部分反馈神经网络对腊样芽孢杆菌DM423分批培养过程中生物量的软测量

神神经网络以其较强的非线性处理能力在生物化工中有着广泛的应用前景，本文应用部分反馈神经网络对分批培养过程中的腊样芽孢杆菌DM423的生物量进行软测量，构建了拓扑结构为11-5-1的部分反馈神经网络。网络的输入量为pH、温度、溶氧量和葡萄糖浓度的延时量，并也将网络输出的生物量浓度进行延时、反馈作为网络输入量，生物量浓度当时值为输出量，网络的泛化能力较好，测试样本的均方差为0.5610-3。此外，所建立的部分反馈神经网络具有良好鲁棒性和预测能力。

Partial Recurrent Neural Network-based Soft-sensor for Baeillus cereus DM423 Biomass during Reactor Batch Cultivation

Abstract: Neural network is promising to deal with nonlinear problems in biochemical industry. In this paper, the biomass of Baeillus cereus DM423 during batch cultivation was measured by the soft-sensor of partial recurrent neural network. A partial recurrent neural network was constructed with the topology of 11-5-1, the 11 input variables being delays of pH, temperature, dissolved oxygen, glucose concentration at two previous times and delays of estimated biomass concentration at three previous times, and the output variable being biomass concentration at present time. The constructed network had good generalization with the mean square error of 0.5610-3 for the testing samples. Moreover, the network was evaluated with good robustness and prediction ability.