返回
An interpretable machine learning framework for identifying operational risk boundaries in runway excursion using flight data
DOI:10.1016/j.ress.2026.113119.png)
摘要
En 中文
• 来自2362架次飞行记录器数据支持风险边界分析。
• 可解释机器学习将着陆状态与风险前兆联系起来。
• 弱垂直闭合度将长着陆概率从0.011提高到0.144。
• 侧风变异性将滑跑不稳定性从0.008提高到0.337。
• 六个有序风险等级支持可解释的飞行安全监控。
Keyword:
Flight safety
Quick access recorder (QAR)
Interpretable machine learning
Runway excursion risk
期刊
R
IF:
11
论文数:
1.1K
被引数:
0
机构
引用论文
Modeling for evaluation of safety instrumented systems with heterogeneous components用于评估具有异构组件的安全仪表系统的建模
A Runway Overrun Risk Assessment Model for Civil Aircraft Based on Quick Access Recorder Data基于快速存取记录器数据的民用飞机跑道超限风险评估模型
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead停止解释高风险决策的黑盒机器学习模型,而改用可解释的模型
A novel hybrid STL-transformer-ARIMA architecture for aviation failure events prediction一种用于航空故障事件预测的新型stl-transform-arima混合架构
Safety barriers: Research advances and new thoughts on theory, engineering and management安全屏障: 理论、工程和管理的研究进展与新思路

