s)

Organizational Unit

Daniel Guggenheim School of Aerospace Engineering

Full item page

Publication Search Results

Now showing 1 - 9 of 9

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

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

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.
BERT for Aviation Text Classification

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

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.
An Integrated Framework to Evaluate Off-Nominal Requirements and Reliability of Novel Aircraft Architectures in Early Design

(Georgia Institute of Technology, 2021-04-29) Bendarkar, Mayank

One of the barriers to the development of novel aircraft architectures and technologies is the uncertainty related to their reliability and the safety risk they pose. In the conceptual and preliminary design stages, traditional system safety techniques rely on heuristics, experience, and historical data to assess these requirements. The limitations and off-nominal operational considerations generally postulated during traditional safety analysis may not be complete or correct for new concepts. Additionally, dearth of available reliability data results in poor treatments of epistemic and aleatory uncertainty for novel aircraft architectures. Two performance-based methods are demonstrated to solve the problem of improving the identification and characterization of safety related off-nominal requirements in early design. The problem of allocating requirements to the unit level is solved using a network-based bottom-up analysis algorithm combined with the Critical Flow Method. A Bayesian probability approach is utilized to better deal with epistemic and aleatory uncertainty while assessing unit level failure rates. When combined with a Bayesian decision theoretic approach, it provides a mathematically backed framework for compliance finding under uncertainty. To estimate multi-state reliability of complex systems, this dissertation contributes a modified Monte-Carlo algorithm that uses the Bayesian failure rate posteriors previously generated. Finally, multi-state importance measures are introduced to determine the sensitivity of different hazard severity to unit reliability. The developed tools, techniques, and methods of this dissertation are combined into an integrated framework with the capability to perform trade-studies informed by safety and reliability considerations for novel aircraft architectures in early preliminary design. A test distributed electric propulsion (T-DEP) aircraft inspired by the X-57 is utilized as a test problem to demonstrate this framework
Evaluation of Off-Nominal Performance and Reliability of a Distributed Electric Propulsion Aircraft during Early Design

(Georgia Institute of Technology, 2021-01-04) Bendarkar, Mayank ; Sarojini, Darshan ; Harrison, Evan D. ; Mavris, Dimitri N.

General Aviation (GA) is likely to be at the forefront of a paradigm change in aviation, where the introduction of novel concepts such as Urban Air Mobility (UAM), architectures like e-VTOL, and technologies like Distributed Electric Propulsion (DEP) are expected to make aircraft more efficient and reduce their environmental footprint. However, these architectures carry with them an uncertainty related to the off-nominal operational risk they pose. The limitations and off-nominal operational considerations generally postulated during traditional safety analysis may not be complete or correct for new technologies. While a lot of the literature surveyed focuses on improving traditional methods of safety analysis, it still does not completely address the limitations caused due to insufficient knowledge and experience with transformative technologies. The research objective of the present work is to integrate the Bayesian safety assessment framework developed previously by the authors with conceptual and 6-DoF performance models for DEP aircraft to evaluate off-nominal performance and reliability using information that is typically available in conceptual or preliminary design phases. A case study on the electric power architecture of the the NASA Maxwell X-57 Mod. IV is provided. A maximum potential flight path angle metric, as well as trimmability considerations using a 6-DoF model constructed using available literature help determine hazard severity of power degradation scenarios. Bayesian failure rate posteriors are constructed for the different components in the traction power system, which are used in a Bayesian decision framework. The results indicate that while most of the components in the traction power architecture of the X-57 Mod. IV are compliant with failure rate requirements generated, the batteries, cruise motors, and cruise motor-inverters do not meet those requirements.
A Model-Based Aircraft Certification Framework for Normal Category Airplanes

(Georgia Institute of Technology, 2020-06) Bendarkar, Mayank ; Xie, Jiacheng ; Briceno, Simon ; Harrison, Evan D. ; Mavris, Dimitri N.

A typical aircraft certification process consists of obtaining a type, production, airworthiness, and continued airworthiness certificate. During this process, a type certification plan is created that includes the intended regulatory operating environment, the proposed certification basis, means of compliance, and a list of documentation to show compliance. This paper extends previous work to demonstrate a model-based framework for the management of these certification artifacts for normal category airplanes. The developed framework integrates the regulatory rules and approved means of compliance in a single model while using best-practices found in Model-Based Systems Engineering (MBSE) literature. This framework, developed using SysML in MagicDraw captures not just the textual requirements and verification artifacts, but also their relationships and any inherent meta-data properties via custom defined stereotype profiles. Additionally, a simulation capability that automates the extraction and export of the applicable rules (certification basis) and corresponding means of compliance for any aircraft under consideration at the click of a button has been developed. The framework also provides numerous additional benefits to different stakeholders that have been described in detail with examples where necessary.
Evaluating Optimal Paths for Aircraft Subsystem Electrification in Early Design

(Georgia Institute of Technology, 2019-06) Bendarkar, Mayank ; Rajaram, Dushhyanth ; Yu, Cai ; Briceno, Simon ; Mavris, Dimitri N.

The aerospace industry’s push for More-Electric Aircraft (MEA) has motivated numerous studies to quantify and optimize the impact of subsystem electrification in early design phases. Past studies on multi-objective optimization of MEA show a clear benefit over conventional architectures when no constraints are placed on the number of subsystems electrified at once. In reality however, aircraft manufacturers are more likely to progressively electrify subsystems over multiple aircraft generations. While step-by-step electrification may lead to sub-optimal intermittent MEA architectures when compared with scenarios with no such imposition on number of subsystems electrified, little or no literature was found to address the optimal paths towards such electrification changes. The primary aim of this study is the creation of a mathematically defensible methodology that provides decision makers with the ability to analyze several paths for electrification of MEA subsystems while considering Pareto-optimality and other metrics based on objectives of interest in early design. It is hoped that decision makers will be able to understand the performance trade-offs between different electrification paths under different scenarios, constraints, and uncertainties. The resulting methodology is demonstrated on an exercise in the electrification of Small Single Aisle aircraft.
A Model-Based System Engineering Approach to Normal Category Airplane Airworthiness Certification

(Georgia Institute of Technology, 2019-06) Bleu-Laine, Marc-Henri ; Bendarkar, Mayank ; Xie, Jiacheng ; Briceno, Simon ; Mavris, Dimitri N.

Airworthiness certification is to ensure the safety of aircraft. With the surge in novel general aviation aircraft configurations and technologies, the Federal Aviation Administration replaced prescriptive design requirements with performance-based airworthiness standards in Federal Aviation Regulations Part 23 that governs the airworthiness of normal category airplane. The amendment ported over the accepted means of compliance (MoC) from prescriptive advisory circulars to a number of consensus standards from aviation community. Because these MoCs are scattered in multiple documents and cross-reference one another, the certification practice with this new format may be cumbersome and time-consuming.This paper proposes a Model Based System Engineering (MBSE) approach that is envisioned to parametrically transform the document-centric exercise to a model-based process. The approach helps collect the FAR23 regulations and the associated MoC in an integrated system model along with the relevant mappings between them. This allows users to automatically generate a compliance checklist for any specific certification requirement. Other benefits of the MBSE approach include circular referencing check, automatically propagating any future changes to the FARs or MoC standards through the model, and potential incorporation with early aircraft design.
A Bayesian Safety Assessment Methodology for Novel Aircraft Architectures and Technologies using Continuous FHA

(Georgia Institute of Technology, 2019-06) Bendarkar, Mayank ; Behere, Ameya ; Briceno, Simon ; Mavris, Dimitri N.

Novel architectures and technologies carry with them an uncertainty related to their reliability and associated safety risk. Existing safety assessment methods involve determining the severity of discrete functional failure and the corresponding probability. However, with the advent of novel aircraft architectural and operational concepts, traditional methods of establishing severity and probabilities failures are found lacking due to the scarcity of available data. The current work proposes a safety assessment method that uses architecture-specific performance models along with continuous functional hazard assessments to inform hazard severity. The probability of failures is determined using a Bayesian framework that does not falter when data is scarce. Taken together, it is expected that this new proposed methodology will enable a more accurate safety assessment of novel aircraft architectures and technologies. A safety assessment of an electric propulsion system powered by a fuel cell is conducted using the proposed methodology to serve as a proof of concept.
Rapid Assessment of Power Requirements and Optimization of Thermal Ice Protection Systems

(Georgia Institute of Technology, 2018-06) Bendarkar, Mayank ; Chakraborty, Imon ; Garcia, Elena ; Mavris, Dimitri N.

A thermal ice protection system prevents or dispatches ice formed on critical aircraft components like wings or nacelles by heating them either through electro-thermal or pneumatic means. The power requirements for such a system are a function of flight and atmospheric conditions and protected surface area. The developed analysis framework allows evaluation of transient and steady-state cases, anti-icing and de-icing designs, as well as evaporative and running-wet operation. To enable these analyses, a flow solver is first used to calculate local water catch efficiencies and convective heat transfer coefficients on an airfoil. These are then used within a thermal solver which evaluates water and ice accumulations over multiple control volumes under different cases of interest. This control volume approach includes both thermal and mass balances to track temperatures of the protected surface, ice, and water, as well as water/ice layer thicknesses and the water mass flow in or out of the control volume through evaporation or runback. Finally, this tool can yield power requirements for different system layouts and operating conditions, or optimize the protected surface area for a given airfoil under given operating conditions. This can help designers get an estimate of the power draw, and obtain more information on placement of the IPS on novel configurations during the design space exploration phase itself with greater fidelity and minimal computational costs.