基于多“内在传感器”逆的诺西肽发酵过程生化参数软测量模型  

Soft-sensing model for biochemical parameters in Nosiheptide fermentation process based on multiple "inherent sensor" inversion

作　　者：杨强大[1] 王福利[1] 常玉清[1] 

机构地区：[1]东北大学流程工业综合自动化教育部重点实验室,沈阳110004

出　　处：《仪器仪表学报》2007年第12期2163-2168,共6页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60374003);973子课题(2002CB312200)资助项目

摘　　要：针对诺西肽发酵过程中关键生化参数难以在线测量的问题,提出了一种基于多"内在传感器"逆的软测量模型。在诺西肽发酵过程非结构模型的基础上,建立了多个包含在原系统中的"内在传感器"子系统。经过数学推导证明了各子系统的可逆性,并利用神经网络分别拟合各子系统的逆,实现了诺西肽发酵过程中菌体浓度和基质浓度的软测量。实际应用表明,该软测量模型能够较好地预估菌体浓度和基质浓度,其平均相对误差都在5%以内,且所提软测量建模方法是有效的。A soft-sensing model based on multiple ＂inherent sensor＂ inversion is proposed to solve the difficulties of crucial biochemical parameters＇ on-line measurement in Nosiheptide fermentation process. Based on the unstructured model of Nosiheptide fermentation process, some ＂inherent sensor＂ subsystems embodied in the primary system are built, and the invertibility of each subsystem is testified by mathematical inference. Then the estimation of biomass concentration and substrate concentration in Nosiheptide fermentation process is achieved by a soft-sensing model, which is developed using multiple neural networks to fit the inversion of each subsystem. Practical application shows that the soft-sensing model can well predict biomass concentration and substrate concentration with average relative errors within 5% , and the proposed approach is effective for the development of soft-sensing model.

关 键 词：逆系统 多模型 软测量 神经网络 发酵 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]