主元个数是PCA模型的关键参数,其选取直接决定PCA的故障诊断性能；针对传统主元个数选取方法主观性较大,且不考虑故障诊断要求的缺点,提出一种改进的主元个数确定方法；该方法将传统的累积方差贡献率与故障检测率相结合,首先利用累积方差贡献率初步确定主元个数,然后确定满足故障检测率要求的主元个数,将两个主元个数进行比较,从而获得最佳主元个数；与单纯累积方差贡献率方法相比,提高了主元模型的精度,减少了以往方法中人为因素的影响；通过对卫星控制系统的故障检测,证实了该方法可大大提高故障检测准确率.%The number of principal components is the essential parameter of PCA and ultimately determines the performance of this use-ful statistical method. Traditional selection methods are very subjective without fault diagnosis demands. A new determination of principal component in PCA model by using traditional cumulative percent variance (CPV) and fault detection accuracy jointly is proposed. Firstly, the number of principal components is determined by cumulative percent variance. Then, the number of principal components is determined by fault detection accuracy again, the two numbers of principal components are compared before obtain the prime principal component. Not only improving precision of principal component model but also avoiding other subjective factors compared with using only cumulative percent vari-ance. This method is applied to satellite control system fault detection. The simulation results show that the new method enhances the fault detection accuracy.
展开▼
