基于支持向量机的航空发动机故障诊断

Aero-Engine Fault Diagnosis Based on Support Vector Machine

支持向量机是一种具有完备统计学习理论基础和出色学习性能的新型机器学习方法,它能够较好地克服神经网络容易出现的过学习和泛化能力低等缺陷。提出一种基于支持向量机的航空发动机故障诊断方法,应用该方法成功地对发动机气路部件的几种典型故障进行了正确诊断。在对检验样本施加噪声后,支持向量机构成的故障分类器仍然能够满足发动机故障诊断的要求,表明提出的故障诊断算法具有良好的鲁棒性,可以作为工程应用的基础。

The capabilities of Support Vector Machine (SVM) applied to aero-engine fault diagnosis (i.e. classification of multiple faults) were investigated. The gas path components of a jet engine were selected to gather datasets for the evaluation of the classification capabilities. With the proposed approach, 24 sets of component single fault testing data from the datasets were classified into 5 single faults and 16 sets of multiple faults testing data were classified into 8 faults. The single faults are low pressure compressor (LP) fault, high pressure compressor (HP) fault, low pressure turbine (LT) fault, high pressure turbine (HT) fault and no fault; while the multiple faults are LP + HP, HP + HT, HT + LT, LP + LT and LP + HP + LT + HT faults. There is no misclassification for all of the testing data using SVM. When the datasets are masked with noise as great as 12% in single faults and 10% in multiple faults, the effectiveness and robustness of the fault diagnosis algorithms are still satisfactory.