发酵过程状态预估的BP神经网络模型实现

The Realization of Fermentation Process Status Pre-estimating Model Based on BP NN

论述了利用改进的BP神经网络实现发酵过程状态预估模型的设计原理和方法,包括BP神经网络的拓扑结构选取、学习和测试样本的选择及处理,变步长引入动量项BP神经网络的训练方法以及全局收敛法的实现等。此外,用VC实现了发酵过程BP神经网络建模平台。经聚赖氨酸发酵过程验证,其模型具有良好的收敛性能和泛化性能,可应用于发酵过程状态参数的在线预估和测量。

It was discussed that designing principle and method implemented fermentation process status preestimating model based on BP NN, including the choice of network architecture, the processing of learning and testing sample, the training of BP NN based on variable learning rate and momentum, the realization of global convergence method. Moreover, the software platform of fermentation process status preestimating model was implemented by VC. Through lysine fermentation process testing, The model possessed advanced convergence and generalization performance, being applied in fermentation process status parameter online preestimated and measured.