基于模式识别的航空发动机燃油控制系统传感器故障诊断

Fault Diagnosis of Fuel Control System Sensor for Aeroengine Based on Pattern Recognition

航空发动机燃油控制系统执行机构故障有可能导致参数测量传感器出现较大偏差,而采用传统的传感器故障诊断方法易误诊为传感器故障。为此,引入修正因子作为传感器故障模式样本,通过聚类分析获得样本的特征向量;按照卡洛南—路伊变换(K—L变换)原理,对传感测量信息进行变换,构成了新的正交变换矩阵,减弱了各特征向量的相关性,突出了差异性,加强了对故障传感器和发动机燃油控制系统执行机构故障的特征识别能力;利用多组学习训练样本,设计了发动机不同参数测量传感器故障模式的判别函数。经仿真试验验证,该方法可以有效识别、诊断传感器故障。

The actuator fault of fuel control system for aeroengine could cause greater deviation of the parameter measurement sensor, but sensor fault was easily misdiagnosed by adopting the traditional sensor fault diagnosis method. Therefore, the correction factor was introduced as the sensor fault pattern sample, and the sample feature vector was obtained through the clustering analysis. Based on the Karhunen -Loeve transform （K- L transform） theory, the sensor measurement information was transformed to form the new orthogonal transfer matrix, weaken the correlation of the dif- ferent feature vectors, make the variability outstanding , and strengthen the features recognition ability of the fault sensor and engine fuel control system actuator fault. The discriminant functions of different parameter measurement sensor fault patterns were designed for the aeroengine by using multi - group studying and training samples. The method can effectively recognize and diagnose the sensor fault through the simulation experimental validation.