基于BP神经网络的洋葱伯克霍尔德菌脂肪酶发酵软测量建模  

Soft Sensor Modeling of Burkholderia cepacia Lipase Fermentation Process Based on BP Neural Network

作　　者：汪小锋[1] 付强[1] 申旭光[1] 杨江科[1] 闫云君[1] 

机构地区：[1]华中科技大学生命科学技术学院,教育部分子生物物理重点实验室,湖北武汉430074

出　　处：《食品与生物技术学报》2010年第6期921-926,共6页Journal of Food Science and Biotechnology

基　　金：国家863计划重点项目(2006AA020203,2007AA100703,2009AA03Z232);国家863计划专题项目(2007AA05Z417);教育部新世纪优秀人才支持计划项目(NCET-07-0336);武汉市科技攻关计划重点项目(200720422138)

摘　　要：为建立洋葱伯克霍尔德菌(Burkholderia cepacia)脂肪酶发酵过程的软测量模型,运用BP神经网络对洋葱伯克霍尔德菌脂肪酶的发酵过程进行软测量建模,并利用遗传算法对神经网络的初始权值和阈值进行优化,实现模型加快收敛速度,达到全局最优解效果。该模型能够比较精确地模拟菌体生长、底物消耗以及发酵产酶的过程动态,具有良好的泛化能力,说明BP神经网络结合遗传算法在洋葱伯克霍尔德菌脂肪酶发酵过程的模拟与预测中是一种高效快速的方法。The aim of this study was to establish a soft sensor model of the Burkholderia cepacia lipase fermentation process.Procedure and method：for this,the backpropagation neural network was applied to rebuild a soft sensor model for the fermentation process of Burkholderia cepacia lipase,and genetic algorithms was further employed to optimize the initial weights and thresholds of neural networks,so as to speed up the convergence rate and achieve a comprehensive optimum result.Result：The model could imitate precisely the process of bacteria growth,culture medium consumption and lipase production,which suggests that the model possesses good reliability and generalization ability.The backpropagation neural network combined with genetic algorithms is an effective and quick method for simulating and predicating fermentation process of B.cepacia lipase.

关 键 词：洋葱伯克霍尔德菌脂肪酶 BP神经网络 遗传算法 发酵 模型 

分 类 号：TQ925.6[轻工技术与工程—发酵工程]