| 摘要: |
| 针对民机故障诊断中真实故障数据稀缺问题，建立了横侧向飞控系统的高保真仿真模型，分析典型故障模式并构建故障仿真模型，模拟真实飞行故障情况，为智能诊断提供数据支撑。为解决退化故障数据不足的问题，提出了基于蒙特卡罗仿真的失效数据集生成流程，结合随机采样、概率建模和仿真验证技术，在不同运行条件和故障程度下生成多样化故障数据集，增强数据覆盖度、代表性和适应性。该方法不仅可生成已知故障模式下的丰富数据样本，还能通过随机扰动和参数优化，提高对未知故障模式的外推能力，提升诊断系统学习能力。通过突发型故障注入实验验证了数据集有效性，结果表明该方法能准确再现不同故障模式下的动态特征，确保数据的物理合理性和诊断价值。 |
| 关键词: 飞控系统 智能故障诊断 退化模型 故障注入 失效数据集生成 |
| DOI：10.20079/j.issn.1001-893x.241126002 |
|
| 基金项目:国家自然科学基金资助项目（61773262，62006152）；中国航空科学基金（2022Z071057002）；上海市自然科学基金（24ZR1434400） |
|
| Generation of Flight Control System Failure Datasets Based on Degradation Model |
| ZHAO Guang,HU Shiqiang,GUO Qiang,LUO Lingkun |
| (1.School of Aeronautics and Astronautics，Shanghai Jiao Tong University，200000，China; 2.Commercial Aircraft Corporation of China,Ltd.，Shanghai 200126，China) |
| Abstract: |
| For the issue of scarce real fault data in civil aircraft fault diagnosis，a high-fidelity simulation model of the lateral-directional flight control system is established.It analyzes typical fault modes and constructs a fault simulation model to replicate real flight fault scenarios，providing data support for intelligent diagnosis.To tackle the insufficiency of degradation fault data，a failure dataset generation process based on Monte Carlo simulation is proposed.This process integrates random sampling，probabilistic modeling，and simulation verification techniques to generate diverse fault datasets under various operating conditions and fault severity levels，thereby enhancing data coverage，representativeness，and adaptability.The method not only generates abundant data samples for known fault modes but also improves the extrapolation capability for unknown fault modes through random perturbation and parameter optimization，thereby enhancing the learning ability of the diagnostic system.The effectiveness of the generated dataset is validated through sudden fault injection experiments.The results show that the method can accurately reproduce the dynamic characteristics of different fault modes，ensuring the physical rationality and diagnostic value of the data. |
| Key words: flight control system intelligent fault diagnosis degradation model fault injection failure dataset generation |
《电讯技术》编辑部 