摘要
通过对萃取固定床层进行质量衡算建立了超临界CO2流体萃取动力学模型.然后采用BP神经网络拟合实验条件下的萃取动力学曲线,并将之与动力学模型试差法求解所得的萃出曲线进行对照确定模型参数kLa.结果表明,BP神经网络能很好地模拟大黄蒽醌的萃取动力学曲线,网络训练误差和预测误差分别为1.5%和3.5%.确定参数后的动力学模型可用于对萃取床层作较为精确的定量描述:模拟所得萃取穿透曲线与大黄蒽醌萃取实验结果相比AARD误差在10%左右.与前人模型相比,该模型具有精度好,获取参数时的实验条件相对宽松等优点.
A kinetic model for supercritical carbon dioxide calculation of the extraction fixed bed. To determine the extraction was developed by mass conservation model parameter k1.a, BP neural network was used to fit the experimental extraction elution curve, which was then compared with the elution curve calculated by the kinetic model with suggested parameter values. The parameter value was eventually determined through trial and error in the fitting process. The results showed that BP neural network could successfully fit the experimental data of the Rhubarb anthraquinone extraction kinetics curves, with 1.5% of training error and 3.5% of prediction error. The kinetic model with determined parameter can be used to describe the extraction bed with higher accuracy. The simulated extraction elution curve through the model had about 10% AARD error compared with the anthraquinone extraction experimental data. This model has the advantage over the others not only for better precision it offers, but also for the fact that it only requires moderate experimental conditions.
出处
《化工学报》
EI
CAS
CSCD
北大核心
2006年第3期582-589,共8页
CIESC Journal
关键词
超临界二氧化碳流体
大黄蒽醌
萃取动力学模型
神经网络
supercritical CO2 fluid
Rhubarb anthraquinone
extraction kinetics model
neural network model
