摘要
本文利用深度学习构建卷积神经网络算法,成功实现石墨烯/氮化硼二维异质结构的热导率预测。基于非平衡态分子动力学模拟计算得到不同拓扑图案异质结构的热导率构建机器学习数据库,将不同拓扑图案的异质结构图片和对应的热导率作为训练样本数据,搭建卷积神经网络。此外,详细分析了卷积神经网络中超参数对热导率预测准确性的影响。本研究有助于快速预测不同拓扑图案对应的热导率,同时对探索二维异质结构与热导率的热传导物理机制具有重要意义。
In this study deep learning is utilized to construct a convolutional neural network algorithm in order to predict the thermal conductivity of graphene/boron nitride two-dimensional heterostructures. Based on non-equilibrium molecular dynamics simulations, the thermal conductivities of heterogeneous structures with different topological patterns are calculated as a training set for input file, and the heterogeneous structure pictures and the corresponding thermal conductivity are used as training sample data to build a convolutional neural network. In addition, the influence of hydroparameters of the convolutional neural network on the accuracy of thermal conductivity prediction is discussed in detail. This study would be helpful to fast design topological patterns of heterostructures with specified thermal conductivity, and be of great significance for exploring the physical mechanism of heat conduction in two-dimensional heterostructures.
作者
安盟
宋福鑫
陈相全
马维刚
张兴
AN Meng;SONG Fu-Xin;CHEN Xiang-Quan;MA Wei-Gang;ZHANG Xing(Key Laboratory of Thermal Science and Power Engineering,Ministry of Education,Department of Engineering mechanics,Tsinghua University,Beijing 100084,China;College of Mechanical and Electrical Engineering,Shaanxi University of Science and Technology,Xi'an 710021,China)
出处
《工程热物理学报》
EI
CAS
CSCD
北大核心
2022年第3期729-733,共5页
Journal of Engineering Thermophysics
基金
国家自然科学基金项目(No.52006130)
陕西省自然科学基金项目(No.2020JQ-692)。
关键词
深度学习
卷积神经网络
分子动力学模拟
热导率
deep learning
convolution neural network
molecular dynamics simulation
thermal conductivity
