一种基于自适应核主元分析的故障检测方法

An Fault Detection Method Based on Kernel Principal Component Analysis

针对用一个时不变的固定KPCA模型来监控航空发动机这类时变系统的性能时,可能会引起故障检测、诊断的偏差的问题,提出了基于自适应核主元分析的航空发动机故障检测方法。该方法利用滑动窗口的机制,通过不断加入实时采集的数据,自动更新监控模型,使KPCA监控模型能适应这种时变系统的正常参数漂移。对某型涡扇发动机进行故障检测的应用结果表明,与静态KPCA检测模型相比,自适应KPCA检测模型具有更好的故障检测效果,可提高故障检测的快速性及准确率。

In order to solve the problem that the deviation of detection and diagnosis is brought when a fixed KPCA model is used to monitor the performance of varying system(such as aeroengine),a novel fault detection method based on adaptive kernel component analysis(AKPCA)is put forward.The proposed method automatically updatas KPCA model by continuously adding online data based on moving window.The updated KPCA model can match the normal parameter excursion of varying system.The practical application results show that the performance of AKPCA is better than that of fixed KPCA,and AKPCA can enormously increase the rapidity and accuracy of fault-detecting.