透過您的圖書館登入
IP:3.230.76.153
  • 學位論文

運用多層感知器預測航空維修品質績效指標

Using MLP to Predict the Trends of Quality Performance Indicators of Aviation Maintenance

指導教授 : 蕭育霖

摘要


航空業是由眾多專業組織彼此依賴、協同運作之產業,環境複雜且多變,而飛航安全作為所有航空業者在營運管理時的重要指標,如何有效提升航空器運作安全成為各航空公司戮力追求的方向。其中掌握航空器之運作安全、穩定的幕後功臣便為航機維護組織,維修人員除需必備的專業技術之外,也需有高度的集中力與耐心。除此之外,工作環境之妥善程度以及制度管理等都可能影響到維修組織之維修品質。然而,預防重於治療,與其在事故發生後逐步檢討,不如在事前分析可能影響維修品質的因素並做出預測,防範於未然。本研究旨在為航機修護組織建立維修品質預測模型,用以提升安全管理能量。首先根據業者自我督察或受相關單位稽核之歷史報告進行分類,依據報告因素按月份分為三大類:人員、環境及管理因素,再根據報告重要程度計算出三類因素之權重與品質績效指標(Quality Performance Indicators, QPI)。其次,利用連續三月之QPI作為輸入資料,以倒傳遞學習演算法(Back-propagation, BP)訓練權重後建立多層感知器(Multi-Layer Perceptron, MLP)預測模型,模型輸出結果為下月QPI之趨勢,包括上升、持平及下降三種變化。本研究之模型預測準確率,人員因素達70.0%、環境因素達81.8%、管理因素達66.7%,整體準確率可達七成。本研究顯示此類預測結果應具有實務應用上的參考價值,可納入作為業者建立安全管理預警指標之參考。

並列摘要


The aviation industry is highly complicated and is operated by many professional organizations. From the perspective of management, improving flight safety is the most important foundation of aircraft operations. Among all kinds of operations, aircraft maintenance repair organization (MRO) plays an essential role. The maintenance personnel must have professional skills and a high degree of concentration and patience. A good working environment and system management could also affect the quality of maintenance. However, prevention is more important than treatment. Instead of improvement after accidents, it is better to analyze the factors that may affect the quality of maintenance and make predictions beforehand. This research aims to establish a maintenance quality prediction model for an aircraft maintenance organization to improve its safety management. First, the historical audit reports of the company were categorized and divided into three categories based on the causes: Personnel, Environmental and Management. Quality Performance Indicators (QPI) were measured based on the importance of the report. Then QPI of the three categories were used as input data to establish the prediction model via Multi-Layer Perceptron (MLP) method to predict the future QPI. The predictive accuracy of the model was 70.0% for Personnel, 81.8% for Environmental, and 66.7% for Management factor. The results demonstrated the potential of the practical application of quality prediction, and can be taken as a reference for the industry to establish early warning indicators for aviation organizations.

參考文獻


中文部分
1. 國家運輸安全調查委員會(2019)。台灣飛安統計2009-2018。
2. 交通部民用航空局(2015)。民航通告-安全績效指標。
3. 鄭文通(2007)。安全管理系統與飛航服務安全查核。飛航天氣,第七期。
4. 交通部民用航空局(2002)。民航通告-機務自我督察作業。

延伸閱讀