臺灣博碩士論文加值系統

自2020年新冠肺炎疫情爆發以來，航空業面臨著多方面的嚴峻挑戰。根據國際間統計資料顯示，其中一項隱憂為飛機不穩定進場的情形較以往更加頻繁，可能的原因包含人員熟練程度下降、飛機重量、非典型的飛航管理模式等多重因素。有鑑於不穩定進場可能帶來飛行中失控、可控飛行撞地、衝出跑道等嚴重後果，且尚無相關研究針對我國探討疫情後航班進場安全的變化，因此本研究以桃園國際機場2019年8月與2020年8月的降落航班作為研究對象。考量到快速存取記錄器（QAR）的機密性質，本研究採用公開的廣播式自動相關監視（ADS-B）飛行數據進行後續分析，利用能量管理之概念取代傳統航空公司對不穩定進場的超限事件定義，改以能量超限面積作為評估飛機不穩定進場的依據，並比較疫情前後變化以觀察航班進場安全趨勢。為了更廣泛討論可能影響飛機能量狀態的潛在因素，本研究納入不同層面的參數包含天氣條件特徵、飛航管理特徵和航班進場特徵等，透過提取上述各層面的參數特徵後，本研究首先盤點出各特徵在兩年內的變化情形與分布狀況，初步證實疫情後、在飛航管理條件改變下，對航班進場特徵與能量超限面積皆有顯著影響。此外，本研究進一步應用機器學習演算法來建立回歸預測模型，以利深入探究上述特徵間的非線性關係，並輔以參數重要性分析來找出導致飛機能量管理超限的主要原因，以此突破了現有文獻對飛航管理因素的量化與分析方法，不僅成功驗證航班進場特徵與飛航管理特徵皆對於飛行能量管理具有顯著影響力，亦提供了飛航管理與飛行能量狀態之間潛在關係的可靠見解。

Since the outbreak of COVID-19 pandemic commenced in 2020, the aviation industry has faced severe challenges in multiple aspects. One of the arising concerns shows that flights tended to suffer unstable approaching more frequently than before due to the complex factors behind them, such as declining human proficiency and atypical air traffic patterns. Given the serious consequences that unstable approach might bring with, this thesis first aims at inspecting the variation in flight trajectories and energy management landing at Taoyuan International Airport (ICAO code: RCTP) over August 2019 and August 2020 as a comparison. Considering the confidentiality of Quick Access Recorder (QAR), the open-source flight data Automatic Dependent Surveillance-Broadcast (ADS-B) is utilized for analysis in this thesis. The concept of energy management is introduced to assess the flight approaching safety by constructing the energy boundaries and deriving energy outer area to replace conventional exceedance events of airlines’ Flight Operations Quality Assurance (FOQA) programs. To find out the potential reasons causing anomalies in flight energy state, several features are considered, including weather conditions, air traffic, and flight operation features. Furthermore, machine learning algorithms are applied to dig into the hidden non-linear relationship within these features. Through the comprehensive comparison over two years of datasets and thorough investigation of potential features, the result first shows that flight operation features and energy management significantly changed under the declined air traffic features. Moreover, the training results of machine learning models also verify that flight approaching features and air traffic features rank as the most significant features which could have an impact on energy outer area. The contribution of this thesis provides informative insights into the relevance between multiple features and flight energy states meanwhile better quantifying the impact of air traffic features.

中文摘要 I
Abstract II
Acknowledgments IV
Contents V
List of Tables VIII
List of Figures IX
I. Introduction 1
1.1. Research Background 1
1.1.1. COVID-19 pandemic challenges to air transport 1
1.1.2. Aviation Safety Issues Related to COVID-19 2
1.1.3. Unstable Approach Consequences: ICAO G-HRCs 4
1.2. Literature Review 5
1.2.1. Approaching Safety Analysis 5
1.2.2. Flight Energy Management 5
1.2.3. Impact of COVID-19 on aviation safety 7
1.2.4. Feature Importance Analysis for precursor identification 7
1.2.5. Summary and Discussion 9
1.3. Motivation & Objectives 11
1.4. Thesis Organization 12
II. Methodology 13
2.1. Research Process 13
2.2. Data Overview 15
2.2.1. Automatic Dependent Surveillance-Broadcast (ADS-B) 15
2.2.2. Aeronautical Information Publication Taipei FIR (AIP Taipei FIR) 16
2.2.3. Meteorological Aerodrome Reports (METARs) 18
2.3. Energy Boundary Construction 20
2.3.1. Energy Boundary 22
2.3.2. Ideal Energy State 23
2.3.3. Energy Outer Area 26
2.4. Machine Learning Algorithms 28
2.4.1. Random Forest 29
2.4.2. extreme Gradient Boosting (XGBoost) 30
III. Data preprocessing 34
3.1. METAR Data 34
3.2. ADS-B Data 39
3.2.1. Air Traffic Features 39
3.2.2. Flight Approaching Features 44
3.2.3. Energy Outer Area 48
IV. Evaluation of Approaching Trend under Epidemic 50
4.1. Flights Approach Profile Overview 50
4.2. Features Overview 54
4.2.1. Weather Features 55
4.2.2. Air Traffic Features 56
4.2.3. Flight Approaching Features 59
4.2.4. Energy Outer Area 60
V. Model Training and Discussion 62
5.1. Overview of Machine Learning Model 62
5.2. Feature Engineering 63
5.2.1. Feature Selection 63
5.2.2. Feature Scaling and Normalization 67
5.3. Model Architecture 70
5.3.1. Data Splitting: K-Fold Cross Validation 70
5.3.2. Hyperparameters Setting 71
5.3.3. Permutation Feature Importance 72
5.4. Results and Discussion 74
5.4.1. Scenario 1 (time window: 0-10 nm) 75
5.4.2. Scenario 2 (time window: 0-3 nm) 78
5.4.3. Casy studies 80
VI. Conclusion and Future work 86
Reference 89

[1]International Civil Aviation Organization (ICAO). "Effects of Novel Coronavirus (COVID-19) on Civil Aviation: Economic Impact Analysis." https://www.icao.int/sustainability/Documents/Covid-19/ICAO_coronavirus_Econ_Impact.pdf (accessed 27 June 2023).
[2]M. R. O. C. Civil Aeronautics Administration. "Traffic of Civil Aviation by Airports." https://www.caa.gov.tw/article.aspx?a=1736&lang=1 (accessed 27 June 2023.)
[3]International Civil Aviation Organization (ICAO). "COVID OPS Guidance." https://www.icao.int/safety/OPS/OPS-Normal/Pages/Covidopsguidance.aspx (accessed 27 June 2023.)
[4]International Air Transport Association (IATA). "Examining Unstable Approaches - Risk Mitigating Efforts Unstable Approach - Safety Analysis Project Team." https://www.iata.org/contentassets/7a5cd514de9c4c63ba0a7ac21547477a/examining-unstable-approaches_risk-mitigating-efforts.pdf (accessed 27 June 2023.)
[5]European Union Aviation Safety Agency (EASA). "Review of Aviation Safety Issues Arising from the COVID-19 Pandemic." https://www.easa.europa.eu/en/document-library/general-publications/review-aviation-safety-issues-arising-covid-19-pandemic-0 (accessed 27 June 2023.)
[6]International Air Transport Association (IATA). "Guidance for Flight Operations During and Post Pandemic." https://www.iata.org/contentassets/d0e499e4b2824d4d867a8e07800b14bd/iata-guidance-flight-operations-during-post-pandemic.pdf (accessed 27 June 2023.)
[7]Airbus. "Flight Operations Briefing Notes - Aircraft Energy Management during Approach." https://skybrary.aero/sites/default/files/bookshelf/166.pdf (accessed 27 June 2023.)
[8]International Air Transport Association (IATA). "2021 Safety Report." https://libraryonline.erau.edu/online-full-text/iata-safety-reports/IATA-Safety-Report-2021.pdf (accessed 10 July 2023.)
[9]International Air Transport Association (IATA). "Unstable Approaches - Risk Mitigation Policies, Procedures and Best Practices 2nd Edition." https://www.iata.org/contentassets/b6eb2adc248c484192101edd1ed36015/unstable-approaches-2016-2nd-edition.pdf (accessed 27 June 2023.)
[10]International Civil Aviation Organization (ICAO). "GASP Doc. 10004." https://www.icao.int/safety/GASP/Pages/GASP-Doc.-10004.aspx (accessed 27 June 2023.)
[11]D. Moriarty and S. Jarvis, "A systems perspective on the unstable approach in commercial aviation," Reliability Engineering & System Safety, vol. 131, pp. 197-202, 2014.
[12]H.-Y. Lai, C.-H. Chen, L.-P. Khoo, and P. Zheng, "Unstable approach in aviation: Mental model disconnects between pilots and air traffic controllers and interaction conflicts," Reliability Engineering & System Safety, vol. 185, pp. 383-391, 2019.
[13]S. J. Corrado, T. G. Puranik, O. J. Pinon, and D. N. Mavris, "Trajectory clustering within the terminal airspace utilizing a weighted distance function," in Proceedings, 2020, vol. 59, no. 1: MDPI, p. 7.
[14]S. J. Corrado, T. G. Puranik, O. P. Fischer, and D. N. Mavris, "A clustering-based quantitative analysis of the interdependent relationship between spatial and energy anomalies in ADS-B trajectory data," Transportation Research Part C: Emerging Technologies, vol. 131, p. 103331, 2021.
[15]G. Jarry, D. Delahaye, F. Nicol, and E. Feron, "Aircraft atypical approach detection using functional principal component analysis," Journal of Air Transport Management, vol. 84, p. 101787, 2020.
[16]Y. Jiao, H. Sun, C. Wang, and J. Han, "Research on unstable approach detection of civil aviation aircraft," Procedia computer science, vol. 131, pp. 525-530, 2018.
[17]T.-Y. Chiu and Y.-C. Lai, "Unstable Approach Detection and Analysis Based on Energy Management and a Deep Neural Network," Aerospace, vol. 10, no. 6, p. 565, 2023.
[18]D. Martınez et al., "Forecasting unstable approaches with boosting frameworks and lstm networks," 9th SESAR Innovation Days, 2019.
[19]J. L. Ackley, T. G. Puranik, and D. Mavris, "A supervised learning approach for safety event precursor identification in commercial aviation," in AIAA Aviation 2020 Forum, 2020, p. 2880.
[20]Flight Safety Foundation (FSF). "FSF ALAR Briefing Note 4.2 - Energy Management." https://flightsafety.org/wp-content/uploads/2016/09/alar_bn4-2-energymgmt.pdf (accessed 27 June 2023.)
[21]Flight Safety Foundation (FSF). "FSF ALAR Briefing Note 7.1 - Stabilized Approach." https://flightsafety.org/wp-content/uploads/2016/09/alar_bn7-1stablizedappr.pdf (accessed 28 June 2023.)
[22]Federal Aviation Administration (FAA). "Airplane Flying Handbook (FAA-H-8083-3C) Chapter 4: Energy Management: Mastering Altitude and Airspeed Control." https://www.faa.gov/regulations_policies/handbooks_manuals/aviation/airplane_handbook/media/05_afh_ch4.pdf (accessed 27 June 2023.)
[23]J. Sembiring, C. Liu, P. Koppitz, and F. Holzapfel, "Energy management for unstable approach detection," in 2018 IEEE International Conference on Aerospace Electronics and Remote Sensing Technology (ICARES), 2018: IEEE, pp. 1-6.
[24]S. G. Kumar, S. J. Corrado, T. G. Puranik, and D. N. Mavris, "Application of isolation forest for detection of energy anomalies in ADS-B trajectory data," in AIAA SCITECH 2022 Forum, 2022, p. 2441.
[25]G. Jarry, D. Delahaye, and E. Feron, "Flight safety during Covid-19: A study of Charles de Gaulle airport atypical energy approaches," Transportation research interdisciplinary perspectives, vol. 9, p. 100327, 2021.
[26]W.-C. Li, A. Nichanian, J. Lin, and G. Braithwaite, "Investigating the impacts of COVID-19 on aviation safety based on occurrences captured through Flight Data Monitoring," Ergonomics, no. just-accepted, pp. 1-39, 2022.
[27]M. R. O. C. Civil Aeronautics Administration. "eAIP TAIPEI FIR." https://ais.caa.gov.tw/eaip/ (accessed 27 June 2023.)
[28]Flight Safety Foundation (FSF). "FSF ALAR Briefing Note 8.2 - The Final Approach Speed." https://flightsafety.org/files/alar_bn8-2-apprspeed.pdf (accessed 13 July 2023.)
[29]A. Maheshwari, N. Davendralingam, and D. A. DeLaurentis, "A comparative study of machine learning techniques for aviation applications," in 2018 Aviation Technology, Integration, and Operations Conference, 2018, p. 3980.
[30]G. Gui, F. Liu, J. Sun, J. Yang, Z. Zhou, and D. Zhao, "Flight delay prediction based on aviation big data and machine learning," IEEE Transactions on Vehicular Technology, vol. 69, no. 1, pp. 140-150, 2019.
[31]J. K. Williams, "Using random forests to diagnose aviation turbulence," Machine learning, vol. 95, pp. 51-70, 2014.
[32]T. G. Puranik, N. Rodriguez, and D. N. Mavris, "Towards online prediction of safety-critical landing metrics in aviation using supervised machine learning," Transportation Research Part C: Emerging Technologies, vol. 120, p. 102819, 2020.
[33]S. Li, J. Qin, M. He, and R. Paoli, "Fast evaluation of aircraft icing severity using machine learning based on XGBoost," Aerospace, vol. 7, no. 4, p. 36, 2020.
[34]A. D. Midtfjord, R. De Bin, and A. B. Huseby, "A decision support system for safer airplane landings: Predicting runway conditions using XGBoost and explainable AI," Cold Regions Science and Technology, vol. 199, p. 103556, 2022.
[35]Y. Liu, R. Sun, and P. He, "Research on the Pre-Warning Method of Aircraft Long Landing Based on the XGboost Algorithm and Operation Characteristics Clustering," Aerospace, vol. 10, no. 5, p. 409, 2023.
[36]Abhishek Sharma. "Random Forest vs Decision Tree | Which Is Right for You?" https://www.analyticsvidhya.com/blog/2020/05/decision-tree-vs-random-forest-algorithm/ (accessed 27 June 2023.)
[37]T. Chen and C. Guestrin, "Xgboost: A scalable tree boosting system," in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 2016, pp. 785-794.
[38]Kaggle. "XGBoost." https://www.kaggle.com/code/alexisbcook/xgboost (accessed 27 June 2023.)
[39]Amazon Web Services (AWS). "How XGBoost Works." https://docs.aws.amazon.com/sagemaker/latest/dg/xgboost-HowItWorks.html (accessed 27 June 2023.)
[40]S. K. Goh, N. P. Singh, Z. J. Lim, and S. Alam, "Interpretable Tracking and Detection of Unstable Approaches Using Tunnel Gaussian Process," IEEE Transactions on Aerospace and Electronic Systems, 2022.)
[41]Flight Safety Foundation (FSF). "FSF ALAR Briefing Note 4.1 — Descent-and-approach Profile Management." https://flightsafety.org/wp-content/uploads/2016/09/alar_bn4-1-profilemgmt.pdf (accessed 14 July 2023.)
[42]Aviation Weather Center. "Static TAF Plots." https://www.aviationweather.gov/taf/help?page=plot (accessed 14 July 2023.)
[43]D. Xue, L.-T. Hsu, C.-L. Wu, C.-H. Lee, and K. K. Ng, "Cooperative surveillance systems and digital-technology enabler for a real-time standard terminal arrival schedule displacement," Advanced Engineering Informatics, vol. 50, p. 101402, 2021.
[44]International Civil Aviation Organization (ICAO). "Aircraft Operations - Volume I Flight Procedures (Doc 8168)." https://ffac.ch/wp-content/uploads/2020/11/ICAO-Doc-8168-Volume-I-Flight-Procedures.pdf (accessed 14 July 2023.)
[45]A. F. C. H. Civil Aeronautics Administration. "Air Traffic Management Procedures (ATMP) 16th edition." https://www.caa.gov.tw/article.aspx?a=260&lang=2 (accessed 14 July 2023.)
[46]G. Chen, J. Rosenow, M. Schultz, and O. Okhrin, "Using Open Source Data for Landing Time Prediction with Machine Learning Methods," in Proceedings, 2020, vol. 59, no. 1: MDPI, p. 5.
[47]Jeppesen, EASA ATPL Training - Radio Navigation. 2013.
[48]S. Dipanjan, B. Raghav, and S. Tushar. "Feature Engineering and Selection." https://link.springer.com/chapter/10.1007/978-1-4842-3207-1_4 (accessed 28 June 2023.)
[49]D. Muñoz-Esparza, R. D. Sharman, and W. Deierling, "Aviation turbulence forecasting at upper levels with machine learning techniques based on regression trees," Journal of Applied Meteorology and Climatology, vol. 59, no. 11, pp. 1883-1899, 2020.
[50]R. A. Burnett and D. Si, "Prediction of injuries and fatalities in aviation accidents through machine learning," in Proceedings of the International Conference on Compute and Data Analysis, 2017, pp. 60-68.
[51]Rebecca Patro. "Cross-Validation: K Fold vs Monte Carlo." https://towardsdatascience.com/cross-validation-k-fold-vs-monte-carlo-e54df2fc179b (accessed 27 June 2023.)
[52]Scikit-learn (SKlearn). "sklearn.ensemble.RandomForestRegressor." https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestRegressor.html (accessed 15 July 2023.
[53]Kaggle. "A Guide on XGBoost hyperparameters tuning." https://www.kaggle.com/code/prashant111/a-guide-on-xgboost-hyperparameters-tuning (accessed 15 July 2023.)
[54]Scikit-learn (SKlearn). "Permutation feature importance." https://scikit-learn.org/stable/modules/permutation_importance.html#id2 (accessed 27 June 2023.)