"CAFUC2"

"This dataset is dedicated to anomaly detection in flight training scenarios, constructed based on 120 original normal flight training CSV files with a total of 993,655 valid normal data records. To support the training, validation, and performance evaluation of anomaly detection models, four types of typical flight anomalies\u2014throttle surge, course deviation, engine failure, and excessive pitch\u2014have been injected into the normal data. The anomaly injection ratio is controlled within the range of 2% to 5%, which aligns with the frequency of common abnormal events in actual flight training.The dataset retains core flight training feature fields with a unified CSV format, free from missing values or invalid records. It can be directly applied to supervised, semi-supervised, or unsupervised anomaly detection algorithms such as Isolation Forest, Autoencoder, and LSTM. Additionally, this dataset provides high-quality, real-scenario data support for the development, optimization, and validation of flight training anomaly identification systems, facilitating the improvement of flight safety monitoring capabilities in aviation training."

本数据集面向飞行训练场景下的异常检测任务，基于120份原始正常飞行训练逗号分隔值（Comma-Separated Values，CSV）文件构建，总计包含993655条有效正常数据记录。为支撑异常检测模型的训练、验证与性能评估工作，研究人员向正常数据中注入了四类典型飞行异常：油门喘振、航向偏离、发动机故障及俯仰角超限。异常数据注入比例控制在2%至5%区间内，贴合实际飞行训练中常见异常事件的发生频率。本数据集保留了飞行训练的核心特征字段，采用统一的CSV格式，无缺失值与无效记录，可直接应用于有监督、半监督或无监督类异常检测算法，例如孤立森林（Isolation Forest）、自编码器（Autoencoder）及长短期记忆网络（LSTM）。此外，本数据集可为飞行训练异常识别系统的开发、优化与验证提供高质量的真实场景数据支撑，助力提升航空训练中的飞行安全监控能力。