两相流实验智能化升级及教学研究  

Intelligent upgrades and teaching research of a two-phase flow experiment

作　　者：李辉[1] 吕卓然 符泰然[1] 霍雨佳 许兆峰 陆规[2] LI Hui;LYU Zhuoran;FU Tairan;HUO Yujia;XU Zhaofeng;LU Gui(Key Laboratory for Thermal Sciences and Power Engineering of Ministry of Education,Tsinghua University,Beijing 100084,China;Key Laboratory of Power Station Energy Transfer Conversion and System of Ministry of Education,North China Electric Power University,Beijing 102206,China)

机构地区：[1]清华大学热科学与动力工程教育部重点实验室,北京100084 [2]华北电力大学电站能量传递转化与系统教育部重点实验室,北京102206

出　　处：《实验技术与管理》2025年第3期174-180,共7页Experimental Technology and Management

摘　　要：两相流热工参数测量是能源动力学科一项重要的教学内容,可视化实验系统能够帮助学生直观认识两相流基本现象、流型及其演化规律,在流体力学、传热学及多相流教学中具有重要作用。该文根据两相流实验教学需求,结合最新的人工智能及数字孪生技术,在原先开发的数字化两相流流型演示实验系统基础上做了智能化升级,采用小波分析和灰度直方图分析两种特征向量提取方法,以及特征向量法及卷积神经网络直接图像识别法这两种智能算法用于识别两相流流型,拓展了实验台功能,丰富了教学内容,实现了多学科交叉融合。该文开发的基于人工智能算法的流型识别方法,也为目前两相流含气率测量无法兼顾精度和效率的瓶颈问题提出了新的解决思路。[Objective]The measurement of the thermal parameters of two-phase flow represents a crucial aspect of energy and power engineering,with extensive applications in the teaching and research of nuclear energy,petrochemical processes,and multiphase flow systems.The complex characteristics of two-phase flow—multiple phases and diverse flow patterns across varying scales—pose significant challenges in accurately capturing these phenomena.Therefore,traditional experimental teaching systems often fail to provide intuitive demonstrations of these dynamic characteristics,which limits students’comprehensive understanding of intricate multiphase flow behaviors.To solve this problem,this study introduces an intelligent upgrade of the existing two-phase flow experimental system based on advanced artificial intelligence(AI)and digital twin technologies.The primary purpose of this upgrade is to optimize the efficiency and accuracy of flow pattern recognition while enhancing the overall effectiveness of experimental teaching.Moreover,this innovative approach solves the persistent challenges of balancing precision and efficiency in measuring gas fractions under two-phase flows.[Methods]In this research,two sophisticated feature extraction methods—wavelet analysis and grayscale histogram analysis—are employed to capture the multidimensional characteristics intrinsic to two-phase flow accurately.As a multiscale signal processing method,wavelet analysis can effectively capture transient changes and local factors in two-phase flow.It is particularly suitable for processing nonlinear and nonstationary signals in flow pattern evolution.Meanwhile,grayscale histogram analysis is utilized to statistically derive key image features based on pixel intensity distributions,offering a straightforward yet highly effective method for feature extraction.To further improve the flow pattern recognition process,we introduce two intelligent algorithms:the eigenvector method and convolutional neural network(CNN)image recognition.The eigenvect

关 键 词：气液两相流 流型识别 含气率 人工神经网络 特征提取 

分 类 号：O359[理学—流体力学]