A Causal Correlations Analysis with Multi-sample Controlled Flight Into Terrain Incidents
-
摘要: 为深入探究不同等级后果事件致因关联研究不足的问题，基于“安全Ⅱ”视角，以可控飞行撞地（controlled flight into terrain，CFIT）类高风险事件为例，分析不同后果样本事件致因及其关联性。选取航空安全网（aviation safety network，ASN）128篇、航空安全报告系统（aviation safety reporting system，ASRS）354篇CFIT事件为样本，依据威胁与差错管理模型（threat and error management，TEM）分析事件潜在的条件、威胁、机组差错、航空器非预期状态和措施，识别出3 367个致因因素和2 169条因果/时序关系，经语义分析整合、因果/时序关系梳理建立基于46个贝叶斯网络节点的样本事件演化网络模型；依据贝叶斯网络和关联规则计算节点间先验概率、条件概率及相关性，确定样本事件高概率节点和链路。结果表明：①运行压力—机组疲劳—失去情景意识—错误设置高度表—航空器偏离高度路径是导致CFIT事故的高风险路径；②机组疲劳（57%）、航司/局方管理不足（17%）等消极因素出现时，发生可控飞行撞地事故概率增加24%；③从“安全Ⅱ”的视角展开研究，航空系统的应对措施（78%）和执行改出动作（84%）等积极因素生效时，稳定进近的概率增加34%，可有效降低发生可控飞行撞地的发生风险。Abstract: To address the research gap regarding the causal correlations among incidents with varying consequence severities, this study adopts a Safety-Ⅱ perspective and takes controlled flight into terrain (CFIT), a high-risk aviation accident category, as the research object to analyze the causal factors and the interrelationships across samples with distinct outcome levels. A total of 128 CFIT accident investigation reports from the Aviation Safety Network (ASN) and 354 voluntary reports from the Aviation Safety Reporting System (ASRS) are selected as the analytical sample. Guided by the threat and error management (TEM) model, a systematic analysis is conducted to identify latent conditions, threats, flight crew errors, undesired aircraft states, and countermeasures inherent in the sampled accidents, which resulted in the identification of 3 367 causal factors and 2 169 causal-temporal relationships. Following semantic analysis and integration of these relationships, a Bayesian network (BN) model is constructed, resulting in an accident evolution network model comprising 46 BN nodes. Association rule mining is employed to compute conditional probabilities and inter-node correlations, so as to delineate the high-probability causal chains of the samples. Results demonstrate that: ①the primary high-risk pathway leading to CFIT accidents is: operational pressure → flight crew fatigue → loss of situational awareness → incorrect altimeter setting → aircraft deviation from the intended altitude profile. ②When negative factors including flight crew fatigue (accounting for 57% of the identified negative factors) and inadequate management by airlines/regulatory authorities (17%) are present, the probability of CFIT accidents increases by 24%. ③From the Safety-Ⅱ perspective, when positive factors, such as aviation system response measures (with an effectiveness rate of 78%) and the execution of recovery actions (84%) take effect, the probability of a stable approach is enhanced by 34%, thereby significantly mitigating the risk of CFIT occurrences.
-
图 2 样本1 CFIT事故调查报告地域分布
Figure 2. Geographic distribution of CFIT accident investigation reports of sample 1
图 6 样本2高概率因果/时序链路
Figure 6. The high probability causal/chronological pathway of sample 2
图 8 样本2弱控制因果/时序链路
Figure 8. The weakly controlled causal/chronological link of sample 2
表 1 TEM模型重要概念
Table 1. Important concepts of the TEM model
重要概念 概念含义 潜在的条件 事故/事件发生前系统中存在的条件，由各种可能的因素引发 威胁 发生在机组人员影响范围之外的事件或错误，但若要保持安全系数，则需要机组人员的关注和管理 机组差错 观察到的机组偏离组织期望或机组意图的情况 航空器非预期状态 由机组人员引起的明显降低安全系数的航空器状态；错误管理不力导致的安全受损情况。非预期航空器状态可恢复 措施 帮助机组人员管理威胁、错误和非预期航空器状态的各种工具和技术 
下载: 导出CSV
表 2 1892245号报告影响因素与关系分析
Table 2. Influencing factors and relationships in ACN 1892245 report
1892245号报告内容 因果/时序关系 今晚日落后从CNO直飞ZZZ时，我计划爬升到792.48 m。但我误读了高度计的读数，在487.68 m时就平飞了。黄昏时分光线昏暗，我的头灯角度向下，高度计受到遮挡，这些都是造成误差的原因。在天色渐暗但尚未完全黑透的情况下，很难判断我在地面上的高度。当我接近奇诺山时，GPS地形警告系统启动，促使我重新检查高度计读数，发现我比计划高度低了304.8 m。我立即开始爬升，并在762 m（MSL）的高度飞离了丘陵 昏暗的照明条件→误读高度计 头灯角度向下→误读高度计 误读高度计→在487.68 m（MSL）高度处平飞 在487.68 m（MSL）高度处平飞→比计划高度低了304.8 m 山脉→GPS地形警告系统启动 比计划高度低了304.8 m→ GPS地形警告系统启动 GPS地形警告系统启动→重新检查高度计读数 重新检查高度计读数→GPS地形警告系统启动
下载: 导出CSV
表 3 致因因素与TEM对应重要概念
Table 3. Influencing factors and corresponding concepts of TEM model
影响因素 TEM对应重要概念 山脉 威胁 昏暗的照明条件 误读高度计 机组差错 在487.68 m高度处平飞
GPS地形告警系统启动航空器非预期状态 重新检查高度计读数
立即爬升措施
下载: 导出CSV
表 4 致因因素的类型及数量
Table 4. Types and number of influencing factors
因素类型 样本1各因素数量 样本2各因素数量 潜在的条件L 93 144 威胁T 355 718 机组差错F 189 518 非预期航空器状态U 167 323 措施C 157 703 共计 961 2 406
下载: 导出CSV
表 5 CFIT贝叶斯网络节点
Table 5. CFIT Bayesian network nodes
重要概念 编码 节点 示例 潜在的条件L L1 陌生/无经验 不熟悉的机场、不熟悉的飞机、机组人员经验不足、度假、加拿大的不同做法 L2 熟悉/有经验 熟悉的机场、以往的经验 L3 生活压力 家庭成员, COVID-19 L4 缺少技术和设备 未安装可用的安全设备（EGPWS、预测风切变、TCAS/ACAS等）、错误航图 L5 完美的天气/极佳的能见度 塔台在望，跑道在望 L6 维修 机械故障，没有氧气补充，没有燃料补充 L7 航空公司/民航局管理不足 L8 缺乏培训 威胁T T1 环境运行压力 TCAS TA/RA、（可能的）交通堵塞、野火、天线不明显、野生动物 T2 恶劣天气 结冰条件、潮湿天气、雾霾、阳光刺眼、雷暴/降雨、湍流/风切变/尾风 T3 疲劳 精神疲劳、机组疲惫、睡眠不足、不能放松 T4 空管威胁 空管员未能提供指导/做出回应、空管员语言困难、空管员跑道/目的地变更、空管员繁忙、空管员缺乏经验 T5 任务饱和 工作量大，同时监控CTAF和中心频率 T6 分心 分心使用iPad、分心进行配置、放松警惕、自满情绪 T7 地形 山脊、山脉、树木、人造建筑、塔台 T8 光线 夜晚、眩光反射、黑洞幻觉、日落 T9 航司运营压力/运营规划和调度 准点率压力、延误、晚到飞机或机组人员、设备变更、航班时刻压力、前一航段延迟到达、长时间飞行、飞行计划变更、健康和福利问题 T10 飞机自身威胁 自动驾驶仪/ILS/FMS失效、系统/发动机/飞行控制/自动化/无线电/高度表异常、发动机噪音、TCAS未正常工作、未收到GPWS警告、风扇声音大、遮阳板、无航图 T11 能见度差 机组差错F F1 自动化错误/飞行控制错误/系统无线电仪表问题 高度设置不正确、未使用VNAV、航向设置不正确、DME设置不正确、频率拥挤、飞行控制垂直/横向和/或速度偏差、未设置/检查高度计 F2 检查单问题 未完成着陆检查单，未使用检查单/在错误的时间执行检查单 F3 违反标准操作程序/无标准操作程序交叉验证 未遵守SOP、无菌环境、CRM不佳、未监控飞行链路 F4 机组对外理解错误 误解指令/自动操作，未与ATC沟通，与现场失去联系，未尝试进行澄清 F5 失去情景意识 机组人员忽略了潜在的障碍物，固步自封，只关注目视程序 F6 继续飞行 F7 对警报迟钝 对EGPWS警告不敏感 航空器非预期状态U U1 飞机操作/手动飞行偏差 垂直/横向/速度偏差、俯仰下降、进场更快、下降过晚/过早、航向错误、矢量过高过紧、过冲最终航向、手飞垂直偏差 U2 保持高下降率 U3 进近不稳定 U4 未经许可进入空域 军队空域 U5 超出飞机限制的操作 超过速度限制 U6 向地形飞行 受控飞向地形（人造建筑/桥梁/山脉/树木） U7 低于指定高度 低于MDA、低于MSA、低于MVA、低于MEA、低于FAF高度、低于拦截高度、未能保持足够高度 U8 低于/高于下滑道 U9 地面导航 走错滑行道、匝道、登机口或停机位 措施C C1 基于航空系统的应对措施 TAWS/GPWS/EGPWS、红色PAPI灯/红色VASI灯 C2 排除自动化设置故障 手动输入航向和频率，重设高度，排除无线电设置故障 C3 监控/交叉检查 C4 遵守RA C5 来自ATC的低空警报 C6 来自ATC的指导 C7 机组人员回复外部/请求指导 C8 执行改出动作 脱离自动驾驶仪、增加动力、平飞、开始爬升、减少推力、绕行/错过进近、右转 C9 警告设备无响应 没有EGPWS/GPWS警告 C10 稳定进近 C11 造成事故
下载: 导出CSV
表 6 样本2致因因素先验概率计算结果（部分）
Table 6. Results of a priori probability calculations of influencing factors for Sample 2 (partial)
节点 状态发生 状态不发生 T2恶劣天气 0.26 0.74 T 10飞机自身威胁 0.07 0.93 T9航司运营压力/运营规划和调度 0.07 0.93 T4空管威胁 0.13 0.87 F 1自动化错误/飞行控制错误/系统无线电仪表 0.17 0.83 T6分心 0.06 0.94 T 1环境运行压力 0.1 0.9 L2熟悉/有经验 0.03 0.97 T8光线 0.04 0.96 L5完美的天气/极佳的能见度 0.03 0.97 T7地形 0.23 0.77 L 1陌生/无经验 0.18 0.82
下载: 导出CSV
表 7 样本2致因因素条件概率计算结果（部分）
Table 7. Results of conditional probability calculations for Sample 2 influencing factors (partial)
规则前项 规则后项 置信度 L5完美的天气/极佳的能见度 C 10稳定进近 0.666 7 L2熟悉/有经验 C 10稳定进近 0.555 6 L4缺少技术和设备 U9地面导航 0.200 0 U4未经许可进入空域 C6来自ATC的指导 0.125 0 T8光线 C7机组人员回复外部/请求导航 0.071 4
下载: 导出CSV
表 8 T5节点的父节点对应条件概率
Table 8. Conditional probabilities corresponding to the parent node of T5 node
节点 状态 T4空管威胁 Y N N T2恶劣天气 N Y N P（T5任务饱和=Y|Node i） 0.11 0.03 0
下载: 导出CSV
表 9 Noise or Gate方法计算T5节点概率示例
Table 9. Example of Noise or Gate method to calculate probabilities of T5 node
节点状态 P(T5=Y|Node i) P（T5=N|Node i） T4=N T2=Y 0.03 0.97 T4=Y T2=N 0.11 0.89 T4=Y T2=Y 0.14 0.86
下载: 导出CSV
-
[1] IATA. IATA controlled flight into terrain accident analysis report 2008-2017 data[R]. Montreal, Canada: International Air Transport Association, 2018. [2] ICAO. Celebrating TAWS 'Saves': But lessons still to be learnt[EB/OL]. (2006-06-29)[2024-10-01]. https://www.icao.int/safety/implementation/lists/library1/public.aspx .[3] ICAO. Annual ICAO Safety Report[R/OL]. (2023-08-23)[2023-10-14]. https://www.icao.int/safety/Pages/Safety-Report.aspx .[4] 刘清贵. 韩亚航空214航班事故解析[J]. 中国民用航空, 2013(8): 27-28.LIU Q G. What Asiana airlines flight 214 accident tells us[J]. China Civil Aviation, 2013(8): 27-28. (in Chinese). [5] KELLY D, EFTHYMIOU M. An analysis of human factors in fifty controlled flight into terrain aviation accidents from 2007 to 2017[J]. Journal of Safety Research, 2019, 69: 155-165. doi: 10.1016/j.jsr.2019.03.009 [6] 王洁宁, 钟彬. 人为差错导致的可控飞行撞地风险研究[J]. 中国民航大学学报, 2019, 37(3): 38-43.WANG J N, ZHONG B. CFIT risk caused by human errod[J]. Journal of Civl Aviaiton University of China, 2019, 37 (3): 38-43. (in Chinese) [7] GUO C, SUN Y, SU S, et al. Risk assessment method for controlled flight into terrain of airlines based on QAR data[J]. Aircraft Engineering and Aerospace Technology, 2023, 95 (8): 1184-1193. doi: 10.1108/AEAT-10-2022-0269 [8] 孙瑞山, 占欣. 基于ISM—CRITIC法的通用航空可控飞行撞地影响因素分析[J]. 中国安全生产科学技术, 2018, 14(1): 129-135.SUN R S, ZHAN X. Analysis on influence factors of controlled flight into terrain of general aviation based on ISM-CRITIC method[J]. Joural of Safety Science and Technology, 2018, 14(1): 129-135. (in Chinese) [9] 魏燕明, 甘旭升, 游雯雯, 等. 基于ISM模型的可控飞行撞地诱因关系分析[J]. 火力与指挥控制, 2018, 43(2): 172-176.WEI Y M, GAN S X, YOU W W, et al. Inducement relationship analysis of controlled flight into terrain based on ISM model[J]. Fire Control & Command Control, 2018, 43(2): 172-176. (in Chinese) [10] MENG B, LU N. A hybrid model integrating HFACS and BN for analyzing human factors in CFIT accidents[J]. Aerospace, 2022, 9(11): 711. doi: 10.3390/aerospace9110711 [11] 张晓全, 吴贵锋. 功能共振事故模型在可控飞行撞地事故分析中的应用[J]. 中国安全生产科学技术, 2011, 7(4): 65-70.ZHANG X Q, WU G F. Application of FRAM model in CFIT accident analysis[J]. Joumal of Safety Science and Technology, 2011, 7(4): 65-70. (in Chinese) [12] 汪磊, 孙瑞山. CFIT事故中的人为因素[J]. 中国民航大学学报, 2007(增刊1): 80-82.WANG L, SUN R S. Human factors in CFIT accidents[J]. Journal of Civil Aviation University of China, 2007(S1): 80-82. (in Chinese) [13] 杜尹岚. 基于贝叶斯网络的CFIT事件量化分析[D]. 天津: 中国民航大学, 2021.DU Y L. Quantitative analysis of CFIT events based on Bayesian network[D]. Tianjin: Civil Aviation University of China, 2021. (in Chinese) [14] 李冰滔. 基于贝叶斯网络模型的可控飞行撞地研究[D]. 广汉: 中国民用航空飞行学院, 2025.LI B T. Controlled flight into terrain study based on Bayesian network modeling[D]. Guanghan: Civil Aviation Flight University of China, 2025. (in Chinese) [15] 杜红兵, 王雪莉. 基于贝叶斯网络的可控飞行撞地事故原因分析方法[J]. 安全与环境学报, 2009, 9(5): 136-139.DU H B, WANG X L. Cause analysis of the controlled flight bumping into terrain based on the Bayesian network[J]. Journal of Safety and Environment, 2009, 9(5): 136-139. (in Chinese) [16] HOLLNAGEL E. Safety-Ⅰ and Safety-Ⅱ: the past and future of safety management[M]. London: CRC press, 2014. [17] KUHN K D. Using structural topic modeling to identify latent topics and trends in aviation incident reports[J]. Transportation Research Part C: Emerging Technologies, 2018, 87: 105-122. doi: 10.1016/j.trc.2017.12.018 [18] DNM Safety. From Safety-Ⅰ to Safety-Ⅱ: a white paper[EB/OL]. (2022-05-21)[2025-01-16]. https://skybrary.aero/sites/default/files/bookshelf/2437.pdf .[19] American Airlines. Trailblazers into Safety-ll: American airlines' learning and improvement team[EB/OL]. (2020-04-23)[2025-01-16]. https://skybrary.aero/articles/trailblazers-safety-ii-american-airlines-learning-and-improvement-team .[20] American Airlines. Charting a new approach: what goes well and why at American airlines[EB/OL]. (2020-04-23)[2025-01-16]. https://skybrary.aero/sites/default/files/bookshelf/AA%20LIT%20White%20Paper%20II%20-%20SEP%202021.pdf .[21] BAO M Y, DING S T. Individual-related factors and management-related factors in aviation maintenance[C]. 3rd International Symposium on Aircraft Airworthiness, Toulouse, France: ISAA, 2014. [22] 刘俊杰, 张丽娟. 飞行疲劳事件语义网络分析[J]. 中国安全科学学报, 2016, 26(1): 34-39.LIU J J, ZHANG L J. Semantic network analysis of flight fatigue events[J]. China Safety Science Journal, 2016, 26(1): 34-39. (in Chinese) [23] GAO Y, HAO Y, WANG S, et al. The dynamics between voluntary safety reporting and commercial aviation accidents[J]. Safety Science, 2021, 141. [24] 袁乐平, 张文东, 赵力梵, 等. 基于TEM模型的民航不安全事件关联规则挖掘与分析[J]. 中国安全生产科学技术, 2022, 18(10): 31-36.YUAN L P, ZHANG W D, ZHAO L F, et al. Mining and analysis of association rules in civil aviation unsafe incidents based on TEM model[J]. Journal of Safety Science and Technology, 2022, 18(10): 31-36. (in Chinese) [25] ICAO. Threat and error management(TEM)in air traffic control[EB/OL]. (2005-10-06)[2024-10-02]. https://flightsafety.org/files/tem_icao_10-05.pdf .[26] IATA. IATA safety report 2022[EB/OL]. (2023-03-07)[2024-10-01]. https://www.iata.org/en/pressroom/2023-releases/2023-03-07-01/ .[27] LI M, WANG D, SHAN H. Risk assessment of mine ignition sources using fuzzy Bayesian network[J]. Process Safety and Environmental Protection, 2019, 125: 297-306. doi: 10.1016/j.psep.2019.03.029 [28] ZHANG C, LIU C, LIU H, et al. Incorporation of pilot factors into risk analysis of civil aviation accidents from 2008 to 2020: a data-driven bayesian network approach[J]. Aerospace, 2023, 10(1): 9. [29] JI C, SU X, QIN Z, et al. Probability analysis of construction risk based on noisy-or gate Bayesian networks[J]. Reliability Engineering & System Safety, 2022, 217: 107974. [30] ICAO. Procedures for air navigation services - training - (Doc 9868 3rd Edition)[EB/OL]. (2023-01-31)[2024-10-01]. https://standart.aero/en/icao/book/doc-9868-training-en-cons .[31] KHATWA R, ROELEN A L C. Controlled flight into terrain (CFIT)accidents of air taxi, regional & major operators[M]. Amsterdam: CRC Press, 2020. -
点击查看大图
图(8) / 表(9)
计量
- 文章访问数: 14
- HTML全文浏览量: 5
- PDF下载量: 3
- 被引次数: 0