基于神经网络的槽式孔板湿气计量修正模型  

Wet gas flow metering correction model of slotted orifice based on neural network

作　　者：华陈权[1,2] 王昌明[1] 耿艳峰[2] 王晓智[2] 

机构地区：[1]南京理工大学机械工程学院,江苏南京210094 [2]中国石油大学信息与控制工程学院,山东东营257061

出　　处：《中国石油大学学报（自然科学版）》2009年第6期152-156,共5页Journal of China University of Petroleum(Edition of Natural Science)

基　　金：国家自然科学基金项目(60672003);山东省自然科学基金项目(Y2006F49)

摘　　要：以空气-水为介质对槽式孔板进行了湿气计量特性试验研究,提出了一种新的基于神经网络的槽式孔板湿气计量修正模型。模型以Lockhart-Martinelli参数X、气体弗劳德准数Frg、密度比Rd、孔径比β四个无量纲参数作为模型的输入,'虚高'OR作为输出。结果表明,在表压为0.25～0.35 MPa,X为0.02～0.6,Frg为0.5～2.7,β为0.5～0.75的测试范围内,模型能够很好地预测实际'虚高',用新修正模型对由于液相存在而引入的气相流量误差进行修正后,气相流量相对误差在95%的置信度下小于±4%,明显优于其他槽式孔板湿气计量修正模型,可以满足生产计量的精度要求。The wet gas metering characteristics of slotted orifice meter were discussed by using air-water as media. A novel wet gas metering correction model based on BP neural network was proposed. In the model, Lockbart-Martinelli parameter X, gas Froude number, the gas to liquid density ratio Rd, and bore diameter ratio β are the inputs, and over-reading is the output. The results show that the new correction model can predict the over-reading accurately, and correct the liquid-induced gas flow rate prediction error of the wet gas flow to ＋4% at 95% confidence level under the conditions of pressure from 0. 25 MPa to 0.35 MPa, X from 0. 02 to 0. 6, gas Froude number from 0. 5 to 2. 7, and β from 0. 5 to 0. 75. The model is superior to other correction model and can satisfy the accuracy requirement of production metering.

关 键 词：神经网络 湿气 槽式孔板 计量修正模型 虚高 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置]