Date of Award
Summer 2024
Access Type
Dissertation - Open Access
Degree Name
Doctor of Philosophy in Aviation
Department
College of Aviation
Committee Chair
Dothang Truong
First Committee Member
David A. Esser
Second Committee Member
Jennifer Thropp
Third Committee Member
Kenneth Kuhn
College Dean
Alan J. Stolzer
Abstract
In the analysis of human error incidents, human factors specialists predominantly follow two schools of thought in categorization strategies. One is to categorize the operator’s actions by the physical properties of the activity (phenotypes); the other focuses on the cognitive behaviors preceding the incident (genotypes). These categorization strategies are intended to foster an understanding of the events, identify risks, and assist in the implementation of risk reduction interventions. Commercial aviation maintenance presents unique challenges to human factor practitioners in terms of task requirements, working environments, communication modalities, and safety standards. Complex tasks are handed from one shift of personnel to the next, and technicians are under constant pressure to rapidly assess an aircraft’s status, perform the needed maintenance, and return the aircraft to revenue service.
This exploratory qualitative research analyzed the way in which aviation maintenance incidents are reported and compared the thematic focus of maintenance reporting to established human factor paradigms. As a result of this analysis, a novel human factor conceptual framework is introduced that improves the alignment of incident investigation with maintenance incident reporting styles and improves the existing communication gap between incident reporter and incident investigator.
The research methodology detailed in this dissertation describes the analysis of a large corpus of aviation maintenance reports. This analysis is accomplished using natural language processing to implement Latent Dirichlet Allocation in a topic modeling strategy. The topic modeling process distilled these reports into a set of topic word groups that reflect the prevalent themes within the corpus of documents, allowing for a reasonable effort of evaluation by subject matter experts. Without the topic modeling process, the volume of data within the selected corpus of documents would be overwhelming and impractical for direct review and evaluation.
The topic modeling process and the topic word group thematic assessment were reinforced with subject matter expert evaluations by human factor and aviation maintenance specialists. The findings of this process are the inspiration for a novel human factor classification strategy focused on the way maintenance personnel describe maintenance events and the organizational responsibilities under whose authority and direction these maintenance activities occur.
This research contributes to aviation maintenance and human factors research by offering a previously unexplored approach combining natural language processing and qualitative evaluation to address the challenges encountered in the analysis of aviation maintenance incidents. The proposed human factor framework is a departure from established human factor paradigms and can, if accepted and effectively implemented, allow aviation maintenance organizations to develop effective risk mitigation strategies and improve technician performance and aviation safety.
Scholarly Commons Citation
Braun, Christopher P., "Exploring a New Conceptual Framework in Aviation Maintenance Incident Reporting Using Natural Language Processing" (2024). Doctoral Dissertations and Master's Theses. 832.
https://commons.erau.edu/edt/832
