机械搅拌釜中翼型组合桨持气特性的神经网络模型  

Fractional Gas Hold-up Performance Investigated on the Basis of Neural Network Model in Mechanically Agitated Reactors with Mixed Hydrofoil Impeller K5 and Rushton Turbine

作　　者：沈春银[1] 陈剑佩[1] 张家庭[1] 戴干策[1] 

机构地区：[1]华东理工大学联合化学反应工程研究所,上海200237

出　　处：《华东理工大学学报（自然科学版）》2003年第5期441-446,共6页Journal of East China University of Science and Technology

摘　　要：研究了内径0.382～1.16m机械搅拌釜中翼型组合桨气液两相的持气特性,建立了气含率与结构参数(包括翼型桨径、桨间距、桨下距离、通气位置、挡板形式及翼型桨排出流方向)和操作参数(包括搅拌转速及通气量)间关联的神经网络模型。考察了所建立的网络模型中各参数对气含率的影响规律。结果表明,模型具有很好的泛化能力,其泛化相对误差在±10%以内。采用开槽挡板、低位通气、适宜的桨径、桨间距和翼型桨的排出流向上的方案,在较宽的桨下距离范围内可获得较高的气含率。由于模型使用了与规模无关的无因次参数建模,因此可用于离线预测、参数优化及放大设计。Fractional gas holdup performance has been investigated with the aid of a neural network model. The neural network model has served as a nonlinear relationship performer correlating the fractional gas holdup in agitated reactor (i.d. range is 0.382 m to 1.16 m), with the structure parameters including the ratio of impeller diameter, distance between impellers, clearance of lower impeller, gas sparger position, the baffle type, and the operating conditions such as agitator speed, the rate of gas, and the pumping mode of axial K5 impeller. The correlation obtained in this way is able to predict fractional gas holdup reasonably well and the relative error of generalization by this neural network model is within ±10%, if the parameter falls into the trained range. The neural network model could be used to offline prediction and parameter optimization, and it's useful for scaleup because dimensionless parameters were used.

关 键 词：人工神经网络 气含率 翼型桨 机械搅拌 挡板 

分 类 号：TQ027.3[化学工程]