节点文献
死端微滤酵母悬浮液比阻的预测研究
Study of predicting the specific resistance of yeast suspensions in the dead-end microfiltration
【作者】 姚金苗; 王湛; 陈德明; 储金树; 孙光民; 张虎; 李兆辉;
【Author】 YAO Jinmiao~1,WANG Zhan~1,CHENG Deming~2,CHU Jinshu~1, SUN Guangmin~2,ZHANG Hu~3,LI Zhaohui~3 (1.Department of Chemistry and Chemical Engineering,College of Environmental and Energy Engineering, 2.School of Electronics Information and Control Engineering,Beijing University of Technology,Beijing 100022,China; 3.Beijing fluid filtration & separation technology research center,Beijing 101312,China)
【机构】 北京工业大学环境与能源工程学院化学化工系; 北京工业大学电子信息与控制工程学院; 北京流体过滤与分离技术研究中心;
【摘要】 首先设计了三因素四水平的正交实验表作为建模样本,其次利用人工神经网络方法和多元线性回归方法分别建立了基于操作条件(压力△P=0.04~0.12 MPa,浓度C=0.3~2.0 g/L,温度T=20~40℃)的比阻预测模型,以期用于死端微滤过程操作条件的优化,最后以检验样本的相对误差作为衡量指标,分别采用BP人工神经网络方法和多元线性回归方法对死端微滤过滤酵母悬浮液时的比阻进行了预测.研究结果表明:(1)在本实验范围内,BP人工神经网络模型的最佳拓朴结构为3-7-1,隐层神经元个数为7,学习速率为0.05,学习函数为traingdx,传递函数为Logsig;用多元线性回归方法得到的比阻与操作条件之间的数学关系式为:α_c-=1.639 883+44.2P+0.862 17C-0.060 7T;(2)利用BP人工神经网络和多元线性回归方法预测死端微滤比阻的平均相对误差分别为3.55%和5.16%.由此可见,这两种方法都可用于死端微滤比阻预测,并且前者优于后者.
【Abstract】 In this study,firstly,an orthonoumal design table used as the modeling sample.Secondly,in order to optimize the operating conditions in the dead - end microfiltration,a predictive model between the operating conditions(pressure,0.04 ~ 0.12 MPa;concentration 0.3 ~ 2.0 g/L;temperature 20 ~ 40℃) and the specific resistance was developed by using an BP artificial neural network method and a multi—regression method,respectively. Finally,an relative absolute error used as a testing index,the specific resistance of yeast suspensions during the dead - end microfiltration was predicted by using an BP artificial neural network method and a multi regression method,respectively.The result showed that,(i) in the experimental rang,a optimal configuration of the available artificial neural network was 3 - 7 - 1,the number of hidden neurons was 7,learning rate was0.05,learning function was traingdx,the transfer function was logsig.In addition,a multi - regression model between the specific resistance and the operating conditions was obtained as follows:α_c = 1.639 883 + 44.2 + 0.862 17 - 0.060 7.(ii) Using an BP artificial neural network method and a multi - regression method, a gotten average relative absolute error was 3.55%and 5.16%,respectively.Therefore an artificial neural network method is much better than a multi - regression method in the study of predicting the specific resistance of yeast suspensions in the dead—end microfiltration.
- 【会议录名称】 第三届中国膜科学与技术报告会论文集
- 【会议名称】第三届中国膜科学与技术报告会
- 【会议时间】2007-10-15
- 【会议地点】中国北京
- 【分类号】TQ028
- 【主办单位】中国膜工业协会、北京工业大学