摘要
建立了用于在线估计高密度重组毕赤酵母培养过程中处于表达阶段的菌体密度软测量模型。分别对比了基于遗传算法(GA)的动力学软测量模型以及基于人工神经网络(ANN)的软测量模型,并对神经网络软测量模型的拓扑结构以及训练参数进行了初步探讨。当采用基于遗传算法(GA)的动力学模型,模型拟合值的最大误差为7.63%;在采用神经网络软测量技术时,选取合适的模型结构和输入参数,最大误差为3.12%,而且软测量模型可以很好地反映菌体浓度实时变化趋势。该研究结果表明,在酵母细胞的高密度培养过程中采用基于神经网络的软测量模型具有较高的准确度,可以较好地实时反映发酵过程中菌体浓度的变化。
A soft-sensor model based on kinetics and neural network was established for online estimation of cell concentration in high-cell density cultivation of recombinant Pichia Pastoris. Performance comparison of the genetic algorithms （GA） and the artificial neural network （ANN） based soft-sensor kinetic models were also conducted, and topological structures and training parameters of the neural network model were computed. The results showed that the maximum error was 7.63% for the GA based model and 3.12% for the ANN model under the condition of the optimized structure and input parameters. Therefore, the ANN model could predict the real time biomass concentration well with a higher accuracy in P. pastoris high-cell-density culture.
出处
《食品与生物技术学报》
CAS
CSCD
北大核心
2006年第5期28-34,共7页
Journal of Food Science and Biotechnology
基金
国家"十五"高新技术发展计划(863计划)项目(2002aa217021)
国家重大科技专项项目(2002aa2z3451)
关键词
巴氏毕赤酵母
高密度培养
在线估计
遗传算法
神经网络
软测量
Pichia Pastoris
high-cell-density cultivation
online estimation
genetic algorithms
artificial neural network
software sensing
