摘要
诺西肽发酵过程中关键生化参数难以在线测量,给控制与优化带来困难。针对这一问题,利用软测量技术来实现关键生化参数的在线估计,并提出了一种分阶段软测量建模方法。首先以分阶段的诺西肽发酵过程非结构模型为基础,根据隐函数存在定理进行辅助变量的合理选择;然后利用模糊c均值聚类算法将建模数据按其所属阶段的不同进行分类,并利用神经网络建立发酵阶段在线识别模型和对应于各个阶段的局部软测量模型。实验结果验证了所提方法的有效性。
Key biochemical parameters are hard to measure on-line in Nosiheptide fermentation process.This brings difficulties to control and optimize Nosiheptide fermentation process.To solve this problem,soft-sensor technique was used,and a staged modeling method was proposed.Firstly,based on the staged unstructured model of Nosiheptide fermentation process,secondary variables were selected according to the implicit function existence theorem.Secondly,the modeling data were classified according to their stages by using fuzzy c-means clustering algorithm.Lastly,a model for the on-line identification of fermentation stages and some local soft-sensor models were developed by using neural network.The testing results showed the effectiveness of the proposed approach.
出处
《化工学报》
EI
CAS
CSCD
北大核心
2011年第6期1612-1619,共8页
CIESC Journal
基金
中央高校基本科研业务费专项资金项目(N090302006)~~
关键词
诺西肽发酵过程
软测量
辅助变量选择
分阶段建模
Nosiheptide fermentation process
soft-sensor
selection of secondary variables
staged modeling