基于DRFNN的赖氨酸发酵生物参数估计  

Biological Parameter Estimation of Lysine Fermentation Based on DRFNN

作　　者：张瑶[1] 

机构地区：[1]南京工程学院先进制造技术工程中心,江苏南京211167

出　　处：《南京工程学院学报（自然科学版）》2012年第4期37-42,共6页Journal of Nanjing Institute of Technology(Natural Science Edition)

摘　　要：赖氨酸发酵过程是一个复杂的非线性强耦合动态过程,发酵过程中某些关键生物参数(如菌体浓度、基质浓度、产物浓度)难以实时在线检测.针对这些问题,建立了基于动态递归模糊神经网络(DRFNN)的赖氨酸发酵生物参数估计软测量模型.利用主元分析法建立发酵过程中在线不可测参数的软测量模型,应用动态递归模糊神经网络所具有的自学习、自适应能力以及对任意非线性的逼近能力,对该模型进行辨识,从而达到预测这些关键生物参数的目的.仿真结果表明该方法能够对赖氨酸发酵过程中不可在线测量的生物参数进行估计,且达到了较高的精度.Real-time online detection of some key biological parameters, such as biomass concentration, substrate concentration, and production concentration in lysine fermentation process is difficult because this process is complex, nonlinear, stnmg coupling and dynamic. Tu solve this problem, a soft sensor model used to detect such key biological parameters, based on dynamic recursive fuzzy neural network is constructed. Firstly, a principal component analysis is carried out to build a soft sensor model for online indeterminable parameters in this process. Then, the n^del is identified by stir- learning, self-adaptive capaeity and arbitrary nonlinear approximation ＇ability of dynamic rcursive fuzzy neural （DRFNN） as to predict the key biological parameters. The simulation results show that this method can accurately estimate the important parameters in the course eft lysine fermentation which cannot be measured online in the course of lysine fermentation.

关 键 词：赖氨酸 动态递归模糊神经网络 主元分析法 辨识 生物参数 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]