Ph.D. Thesis Proposal by

Tejas Puranik

 (Advisor: Prof. Dimitri N. Mavris)

“A Methodology for Quantitative Data-Driven Safety Assessment for General Aviation”

Thursday, May 4, 2017 @ 9:00 a.m.

Weber Space Science and Technology Building
Collaborative Visualization Environment (CoVE)

Abstract:

The safety record of aviation operations has been steadily improving for the past few decades, however, accident rates in General Aviation (GA) have not improved significantly compared to scheduled airline operations. Per the Federal Aviation Administration (FAA), the demand for air travel and traffic is predicted to grow steadily through 2036 at a rate of approximately 1.8% annually with GA set to receive a much-needed revitalization. However, safety remains a major hurdle and with such a large increase in expected operations, there is an ever-increasing demand for improving safety of GA operations

Various data-driven safety programs such as Flight Data Monitoring (FDM) that exist in commercial aviation domain have percolated in GA with the aim of improving safety. These programs typically feature a continuous cycle involving data collection from on-board recorders, retrospective analysis of flight data records, identification of operational safety exceedances, design and implementation of corrective measures, and monitoring to assess their effectiveness. While these programs have been shown to be effective in reducing accident rates, there are certain obstacles in their widespread implementation in the GA domain. The variability in recorded parameters in GA, heterogeneity in GA fleet, different missions flown, etc. are some of the important hurdles. Additionally, existing techniques of analysis such as exceedance detection are designed to identify known unsafe conditions but are potentially blind to safety-critical conditions that may be captured in flight data records but are not present in the set of predefined safety events.

The overarching objective of this dissertation is to develop a methodology that can provide objective metrics for quantifying GA flight safety, enable identification of anomalous operations, and provide predictive capabilities that will complement existing approaches. The methodology presents the use of energy-based metrics as objective currency that can be used for quantifying safety across the heterogeneous GA fleet. These metrics are defined using recorded flight data and aircraft performance models. An anomaly detection framework is then developed using these metrics for identifying different types of anomalies (flight-level and instantaneous) in GA operations. A novel technique of calibrating aircraft performance models used in defining these metrics is proposed using data available in the public domain. The obtained calibrated models are further refined by employing Bayesian updating using actual flight data records. Finally, the calibrated performance models and a flight simulation model will be utilized to construct an offline surrogate model that approximates boundaries or limits of the performance envelope and can be queried online to identify when the aircraft might be drifting outside its safety space. Once the methodology is developed, its implementation on a set of real flight data records will be demonstrated through various experiments to highlight its widespread applicability.

Committee members:
      1.  Prof. Dimitri Mavris (Advisor)
           School of Aerospace Engineering, Georgia Institute of Technology
      2.  Dr. Simon Briceno
           School of Aerospace Engineering, Georgia Institute of Technology
      3.  Prof. Karen Marais
           School of Aeronautics and Astronautics, Purdue University