【作者】 陈兰星； 王晶；

【Author】 CHEN Lan-xing WANG Jing Automation Research Institute,Beijing University of Chemical Technology,Beijing 100029,China

【机构】 北京化工大学自动化研究所；

【摘要】 针对聚酯工业中采用软测量对难以或无法测量的质量指标变量进行估算的问题,以聚酯生产过程中的粘度为例,采用主元分析方法结合SPE和T~2控制限对现场测量数据进行了过失误差剔除等处理,利用RBF网络建立软测量模型,并用VC++编写了相应的软件包。鉴于系统运行的实时性,使用PCA方法实现了输入数据的实时监测,并对故障点进行相应处理,以保证模型估算结果的准确度;实现了在线的PCA主元修正和模型的网络输出以及网络参数的实时校正。更多还原

【Abstract】 To solve the problem of decting the soft-sensor theory to compute the variable of quality criterion in the industry of PET,which is illustrated by the viscosity in the PET industry process,the technique of principal component analysis combining SPE and/T~2 control limits is used to detect gross error.The PET soft-sensor based on RBF neural network is introduced,and by using the VC + + language,a soft-seining software package is developed.Considering the real time feature in the process of the system′s operation,the on-line inspection of input data is realized by the PCA theory.The bad data are processed properly to make sure that the calculating result of the model is reliable; the on-line correction of the principal components and the network output including the network parameters of the model is implemented.更多还原