基于神经网络对枯草芽孢杆菌富硒过程的建模研究

Modelling Research of the Selenium Enrichment by Bacillus subtilis Base on Artificial Neural Network

本文通过神经网络和响应面法建模,研究温度、pH、氯化钠浓度和硒浓度对预测枯草芽孢杆菌的生长,碱性蛋白酶活和富硒的影响。将径向基(Radial Basis Function,RBF)神经网络和反向传播(Back Propagation,BP)神经网络这两种类型的网络,与响应面法作比较。同时,使用遗传算法优化神经网络的内部结构。通过灵敏度分析发现枯草芽孢杆菌的生物量以及碱性蛋白酶活性的变化与氯化钠浓度的变化密切相关,并且温度的变化也影响碱性蛋白酶的活性,而亚硒酸钠的浓度是影响富硒最重要的参数。通过比较实测值和预测值发现:用BP网络所预测的结果比RBF网络和响应面法更加准确。遗传算法-神经网络方法提供了一个可靠的软件工具来预测枯草芽孢杆菌的富硒过程,为实际的规模化生产奠定理论基础。

The aim of this study was to investigate the effect of temperature, p H level, sodium chloride level and selenium concentration on predicting the growth（G）, alkaline protease activity（AC） and selenium conversion rate（R）of Bacillus subtills C-3, by using Artificial Neural Network（ANN） and Response Surface Methodology（RSM）. Two types of ANN models, Radial Basis Function（RBF） and Back Propagation（BP） networks were used to compare with RSM.Meanwhile, Genetic Algorithm（GA） can be particularly important as optimization method to optimize the structure of network. The results were compared with actual values and it was found that the predicted growth, alkaline protease activity and selenium conversion rate by BP network was more accurate than RBF network and Response Surface Response Surface Methodology（RSM）. Moreover, it was found that the change of the biomass growth and alkaline protease activity was strongly correlated with the change of sodium chloride level. Temperature was also affected the activity of alkaline protease. Sodium selenite concentration was determined as the most important parameter to the conversion rate of selenium.The results of this study offered alternative method to predict the relationship between the medium component, the growth rate, enzyme activity and selenium conversion rate.