Identification of Causal Paths and Prediction of Runway Incursion Risk using Bayesian Belief Networks
Date of Award
Dissertation - Open Access
Doctor of Philosophy in Aviation
Dothang Truong, Ph.D.
First Committee Member
Alan J. Stolzer, Ph.D.
Second Committee Member
Steven Hampton, Ed.D.
In the U.S. and worldwide, runway incursions are widely acknowledged as a critical concern for aviation safety. However, despite widespread attempts to reduce the frequency of runway incursions, the rate at which these events occur in the U.S. has steadily risen over the past several years. Attempts to analyze runway incursion causation have been made, but these methods are often limited to investigations of discrete events and do not address the dynamic interactions that lead to breaches of runway safety. While the generally static nature of runway incursion research is understandable given that data are often sparsely available, the unmitigated rate at which runway incursions take place indicates a need for more comprehensive risk models that extend currently available research.
This dissertation summarizes the existing literature, emphasizing the need for cross-domain methods of causation analysis applied to runway incursions in the U.S. and reviewing probabilistic methodologies for reasoning under uncertainty. A holistic modeling technique using Bayesian Belief Networks as a means of interpreting causation even in the presence of sparse data is outlined in three phases: causal factor identification, model development, and expert elicitation, with intended application at the systems or regulatory agency level. Further, the importance of investigating runway incursions probabilistically and incorporating information from human factors, technological, and organizational perspectives is supported. A method for structuring a Bayesian network using quantitative and qualitative event analysis in conjunction with structured expert probability estimation is outlined and results are presented for propagation of evidence through the model as well as for causal analysis.
In this research, advances in the aggregation of runway incursion data are outlined, and a means of combining quantitative and qualitative information is developed. Building upon these data, a method for developing and validating a Bayesian network while maintaining operational transferability is also presented. Further, the body of knowledge is extended with respect to structured expert judgment, as operationalization is combined with elicitation of expert data to create a technique for gathering expert assessments of probability in a computationally compact manner while preserving mathematical accuracy in rank correlation and dependence structure.
The model developed in this study is shown to produce accurate results within the U.S. aviation system, and to provide a dynamic, inferential platform for future evaluation of runway incursion causation. These results in part confirm what is known about runway incursion causation, but more importantly they shed more light on multifaceted causal interactions and do so in a modeling space that allows for causal inference and evaluation of changes to the system in a dynamic setting. Suggestions for future research are also discussed, most prominent of which is that this model allows for robust and flexible assessment of mitigation strategies within a holistic model of runway safety.
Scholarly Commons Citation
Goodheart, Benjamin Jeffry, "Identification of Causal Paths and Prediction of Runway Incursion Risk using Bayesian Belief Networks" (2013). Doctoral Dissertations and Master's Theses. 53.