基于人工神经网络的重组毕赤酵母表达期菌体浓度模型

Artificial Neural Network-based Modeling of Biomass Concentration of Recombinant Yeast Pichia Pastoris at Expression Phase

针对生物反应过程控制关键变量难以测量的问题,提出一种基于人工神经网络的重组毕赤酵母高密度发酵表达期的细胞菌量软测量模型,并对该模型的拓扑结构以及训练参数进行了初步探讨。当选取合适的模型结构和输入参数,模型的预测值最大误差为3.12%,表明该模型的计算值与菌体浓度实验值基本一致。因此,在毕赤酵母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,可以应用于发酵过程中菌体浓度的实时预测。

An artificial neural network （ANN）-based soft-sensor modeling of biomass is established at expression phase of recombinant yeast Pichia Pastoris in high-cell-density culture. The structures and training parameters of setup models axe investigated. It shows that the maximum error is only 3.12% after a proper structure and input parameters of the ANN-based model are optimized, which means the calculated values of biomass according to models are basically consistent with experimental values. Hence, the precision of the soft-sensor model based on ANN is very high, which can be applied in real-time prediction of biomass concentration during recombinant Pichia Pastoris bioprocess in high-cell-density culture.