作　　者： ; ; ; ;

机构地区： 华南理工大学食品与生物工程学院

出　　处： 《华南理工大学学报（自然科学版）》 2004年第12期11-15,共5页

摘　　要： 利用BP神经网络来描述腊样芽孢杆菌 (BaeilluscereusDM4 2 3) (简称DM 4 2 3)摇瓶培养过程中生物量与培养液初始葡萄糖含量和初始生物量之间的关系 .建立了结构为2 6 3的BP神经网络 .用所建立的神经网络对DM 4 2 3摇瓶培养过程的生物量进行估算 ,发现估算值与实测值相当接近 ,表明网络有较好的泛化能力 .此外 ,当初始葡萄糖含量超出测试样本范围时 (分别为 4 .2 5和 2 7.30g/L) ,网络对样本的预测结果也相当理想 ,预测的平均相对误差为 1.95 % . The relationship between the initial conditions (initial glucose content and ini tial biomass) and the biomass during the flask cultivation of Baeillus cereus DM423 was described by BP neural network. A 2-6-3 topo- logy of the BP neural network was then constructed. The estimation results of biomass by the BP neural network show that there is little difference between the calculated valu es and the experimental ones, which means that the constructed neural network ha s a good generalization. The proposed network also promises rather good predicti on results when the initial glucose contents are 4.25 and 27.30 g/L, and the ave rage relative error of the estimation is 1.95%.

关 键 词： 神经网络 摇瓶培养 生物量 估算 泛化能力

领　　域： [轻工技术与工程] [轻工技术与工程] [自动化与计算机技术] [自动化与计算机技术]