基于深度森林及电阻层析成像的气液两相流流型辨识  

Flow Pattern Identification of Gas-Liquid Two-phase Flow Based on Deep Forest and Electrical Resistance Tomography

作　　者：张立峰[1] 佟彤 肖凯 ZHANG Li-feng;TONG Tong;XIAO Kai(Department of Automation,North China Electric Power University,Baoding,Hebei 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003

出　　处：《计量学报》2023年第6期893-898,共6页Acta Metrologica Sinica

基　　金：国家自然科学基金(61973115)。

摘　　要：提出了一种基于深度森林(DF)算法与电阻层析成像技术(ERT)的气液两相流流型辨识方法。首先利用ERT实验装置对4种典型流型进行数据采集,以多帧数据求均值的方式对采集的数据进行预处理;然后选择合适的基本分类器构建深度森林模型,并调整模型的最大层数以保障分类的准确率;最后对多帧数据求均值的有效性和深度森林模型的流型辨识能力进行验证,并与深度神经网络(DNN)及卷积神经网络(CNN)2种传统深度学习算法进行比较。结果表明深度森林的流型辨识准确性优于其他2种算法,平均辨识精度可达98.75%,多帧数据求均值的预处理方法更有利于流型辨识。A flow pattern identification method of gas-liquid two-phase flow based on deep forest algorithm and electrical resistance tomography(ERT)was proposed.Firstly,ERT experimental device was used to collect data of four typical flow patterns,and the collected data was preprocessed by averaging multi-frame data.Then,some proper basic classifiers were selected to construct a deep forest model,and the maximum number of layers of the model was adjusted to ensure the accuracy of classification.Finally,the validity of multi-frame data averaging and the flow pattern identification ability of deep forest model were verified,and compared with two traditional deep learning algorithms,deep neural network and convolutional neural network.The results show that the accuracy of flow pattern identification of deep forest is better than that of other two algorithms,and the average identification accuracy can reach 98.75%.The preprocessing method of multi-frame data averaging is more conducive to flow pattern identification.

关 键 词：计量学 电阻层析成像 流型辨识 深度森林算法 气液两相流 

分 类 号：TB937[一般工业技术—计量学]