本研究主要目的為針對異常事件航班最後進場階段之操作品質表現進行模態判定。本研究以超速事件為例，透過監督式類神經網路訓練異常事件航班每瞬間的所有參數，得到一個以整體概念為出發點之異常事件網路；再接續利用監督式類神經網路判斷飛行參數超標時間區間內何處有縱向模態產生。藉由類神經網路這種非線性統計性數據建模工具的重複使用，確認進場階段異常區間內縱向模態的發生，進而提供FOQA數據更為深入的理解，對評估飛航操作品質有所助益，並可進一步提供飛航安全數據量化評估的基礎。本研究在測驗航班的超速判定中平均誤判率為3.7%，而縱向模態判定之誤判率為2.9%，未來可望藉由更多數據與更多種類之異常事件航班建立更完整之模態網路。

The goal of this study is to identify the model from the FOQA data obtained in the final approach phase of event flight with neural network for the evaluation of the quality of flight operation and the establishment of engineering analysis of flight safety. In order to analysis the FOQA data in a comprehensive aspect, this study used supervised learning neural network to train all 17 parameters of event flight as an input for each second in case of overspeed; then created another supervised learning neural network to identify the longitudinal modes during the overspeed period. By duplicated use of this nonlinear statistical learning model, the neural network, the longitudinal modes of event period in the final approach phase are expected to be identified. With analysis and investigation of these flight modes, the quality of flight operation can be understood more thoroughly and it can be helpful in the establishment of the engineering analysis of flight safety. The results of this study show that the misjudgment rate of overspeed net is 3.7% and the net for longitudinal modes is 2.9%; it is hoped that with more and different kinds of FOQA data the neural network can be used to identify a variety of event in the final approach phase in the follow-up studies

[1] International Air Transport Association, “Safety Report 2013”, 2013.
[2] Federal Aviation Administration, “Introduction to Safety Management Systems for Air Operators”, Advisory Circular AC No：120-92, 2006.
[3] 交通部民用航空局，“安全管理系統”，民航通告編號：AC120-032C，中華民國一佰年一月。
[4] H.W. Heinrich, “Industrial Accident Prevention”, 1931.
[5]Boeing Commercial Airplane Group, “Flight Safety and Accident Investigation Workshop”, IAA, NCKU, 1994.
[6] Reason, J., “Human Errors”, New York, Cambridge University Press, 1990.
[7] Perrow C., “Normal Accidents: Living with High Risk Technologies”, NJ, Princeton University Press, 1984.
[8] U.S. General Accounting Office, Aviation Safety: U.S. Efforts to Implement Flight Operational Quality Assurance Program, Flight Safety Digest, Vol.17, no. 7-9, pp.1-36, 1988.
[9] 景鴻鑫，“本土化之飛安理念”，飛航安全檢討與提昇研討會，國立成功大學，中華民國八十七年。
[10] 鍾華興，“飛航安全之工程分析-線性系統觀點”，國立成功大學民航研究所碩士論文，中華民國九十七年六月。
[11] 陳品妤，“民航操作品質數據之安全性分析”，國立成功大學民航研究所碩士論文，中華民國九十九年六月。
[12] 許澔瑋，“飛航癥候初步探討-幾何分析”，國立成功大學航空太空工程研究所碩士論文，中華民國一佰零一年六月。
[13] 蘇愛琦，“飛航癥候之初步探討-相關性分析”，國立成功大學民航研究所碩士論文，中華民國一佰零一年六月。
[14] Mehryar Mohri, Afshine Rostamizadeh, Ameet Talwalker (2012) Foundations of Machine Learning, The MIT Press ISBN 9780262018258
[15] S. Kotsiantis, Supervised Machine Learning: A Review of Classification Techniques, Informatica Journal 31 (2007) 249-268.
[16] Rumelhart, David E. Hinton, Geoffrey E. Williams, Ronald J.(8 Octorber 1986)” Learning representations by back-propagating errors”. Nature 323 (6088):533-536