基于粗神经网络和特征提取的ECT流型辨识  

Flow Pattern Identification Based on Feature Extraction and Rough Neural Network for Electrical Capacitance Tomography

作　　者：刘延东[1] 李惠强[2] 何在刚 刘浩仟 陈玲[2] 郑静娜 郑斯文[3] 

机构地区：[1]辽宁大学轻型产业学院,辽宁沈阳110036 [2]辽宁大学物理学院,辽宁沈阳110036 [3]沈阳计量测试院,辽宁沈阳110179

出　　处：《辽宁大学学报（自然科学版）》2014年第4期330-337,共8页Journal of Liaoning University：Natural Sciences Edition

基　　金：辽宁省自然科学基金项目(20102082)

摘　　要：针对目前电容层析成像技术流型辨识精度低的问题,提出一种基于粗神经网络与特征提取相结合的方法来辨识两相流流型.该方法首先根据电容层析成像系统和流型的特点来处理电容测量数据,从而完成对各种流型特征的提取;其次对粗神经网络的结构进行设计,并利用典型流型特征参数训练粗神经网络,然后利用此粗神经网络对流型进行辨识;最后进行仿真实验.仿真实验结果表明此种方法较传统的BP神经网络具有较高的识别精度,这也为ECT流型辨识的研究提供了一个新的途径和手段.In response to the inaccuracy of flowpattern identification by Electrical Capacitance Tomography,a newmethod is provided that combines rough neural network with feature extraction,to identify the two-phase flowpatterns. Firstly,this method according to the Electrical Capacitance Tomography system and the characteristics of flowpattern to handle capacitance measurements,thus complete the features of various flowpattern extraction; Secondly,the structure of rough neural network to carry on the design,and use the features parameters of typical flowpattern to train the rough neural network,and then using this rough neural network to identify flowpattern; Finally,the simulation experiment. The simulation experimental results indicate thatthis method has high identification accuracy,comparing with the traditional BP newral network and also provides newways and means for the research of ECT flowpattern identification.

关 键 词：电容层析成像(ECT) 粗神经网络(RNN) 特征提取 流型辨识 

分 类 号：TN911.7[电子电信—通信与信息系统]