基于ReliefF-PNN的发动机气路系统故障诊断

Fault diagnosis of aero-engine air circuit system based on reliefF-PNN

航空发动机气路系统特征参数具有多维度、高冗余和强相关性等特点。为了降低气路系统特征参数的维度、提高诊断精度,提出了一种基于ReliefF-PNN的航空发动机气路系统故障诊断方法。利用ReliefF算法对气路系统的参数作特征评价,确定具备表征气路系统的最优特征参数子集;对得到的最优特征参数子集数据样本训练生成基于PNN的气路系统故障诊断模型;将模型应用于气路系统进行诊断验证,并做结果分析。实验表明,使用ReliefF选择出的最优特征参数进行故障诊断,诊断精度高达100%,结果优于使用所有特征进行诊断。

Aero-engine air circuit system are characterized by multiple dimensions,high redundancy and strong correlation. In order to reduce the dimension and im prove the diagnostic accuracy of the air circuit system,a fault diagnosis method was presented in this paper based on Relief F-PNN. The Relief F algorithm was used to evaluate the characteristic parameters and determine the optimal parameter subsets which can characterize the air circuit system. The subsets as training samples were further used to establish the fault diagnosis model of the system. The model was verified by the application of airway system and then the results were analyzed. Experimental results show that the optimal feature parameters selected by ReliefF are used to diagnose fault,its diagnostic accuracy reaches to 100%,which is better than that using all features.