基于自适应核主元分析的EHA系统传感器故障检测

Sensor Fault Detection for EHA System Based on Adaptive Kernel Principal Component Analysis

关键词:  自适应核函数  核主元分析  传感器  故障检测  EHA系统

Key words:  Adaptive Kernel Function  Kernel Principal Component Analysis  Sensor  Fault Detection  EHA System

摘要:

针对核主元分析(KPCA)方法难以选择合适核函数的问题，提出了一种基于自适应核主元分析的传感器故障检测方法，根据训练数据对核函数进行自适应修正，使核函数适应给定的训练数据。对常规数据标准化处理方法进行了改进，提出了一种“均值化”的处理方法，使处理后的数据既能消除不同变量幅值和量纲的影响，又能反映训练数据的全部信息。将此方法应用于机载电动静液作动器(Electro-Hydrostatic Actuator，EHA)系统的传感器故障检测，结果表明，此方法比常规KPCA方法更为先进，具有更好的故障检测性能。

Abstract:

For the problem that it is hard to choose the kernel function of kernel principal component analysis ( KPCA )， a novel adaptive KPCA method for sensor fault detection is presented. The kernel function is modified adaptively according to the training date ， so the kernel function can adapt to the given training date. Besides ， the common method for data standardization processing is modified ， an‘equalization’processing method is presented ， which can eliminate the influence caused by amplitudes and dimensions of different variables ， and the entire information of the training data can be reflected at the same time. Finally ， an application of the proposed method is given in the sensor fault detection for airborne Electro-Hydrostatic Actuator ( EHA ) system ， and the simulation results show that the proposed method is more advanced than the general KPCA ， and it has much nicer performance in fault detection.