基于非参数回归的发酵过程软测量建模  

A non-parametric soft sensor modeling method for use in fermentation processes

作　　者：高宇[1] 王建林[1] 于涛[1] 赵利强[1] GAO Yu WANG JianLin YU Tao ZHAO LiQiang(College of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029 ,China)

机构地区：[1]北京化工大学信息科学与技术学院,北京100029

出　　处：《北京化工大学学报（自然科学版）》2016年第5期74-78,共5页Journal of Beijing University of Chemical Technology(Natural Science Edition)

基　　金：国家自然科学基金(61240047);北京市自然科学基金(4152041)

摘　　要：为了对发酵过程进行检测并分析发酵中产物的浓度,提出一种基于非参数回归的发酵过程软测量方法。该方法直接利用未经处理过的历史数据,采用多维窗宽选择方法确定窗宽,通过计算待估计点与历史数据库中其他时刻点数据的距离来确定该时刻点的权值,从而达到软测量建模的目的,并将其用于青霉素发酵过程产物浓度的测量。实验研究表明,所得软测量模型精度较高,能够有效地实现对青霉素发酵产物浓度的估计。A soft sensor for use in fermentation processes which allows the process to be followed and product concentration to be analyzed has been simulated.The soft sensor is based on nonparametric regression which leverages untreated historical data directly,uses a multidimensional window width selection method to confirm window width,calculates the distance between the estimation point and other moment points in historical data to confirm the weight of the point at the given moment,in order to simulate the soft sensor and measure the product concentration during a penicillin fermentation process.The results showed that the simulation model is highly precise and it succeeds in estimating the product concentration from penicillin fermentation,which proves the accuracy and effectiveness of the method.

关 键 词：发酵过程 非参数回归 核函数 多维窗宽选择 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]