- 學位論文
運用類神經網路法建立航空安全預測模式
Using Neural Network Methods to Establish a Flight Safety Prediction Model
摘要
因損失成本及耗費社會資源甚鉅，防範航空意外發生一直是航空業者最大的努力目標。近年來，飛航事件頻傳，各國業者對於飛航安全的改善無不更顯重視，而其中因大多數飛安事件的發生與人為因素相關，因此，降低人為疏失的發生便成為了當今飛安管理最重要的課題。據此，本研究以我國民航局飛安查核資料為依據，以HFACS-MA模式分析找出其中存在的人為因素，並依據量化公式建立人為因素指標，運用倒傳遞(Back Propagation Neural Network，BPN)與輻狀基底函數(Radial Basis Function Neural Network, RBF)網路法，建立飛安績效預測模式。研究成果包括找出兩種績效預測模式之最佳參數配置，以及運用敏感度分析，找出飛安績效之重要影響因子，以做為飛安績效改善之參考依據。此研究之貢獻在於探討不同人為因素對於飛安之影響程度，進一步提升飛安管理能量，減少未來飛安事件發生之嚴重性。
並列摘要
The airline industry always do their hard to prevent aviation accidents that will cost too much social resources. In recent years, the airline industry all the world improve their own flight safety actively. Most flight safety system and the occurrence of events related to human factors, therefore, reduce the incidence of human error has become the flight safety management of the most important issues. In this study, the informations based on the Department of Civil Aviation Flight Safety Check data. We create the flight safety performance prediction mode with Back Propagation Neural Network (BPN) network method in previous studies. In this study, we revise the formula of quantification of human factors, and then use the Radial Basis Function Neural Network (RBF) network method to find the best predictive mode parameter configuration We will re-establish the human factor quantization formula and find the best Parameter, and combined with sensitivity analysis to find out the impact flight safety factor as the reference basis for improving performance.