为了加速模型在线更新的速度以更好地适应实际工业过程的动态变化,通过在已有递推主元分析(PCA)算法的基础上简化了自相关矩阵的递推公式,从而改进了基于秩1更新的递推PCA算法,把原来需要进行2次秩1更新的步骤简化为仅仅需要进行一次秩1更新,并在此基础上提出了递推主元回归算法。仿真结果表明,改进后的基于秩1更新的递推PCA算法比原来的基于秩1更新的递推PCA算法缩短了近一半的运算时间,而新的递推主元回归算法,不但能够适应工业过程的动态变化,并且比批处理的方式节约了存储空间与计算时间。
To accelerate the model on-line modification and accommodate the industrial process change, an efficient recursive PCA algorithm using rank-one modification and a novel recursive PCR algorithm are proposed by improving the approach of updating correlation matrix. Simulation results show that the improved recursive PCA based on rank-one modification shorten the computational time in contrast with the existing recursive PCA algorithm. Moreover, the recursive PCR algorithm can adapt process changes and need less computing time and memory usage than batch PCR algorithm.