基于卷积神经网络的微细通道热诊断方法研究  

Study on Thermal Diagnosis Method of Mini-channels Based on Convolutional Neural Network

作　　者：李沛[1] 刘心怡 蒋琳[2] 王秋旺[1] 马挺[1] LI Pei;LIU Xinyi;JIANG Lin;WANG Qiuwang;MA Ting(Key Laboratory of Thermo-Fluid Science and Engineering,MOE,Xi’an Jiaotong University,Xi’an 710049,China;Institute of Applied Electronics,China Academy of Engineering Physics,Mianyang 621900,China)

机构地区：[1]西安交通大学,热流科学与工程教育部重点实验室,西安710049 [2]中国工程物理研究院,应用电子学研究所,绵阳621900

出　　处：《工程热物理学报》2025年第3期933-937,共5页Journal of Engineering Thermophysics

基　　金：国家自然科学基金项目(No.52022080)。

摘　　要：高功率微波在卫星、飞机等电子系统毁伤领域有重要应用前景,其电子束收集极附近存在强烈电磁干扰,需要开发能预测微细通道沸腾传热性能的间接热诊断方法。本文通过微细通道沸腾传热可视化实验结果建立样本数据库,在此基础上构建了基于卷积神经网络的微细通道热诊断模型,并对该方法进行分析和验证。研究结果表明,本文所提出的热诊断方法可以预测实验运行工况和热边界条件,测试准确度可达86%,能够实现流动传热参数的热诊断。High power microwave has important application prospects in the field of electronic system damage such as satellite and aircraft.There is strong electromagnetic interference near the electron beam collector,so it is necessary to develop an indirect thermal diagnosis method that can predict the boiling heat transfer performance of mini-channels.In this paper,the sample database was established through the experimental results of mini-channel boiling heat transfer,and then a thermal diagnosis model based on convolutional neural network was proposed for mini-channels.The method was analyzed and validated.The results indicated that the proposed thermal diagnosis method can predict the experimental operating conditions and thermal boundary conditions,and the test accuracy can reach 86%,which can realize the thermal diagnosis of flow and heat transfer parameters.

关 键 词：卷积神经网络 微细通道 热诊断 沸腾传热 

分 类 号：TK172[动力工程及工程热物理—热能工程]