基于改进K-SVD字典训练的涡扇发动机气路突变故障稀疏诊断方法

Turbofan engine abrupt gas path fault diagnosis method based on improved K-SVD dictionary training and sparse theory

以典型气路突变故障信号的稀疏特性为基础,通过对涡扇发动机部件特征原子组进行分类,提出了改进K-奇异值分解(K-singular value decomposition,K-SVD)字典训练的稀疏诊断方法,并结合气路典型突变故障开展了仿真实验研究。仿真结果表明:相比于拓展卡尔曼滤波(extended Kalman filter,EKF)和无迹卡尔曼滤波(unscented Kalman filter,UKF)方法,改进K-SVD方法对故障定位准确,无故障部件健康参数变化为0,可有效提高故障部件辨识度,避免误诊断;计算耗时与EKF方法基本相等,仅为UKF方法的0.3%,是一种有效的航空发动机气路突变故障在线诊断方法。

The characteristic atomic group of turbofan engine components was classified and exploited to the K-SVD(K-singular value decomposition) based on the sparse characteristics of gas path abrupt fault signals,then an improved K-SVD dictionary training algorithm was proposed and used for abrupt fault diagnosis. The compared results with EKF(extended Kalman filter) and UKF(unscented Kalman filter) showed that the improved K-SVD method was accurate for fault location,and the change of health parameters of no fault components was 0,which can improve the identification of fault components effectively and avoid misdiagnosis;the calculation time was basically the same as EKF method;under similar accuracy,the time consumption of this method was only 0.3% of UKF method,which can be adapted for engine gas path parameter tracking and abrupt faults diagnosis.