中文摘要
題 目：安全裕度-飛航人為疏失風險評估
研 究 生：林鈺峰
指導教授：景鴻鑫

　　本研究的目的，在建立一套可量化的飛航人為疏失風險評估工具。本研究提出”飛航安全裕度”的觀念，來代表飛航組員操作飛機的安全空間，再以該空間被壓縮的程度來呈現人為疏失所造成的風險。由於人為疏失分類上的不易量化，以及風險認知的見仁見智，本研究採用情境參數，來表達人為疏失所造成之影響，從而顯示出其嚴重性。所謂的情境參數，是指SHELL模型中，除了人(L)以外的所有其它參數，包含軟體、硬體、與環境。在給定情境之下，本研究以偏離正常標準情境多遠，也就是飛行員需要多大的所需綜合飛行能力，才能把飛機飛回正常標準情境，來表現該情境之嚴重性。情境偏離標準越遠，表示飛航安全裕度被壓縮得越嚴重，風險也越高。本研究以國內某事件為例，選取十個瞬間，分別收集相關的情境參數，再透過專家訪談，請資深機師提供，將各該情境飛回標準正常情境，所需之綜合飛行能力，換算成飛航安全裕度，再透過類神經網路的學習，建立任意情境與飛航安全裕度的因果關係。結果顯示，本方法可以跨越認知的限制，針對人為疏失的風險，很合理的提供量化評估。如能與模擬機結合，將可進一步驗證本法的正確性。

Abstract
Subject: Safety Margin
-Evaluation of the Risk induced by Human Error in Flight
Student: Yu-feng Lin
Advisor: Hung-Sying Jing

　A scientific tool, Safety Margin, base on the artificial neural network for evaluation of the risks induced by human errors is proposed in this study. The flight safety margin is an expert system designed to numerically evaluate the perceived consequences caused by human errors. It represents how much room still left for the crew member to operate under the threat from the human error viewing as the suppression of the safe operation margin. That is to say the risk is geometrized as the safety margin. In the field of risk analysis, the key difficulty is in that the risk is a perception problem. Although the possibility of occurrence can be defined precisely, the perceived severity will be different from person to person. In addition to the perception problem, the even more difficult part lies in that the outcome will be different even with the same human error given different situations in which the error occurs, not to say the difficulty of quantification of the human errors. Understanding that situations will be changed to be abnormal with the existence of human error, a group of situation parameters are defined from the SHELL model excluding the livewares. All the situation parameters are used as the inputs of the training examples for the neural network. A questionnaire is designed for the pilots to answer that what kind of performance is needed for the crew to recover from the given situations back to the normal, standard situation. The performance is evaluated from both the physiological and psychological points of view. The results are then converted to the flight safety margin, representing the outputs of the training examples. The corresponding expert system can then be established by using the neural network. The tool has been tested with a real case, meaningful results are obtained although there are still much room for improvement. And it will be evaluated well, if this tool and simulator in one combine.

參考文獻
中文文獻
[1].宋明哲. “風險管理”,中華企業管理發展中心, 1984
[2].鄭燦堂. “風險管理-理論與實務”,三民書局, 1989
[3].中華民國管理科學學會, “企業的危機應變與風險管理”, 1992
[4].田德智 “執行失事預防計畫 防範人為錯誤產生”, 中華民國八十三年航空安全研討會, 中正理工學院, 1994。
[5].黃國裕,“財務比率在股票異常報酬之預測能力分析-類神經網路”,中央大學資訊管理研究所,碩士論文,1994。
[6].葉怡成,“應用類神經網路”,台北市,儒林圖書公司,1997。
[7].民用航空局，中華民國八十七年民航統計年報，1998。
[8].景鴻鑫“本土化之飛安理念”，中華民國八十七年，飛航安全檢討與提昇研討會，交通部民用航空局，1998。
[9].何立己, 黑盒子的秘密-航空安全人為因素剖析, AIRWAY世界民航雜誌，1998。
[10].航空器失事調查報告書, 空中巴士 A300B4 – 622R, B-1814 桃園大園 中華民國八十七年二月十六日” 中華民國 台灣 交通部民用航空局. 中華民國八十九年五月十八日, 2000.
[11].中華民國飛行安全基金會”1990’s航空安全回顧”，飛航安全季刊，22期，15~20頁，2000。
[12].陳順宇. 迴歸分析. 華泰書局，2000。
[13].廖達琪、景鴻鑫、楊連誠,“類神經網路系統與選舉預測-2002年北高兩市市議員選舉之案例探討”,選舉預測模型學術研討會,中央研究院中山人文社會科學研究所,2003。

英文文獻
[1].Boeing Commercial Airplane Group “Statistical Summary of Commercial Jet Aircraft Accident Worldwide Operations 1959-1993”, Airplane Safety Engineering, 1994.
[2].Boeing Commercial Airplane Group, Statistical Summary of Commercial Jet Airplane Accidents-Worldwide Operations 1959-2003, 2004.
[3].Berke, L., et al.,“Optimum design of aerospace structural components using neural networks”, Computers ﹠ Structures, 48（6）, pp.1001-1010, 1993.
[4].Edwards, E.”Man and Machine : System for Safety”, Proceedings of the BALPA Technical Symposium, London, 1972.
[5].Garris, M. D. et al.,“Methods for enhancing neural network handwritten character recognition”, IJCNN-91, I, pp.695-700, 1991.
[6].Helmreich, R.L.”Cockpit Management Attitude”, Human Factors, Vol.26, pp.63-72, 1984.
[7].Haykin, s.,“Neural Networks: A comprehensive Foundation”, U.S.: Prentice Hill, 1999.
[8].Perrow, C.”Normal Accidents: Living with High Risk Technologies”, NJ : Princeton University Press, 1984.
[9].Reason, J.”Human Errors”, New York : Cambridge University Press ,1990.
[10].Rosenblatt, F.,“The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain”, Paych. Rev. 65, pp.386-408, 1958.
[11].S. R. Trollip, R. S. Jensen. Human factors for general aviation. Jeppesen Sanderson, Inc., UDA, 1991.