No abstract

No abstract

General Aviation (GA) accounts for the vast majority of aviation accidents in the United States every year. As a result of the implementation of Flight Data Recorders (FDR) in GA aircraft, we now have FDR flight data that we can use to improve safety, by detecting and correcting unsafe behaviors and habits. Because of the high variability in GA operations and GA flight performance, detecting these safety events is not trivial. In this paper, we evaluate the usefulness of an event-driven method in characterizing the safety of the approach phase of a flight. In particular, we develop algorithms to detect safety events during the approach phase from FDR data from a G1000 glass cockpit display on a Cirrus SR-20 fleet. We adjust the current safety event definitions to the SR20, and then change the limits of the safety events, to make the safety events in our dataset more consistent with each other. While changing the definitions gave us meaningful results, we also suggest that new metrics are developed to be used together with safety events in future work.

The probability of an airplane deviation from pre-planned trajectory is a core of aviation safety analysis. We propose to use a mixture of three probability density distribution functions it the task of aviation risk assessment. Proposed model takes into account the effect of navigation system error, flight technical error, and occurrence of rare events. Univariate Generalized Error Distribution is used as a basic component of distribution functions, that configures the error distribution model from the normal error distribution to double exponential distribution function. Statistical fitting of training sample by proposed Triple Univariate Generalized Error Distribution (TUGED) is supported by Maximum Likelihood Method. Optimal set of parameters is estimated by sequential approximation method with defined level of accuracy. The developed density model has been used in risk assessment of airplane lateral deviation from runway centreline during take-off and landing phases of flight. The efficiency of the developed model is approved by Chi-square, Akaike’s, and Bayes information criteria. The results of TUGED fitting indicate better performance in comparison with double probability density distribution model. The risk of airplane veering off the runway is considered as the probability of a rare event occurrence and is estimated as an area under the TUGED.

No abstract