摘要
提出了一种利用支持向量机(SVM)作为弱基分类器、Boosting算法进行加权融合的航空发动机故障诊断算法.该算法具有支持向量机的强分类能力,又具有Boosting算法适合不均衡数据集的特点.为验证算法的有效性,采用外场实测的滑油光谱分析数据针对传动系统的轴承、减速齿轮和滑油系统3类故障进行了验证.为去除实测数据之间的冗余、降低特征维数,提高算法执行效率,采用主元分析(PCA)和粗糙集理论(RST)进行故障特征压缩和提取.利用实测数据构造了Boosting支持向量机分类器.最后,实验结果表明Boosting-SVM算法可以显著提高SVM分类器的推广性能.针对实测数据,3种故障平均识别准确率由79.4%提高到了85.7%.
A fault diagnosis algorithm based on Boosting and support vector machines for aero-engine was proposed.The algorithm adopts support vector machine（SVM） with linear kernel as weak learners and use Boosting algorithm to generate a set of weak learners and then fuse all the learners via weighting.The algorithm both possesses the virtue of SVM and Boosting,which separately shows strong classification ability and superiority for imbalanced datasets.To validate the efficiency of the new algorithm,the typical bearing faults,the main deduction gear faults,and the oil system faults were investigated via real inspection data.In order to reduce the redundancy of the features,principal component analysis（PCA） and rough set theory（RST） were used to determine the richest features.Then,a Boosting-SVM classifier was build using the training data.Results towards the testing data show that the proposed Boosting-SVM algorithm gives a better average success rate than standard SVM,which is significantly improved from 79.4% to 85.7%.
出处
《航空动力学报》
EI
CAS
CSCD
北大核心
2010年第11期2584-2588,共5页
Journal of Aerospace Power
