论文部分内容阅读
提出1种遗失数据重构思想下的软测量方法:先采用主元分析(PCA)离线建立所有变量(包括难测变量)的主元模型,实际应用时,将实时的难测变量看作遗失数据,通过遗失数据重构方法估计出难测变量,增加了软测量方法的灵活性。更进一步,在重构遗失数据时,使用马氏距离取代欧几里德距离作为指标,更准确地反映了过程变量之间的相关关系,由此指标求取软测量值能够大大地改善估计精度。
This paper proposes a soft sensing method under the idea of reconstructing lost data. First, the principal component model of all variables (including unpredictable variables) is established offline by principal component analysis (PCA). In practice, the real-time unpredictable variables are regarded as missing Data to estimate the unpredictable variables through the method of missing data reconstruction, which increases the flexibility of the soft-sensing method. Furthermore, using Mahalanobis distance instead of Euclidean distance as an indicator to reconstruct lost data more accurately reflects the correlation between process variables, and thus obtaining the soft measurement value by the index can greatly improve the estimation accuracy .