论文部分内容阅读
主元分析方法通常采用累积贡献率(CPV)法来确定主元个数,而CPV法具有一定的主观性。本文提出一种基于阈值法来自适应实时确定主元个数的方法,有效克服传统累积贡献率法的缺点;在阈值法的基础上,提出一种对数据样本及协方差矩阵均加入遗忘因子的移动窗递推PCA(MWRPCA)过程监测方法。结合TE过程仿真,该方法能够提高监测精度,降低计算时间。将MWRPCA和RPCA、改进RPCA和MWPCA方法相比较,通过对比分析各主元分析方法TE过程的仿真结果,MWRPCA方法可以运用最少的计算时间达到最高的平均CPV值,同时有效的检测出故障,更利于在线实时监控,仿真结果验证了所提出方法的有效性。
The principal component analysis method usually uses the cumulative contribution rate (CPV) method to determine the number of principal components, while the CPV method has some subjectivity. In this paper, we propose a method based on thresholding method to adaptively determine the number of principal components in real time, which overcomes the shortcomings of the traditional cumulative contribution rate method. On the basis of the threshold method, a new method is proposed to add forgetting factors to both data samples and covariance matrixes Mobile window recursive PCA (MWRPCA) process monitoring method. Combined with TE process simulation, this method can improve the monitoring accuracy and reduce the calculation time. Compared MWRPCA with RPCA, improved RPCA and MWPCA, by comparing and analyzing the simulation results of principal component analysis (TEA) process, MWRPCA can achieve the highest average CPV with the least computation time, meanwhile effectively detect the fault and more Which is beneficial to online real-time monitoring. Simulation results verify the effectiveness of the proposed method.