融合模型基残差分析与数据驱动的气路故障诊断方法研究

Research on Gas Path Fault Diagnosis Method Based on Model-based Residual Analysis and Data-driven

摘要:

为了提高气路故障诊断方法的可靠性,研究聚焦传感器测量噪声、个体差异和性能衰退等不确定性因素影响,导致气路故障诊断方法虚警率过高,无法实现工程应用的问题,开展了融合模型基残差分析与数据驱动的气路故障诊断方法研究。为此,构建了基于发动机模型分析偏差与卷积神经网络建模理论融合的气路故障诊断架构,在建模过程中充分考虑了传感器测量偏差、个体差异和性能衰退等不确定性因素对气路故障诊断结果的影响,据此形成了融合模型基残差分析与数据驱动的气路故障诊断方法。随后,结合多种飞行轨迹和进气条件开展数值模拟分析验证研究,对形成的气路故障诊断方法的虚警率进行了定量验证分析。结果显示,研究提出的融合模型基残差分析与数据驱动的气路故障诊断方法可在多种不确定性因素存在时,可提供满意的故障诊断精度,具有工程应用的潜力。

Abstract:

To improve the reliability of gas path fault diagnosis method, the influence of uncertainty factors such as measurement noise, individual difference and performance degradation of focus sensor is studied carefully, which leads to the problem that the false alarm rate of gas path fault diagnosis method is too high to be applied in engineering. The research of gas path fault diagnosis method based on model-based residual analysis and data-driven is carried out in this paper. Therefore, a gas path fault diagnosis framework based on the fusion of engine model analysis deviation and convolution neural network modeling theory is constructed. In the modeling process, the influence of sensor measurement bias individual difference and performance degradation on the gas path fault diagnosis results is fully considered, and the gas path fault diagnosis method combining model-based residual analysis and data-driven is formed. Then, combined with a variety of flight trajectories and air intake conditions, the numerical simulation analysis was carried out to verify the false alarm rate of the gas path fault diagnosis method. The results show that the proposed method can provide satisfactory fault diagnosis accuracy in the presence of a variety of uncertain factors and has the potential of engineering application.