青霉素发酵过程中基于PLS-VIP的神经网络逆软测量方法  

Neural network inverse soft-sensing method based on PLS-VIP in penicillin fermentation process

作　　者：刘国海[1] 程锦翔[1] 丁煜函[1] 梅从立[1] 

机构地区：[1]f江苏大学电气信息工程学院,江苏镇江212013

出　　处：《计算机与应用化学》2013年第11期1294-1298,共5页Computers and Applied Chemistry

基　　金：国家中小型创新基金项目(12C26213202207);江苏省校优势学科建设工程资助项目([2011]6)资助;中国博士后科学基金(20110491359)

摘　　要：针对复杂发酵过程逆软测量模型难以建立和估计精度不高的问题,提出一种基于偏最小二乘和变量投影重要性指标的PLS-vIP神经网络逆软测量方法。该方法在"内含传感器"逆软测量模型的基础上,通过计算变量投影重要性指标,确定逆模型中各辅助变量对关键生化量的贡献率,建立了辅助变量少、估计精度高的逆软测量模型。与传统逆软测量方法相比,克服了依赖于数学模型和复杂算法推导的问题。将其应用到青霉素发酵过程中生化量估计中,利用Pensim仿真平台进行实验,仿真实验结果表明,该方法的逆模型比传统逆模型构建简单、估计精度高。To overcome the difficulties in establishing an inverse soft-sensing rnodel and its lack of accuracy in the complex fermentation process, an inverse soft-sensing method based on partial least squares and variable importance in projection methods called PLS-VIP is proposed. Based on the ＂assumed inherent sensor＂ inverse soft-sensing model, an inverse soft-sensing model with fewer variables and high accuracy is obtained through calculating VIP values. Compared with traditional inverse soft-sensing method, it is independent of mathematical model and complex derivation. It is applied to a soft sensor modeling for penicillin fermentation process using Pensim simulation platform. Simulation results show that the proposed method has higher precision and better performance than the original inverse soft-sensor method.

关 键 词：神经网络逆 软测量 偏最小二乘 变量投影重要性 青霉素发酵过程 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]