PCA在火箭发动机试车台传感器故障诊断中的应用

Application of PCA in Sensor Fault Diagnosis of Rocket Engine Ground Testing Bed

为了解决液体火箭发动机在试车过程中氢供应系统的关键传感器的故障诊断问题，该文利用主元分析（PCA）方法为几个重要传感器建立了主元分析模型。在所建立模型的基础上，根据平方预报误差（SPE）对传感器故障敏感的特点利用其进行传感器的故障检测。该文根据一种量化的指标参数——传感器有效度指标（SVI）对故障传感器进行辨识，实现故障分离。一般的基于传感器对SPE贡献率的辨识方法只能进行定性分析，而SVI则将辨识参数量化到0～1之间。更具实用价值。同时，证明了迭代收敛的重构解析表达式，去除了故障传感器数据对重构精度的影响，实现了数据重构。通过仿真实例验证了这种传感器故障诊断及重构方法的有效性．为发动机的正常试车提供了有力的保证。

To solve the problem of fault diagnosis for the vital sensors in the hydrogen providing system in the process of testing the liquid rocket engine, a model for these sensors is established with principal component analysis. Based upon the model established here, a square prediction error is used to detect that the sensor fault for SPE （Principle component analysis） is sensitive to the sensor fault. Sensor validation index, which is a quantity parameter, is employed to identify the faulty sensor to separate the fault. Common identifying method with contributions to SPE can just complete qualitative analysis, while the quantitative identifying parameter is between 0 and 1 with SVI, which is more valuable. Meanwhile, the analytical equation for reconstruction is proved to be convergent to eliminate the effect of the faulty sensor, so the data reconstruction is implemented. The simulation result illustrates that the effectiveness of the method for the sensor fault diagnosis and reconstruction provides powerful assurance for the security of the test.