海洋蛋白酶发酵过程生物参数的软测量建模

Soft Sensor Modeling Based on Biological Variables of Marine Protease Fermentation Process

针对海洋微生物低温碱性蛋白酶发酵过程关键生物参数(如菌体浓度、基质浓度、酶活等)的在线测量问题,提出了一种基于径向基函数(RBF)神经网络的软测量建模方法.首先,应用一致关联度法确定软测量模型的辅助变量.然后,针对正交最小二乘、k均值聚类等算法在具体实现时比较繁琐的问题,利用简单且实用的RBF网络设计函数,应用训练样本集对RBF神经网络进行训练,建立了RBF神经网络软测量模型,用测试样本集对建立的软测量模型进行了仿真验证.结果表明,该软测量模型具有很高的逼近精度和良好的实用性.

In order to solve the problem of real-time measurement of crucial biological variables （such as biomass concen- tration, substrate concentration, enzyme activity, and so on） in low-temperature alkaline protease fermentation process of the marine microbe, a soft sensor modeling method based on radial basis function （RBF） neural network is proposed. Firstly, the auxiliary variables of the soft sensor model are determined by means of the uniform incidence degree algorithm. Secondly, for the orthogonal least squares and k-means clustering algorithm cannot be achieved simply, a simple and practical RBF network design function is applied. Training samples are used to train the RBF neural network, and a soft sensor model is established by RBF neural network. The test samples are used to verify the precision of the established soft sensor model. Simulation results show that soft sensor modeling based on RBF neural network has high accuracy and good practicality.