基于LS-SVM逆模型的青霉素发酵软测量方法

针对青霉素发酵过程中的某些关键生物参数(如菌体浓度、基质浓度、产物浓度)难以实时在线测量的问题,提出一种基于最小二乘支持向量机(LS-SVM)逆模型的软测量方法。该方法用具有高斯核函数的LS-SVM,离线建立被测对象的静态非线性逆模型。由静态非线性逆模型外加若干表征非线性动态特征微分器,构成了非线性系统的逆系统,将此逆系统串联在原发酵系统之后,得到"线性化"的伪线性系统。仿真结果表明,该方法能够对青霉素发酵过程中不可在线测量的关键变量进行预测,且达到了较高的测量精度。

Soft-sensor of penicillin fermentation based on LS-SVM inverse system

A soft sensor based on the least-square support vector machine inverse system is proposed to measure the important parameters of penicillin fermentation process, such as matrix concentration, biomass concentration and production concentration. This method is that the inverse system combines with the least-squares support vector machine. The mathematical model of fermentation process is given first, and then the reversibility of system is testified.The state nonlinear inverse model of the plant is off-line built by LS-SVM with a Gaussian kernel. The state nonlinear inverse model and a number of differentiators constitute the non-linear system of inverse system .Based on the theory of inverse system method ,the least-square support vector machine inverse model is cascaded after the fermentation system to be linear pseudo-linear system. The simulation example shows that the method could measure the important parameters which could not be measured online during the course of penicillin fermentation with a high pre- cision.