基于多神经网络的发酵过程菌丝浓度估计

某些发酵过程的参数,如菌丝浓度等难于在线测量,采用软测量的方法来进行估计是一种行之有效的方法。由于发酵过程的复杂性,传统的软测量方法难以获得准确的结果。该文采用多神经网络模型方法,充分利用尽可能得到的可在线测量信息,可有效地提高模型的估计精度和鲁棒性。该文采用的方法较传统的神经网络模型能更好地融合对被估计参数有用的冗余信息,从而达到更好的建模效果。应用实际数据的估计结果表明该软测量方法的优越性。

Combining Multiple Neural Networks to Estimate Biomass in Fermentation

During fermenting process,it is difficult to measure some variables,such as biomass.A usual way is to esti-mate them by software sensor.But to get accurate measurement by using the general modeling approaches is not easy since the natural complexity of the fermenting process.A new software sensor model based on multiple neural networks(MNN)is presented in this paper.In order to get more efficient estimation accuracy and more robust performance,the measurable information on line can be fully adopted in this model.The fusion with all adopted information and available redundant information in this approach is much better than that in the conventional software sensors method.The estima-tion results by using data from real plant illustrates the merits of this new model.