基于粗神经网络特征提取的ECT流型辨识.docVIP

基于粗神经网络和特征提取的ECT流型辨识 刘延东1，李惠强2，何在刚2，刘浩仟2，陈玲2，郑静娜2，郑斯文3 （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） 摘要：针对目前电容层析成像技术流型辨识精度低的问题，本文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。 关键词： 电容层析成像(ECT)；粗神经网络（RNN）；特征提取；流型辨识 Flow pattern identification based on feature extraction and Rough Neural Network for Electrical Capacitance Tomography LIU Yan-dong1，LI HUI-qiang2, HE Zai-gang2，LIU Hao-qian2, CHEN Ling2, ZHENG Jing-na2，ZHENG Si-wen3 (1.Light Industry College，Liaoning University, Shenyang Liaoning 110036, China; 2.College of Physics, Liaoning University, Shenyang Liaoning 110036, China; 3.Shenyang Metrology Testing Institution, Shenyang Liaoning 110179, China) Abstract：In response to the inaccuracy of flow pattern identification by Electrical Capacitance Tomography, a new method is put forward that combines rough neural network with feature extraction, to identify the two-phase flow patterns. Firstly, this method according to the Electrical Capacitance Tomography system and the characteristics of flow pattern to handle capacitance measurements, thus complete the features of various flow pattern extraction; Secondly, the structure of rough neural network to carry on the design, and use the features parameters of typical flow pattern to train the rough neural network,?and then using this rough neural network to identify flow pattern;?Finally, the simulation experiment. The simulation experimental results indicate that this method compared with the traditional BP neural network has high identification accuracy, this also for the research of ECT flow pattern identification provides new ways and means. Key words: Electrical capacitance tomography; Rough Neural Network; feature extraction; flow pattern identification; 1、引言 电容层析成像技术[1-2]是近年发展起来的一项新的检测技术。电容层析成像技术（ECT）因其结构简单、技术成本低、安全性能好等优点而具