木糖醇发酵状态估算、过程预测与优化

State Estimation,Process Prediction and Optimization for Xylitol Fermentation

运用广义对数方程拟合不同初始木糖浓度（４７３～９４７１ｇ／Ｌ）下的木糖醇发酵过程，借助于均匀设计综合考察构建神经网络时第一、二隐层单元数、学习速度、初始权矩阵对模拟结果的影响及不同的初始状态变量对木糖醇发酵过程的影响，建立了一个能够很好地用于不同初糖浓度下木糖醇发酵状态估算和过程预测的四层６７６３拓扑构型的神经网络模型，提出了对改善木糖醇发酵有指导意义的优化方案。

The simulation of the process for xylitol fermentation from different xylose concentrations (4.73～94.71 g/L) was proposed with generalized logistic equation.The effects of the numbers of lst and 2nd hidden units,the learning rate and the initial weight matrix on the neural network training and the different initial state parameters on the process of xylitol fermentation were comprehensive surveied by means of uniform design.Based on above research the four layered network model with 6 7 6 3 topology was constructed,which was accurate enough for state estimation and process prediction of xylitol fermentation from different initial xylose concent rations,and the programme of paramenter optimization for improving xylitol fermentation was presented.