Page 8 - IJAMD-2-1

International Journal of AI for
Materials and Design
Utilizing AI for NTSB UAV accident categorization

decision-making and risk management. The textual AI, particularly machine learning and deep learning,
3
nature of traditional accident reports can often hamper has shown significant promise in UAV data analysis.
more automated structured categorization efforts, Nguyen et al. demonstrated the effectiveness of multitask
6
making it challenging to analyze and interpret the data deep learning in analyzing UAV multisensory data,
efficiently. Advanced methods are gaining prominence in enhancing crop productivity and safety by accurately
aviation research and practical industry application. They predicting equipment malfunctions and environmental
can contribute to the need to process large volumes of conditions. Similarly, the application of large language
unstructured data and extract meaningful insights. The models (LLM), particularly GPT-4, for categorizing UAV
4
National Transportation Safety Board (NTSB) investigates accident reports, as proposed in this study, leverages the
UAV-related accidents and makes detailed reports openly technology to systematically classify incident causes. This
available to support safety analysis. Still, at times, some structured approach offers a more nuanced understanding
of these reports may have missing fields, inconsistent of UAV-related risks. Other researchers have also employed
taxonomies and data formats, and can require natural AI models to parse and interpret large datasets of
language processing (NLP) to extract critical information accident reports, achieving higher accuracy in identifying
and support statistical analysis. 5 patterns and probable causes than manual methods. 4,5,7-9

Artificial intelligence (AI), specifically NLP and For instance, AI categorization techniques have been
machine learning approaches, can enhance the analysis applied in manned aviation and automotive industries,
of UAV accident reports and support the identification demonstrating the versatility and robustness of AI-driven
of safety improvements. By leveraging OpenAI’s GPT-4, analysis. 10
this study aims to evaluate the feasibility of automating Effective data visualization can also be crucial for
the categorization of UAV accident reports and the interpreting complex UAV data and communicating
identification of probable causes and patterns in the data. findings. Techniques such as cluster maps and interactive
AI-driven categorization coupled with data visualization charts are employed to identify geographic and temporal
techniques can provide a deeper understanding of patterns in UAV accidents. These visualization methods
UAV accidents, enabling proactive measures to address facilitate a better understanding of spatial distributions
safety issues and inform policy decisions. Given proper and trends, aiding in the development of targeted safety
consideration of accuracy challenges that could be explored measures. 11,12 Research underscores the importance
in further studies, the approach can also complement the of visualization in making complex data accessible
resource-intensive manual categorization tasks or serve and actionable, particularly for policymakers and
as a capability augmentation tool for such activities. This regulatory bodies aiming to implement data-driven safety
research highlights the potential of AI to revolutionize interventions. 13,14
the way UAV accident data are processed and analyzed,
ultimately contributing to improved safety standards and The integration of AI and advanced data visualization
practices in the UAV industry. techniques offers a powerful framework for enhancing
UAV safety research. AI-driven categorization streamlines
2. Literature review data analysis while visualization tools make the data more
The exponential increase in UAV usage has led to a accessible and interpretable. Studies have shown that this
corresponding rise in safety challenges, necessitating integrated approach not only improves the accuracy of UAV
comprehensive studies to mitigate associated risks. Key data analysis but also supports the development of proactive
safety concerns include loss of control, navigation failures, safety policies. For instance, combining AI categorization
and collisions with other aircraft or infrastructure. These with visual analytics has been used to identify critical
issues often stem from operator error, mechanical failure, safety issues and inform regulatory measures in aviation
4
and adverse environmental conditions. Regulatory and and other high-risk industries. This synergy is essential
2,3
technical challenges further complicate the safe integration for addressing the evolving challenges in UAV operations
6
of UAVs into national airspace systems. Nguyen et al. and ensuring robust safety standards. By leveraging the
highlight difficulties in establishing consistent regulatory strengths of both AI and data visualization, researchers can
frameworks and the technical limitations of current UAV provide deeper insights into accident data, enabling more
systems, which contribute to navigation and control effective and timely interventions to enhance UAV safety.
issues. Synthesizing these studies, it becomes evident that 3. Methods
addressing UAV safety requires a multifaceted approach
involving enhanced risk assessments, robust regulatory The UAV accident reports (n = 34) were sourced from
measures, and technological advancements. the NTSB database, which provides detailed records of

Volume 2 Issue 1 (2025) 2 doi: 10.36922/ijamd.8544