A fast general thermal simulation model based on Multi-Branch Physics-Informed deep operator neural network

解算器 人工神经网络 深度学习 热的 领域(数学) 物理 人工智能 计算机科学 机器学习 数学 气象学 程序设计语言 纯数学
作者
Zibo Lu,Yuanye Zhou,Yanbo Zhang,X Hu,Qiao Zhao,Xuyang Hu
出处
期刊:Physics of Fluids [American Institute of Physics]
卷期号:36 (3) 被引量:3
标识
DOI:10.1063/5.0194245
摘要

Thermal simulation plays a crucial role in various fields, often involving complex partial differential equation (PDE) simulations for thermal optimization. To tackle this challenge, we have harnessed neural networks for thermal prediction, specifically employing deep neural networks as a universal solver for PDEs. This innovative approach has garnered significant attention in the scientific community. While Physics-Informed Neural Networks (PINNs) have been introduced for thermal prediction using deep neural networks, existing methods primarily focus on offering thermal simulations for predefined relevant parameters, such as heat sources, loads, boundaries, and initial conditions. However, any adjustments to these parameters typically require retraining or transfer learning, resulting in considerable additional work. To overcome this limitation, we integrated PINN methods with the DeepONet model, creating a novel model called PI-DeepONet for thermal simulation. This model takes both relevant parameters and coordinate points as simultaneous input functions, presenting a fresh computational perspective for thermal simulation. Based on the PaddlePaddle deep learning framework, our research demonstrates that after sufficient training, this model can reliably and rapidly predict parameter solutions. Importantly, it significantly surpasses traditional numerical solvers in terms of speed by several orders of magnitude, without requiring additional training. This groundbreaking research framework holds vast application potential and promises substantial advancements in the field of thermal simulation.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
更新
PDF的下载单位、IP信息已删除 (2025-6-4)

科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
shlw完成签到,获得积分10
刚刚
11112发布了新的文献求助10
刚刚
希望天下0贩的0应助Skuld采纳,获得10
刚刚
2秒前
6秒前
个性南莲完成签到,获得积分10
8秒前
suiwuya完成签到,获得积分10
10秒前
melone完成签到,获得积分10
12秒前
ambition完成签到,获得积分10
14秒前
14秒前
伯爵的猫完成签到,获得积分10
15秒前
17秒前
沐青完成签到,获得积分10
18秒前
18秒前
yaya完成签到 ,获得积分10
19秒前
沙尔发布了新的文献求助10
20秒前
慢跑跑不动的肥仔完成签到,获得积分10
20秒前
知之然完成签到,获得积分10
22秒前
小明仔驳回了666应助
22秒前
星星完成签到,获得积分10
23秒前
Skuld发布了新的文献求助10
23秒前
25秒前
25秒前
Crazy完成签到 ,获得积分10
28秒前
28秒前
领导范儿应助Notdodead采纳,获得10
28秒前
SYLH应助科研通管家采纳,获得30
28秒前
完美世界应助maidongdong采纳,获得10
28秒前
李爱国应助科研通管家采纳,获得10
28秒前
ding应助科研通管家采纳,获得10
28秒前
orixero应助mint采纳,获得10
29秒前
29秒前
29秒前
烟花应助科研通管家采纳,获得10
29秒前
小二郎应助科研通管家采纳,获得10
29秒前
Hello应助科研通管家采纳,获得10
29秒前
彭于晏应助科研通管家采纳,获得10
29秒前
29秒前
牛牛眉目发布了新的文献求助10
29秒前
SciGPT应助科研通管家采纳,获得10
29秒前
高分求助中
A new approach to the extrapolation of accelerated life test data 1000
Cognitive Neuroscience: The Biology of the Mind 1000
Technical Brochure TB 814: LPIT applications in HV gas insulated switchgear 1000
Immigrant Incorporation in East Asian Democracies 500
Nucleophilic substitution in azasydnone-modified dinitroanisoles 500
不知道标题是什么 500
A Preliminary Study on Correlation Between Independent Components of Facial Thermal Images and Subjective Assessment of Chronic Stress 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 工程类 有机化学 生物化学 物理 内科学 纳米技术 计算机科学 化学工程 复合材料 遗传学 基因 物理化学 催化作用 冶金 细胞生物学 免疫学
热门帖子
关注 科研通微信公众号,转发送积分 3966366
求助须知:如何正确求助?哪些是违规求助? 3511778
关于积分的说明 11159739
捐赠科研通 3246353
什么是DOI,文献DOI怎么找? 1793415
邀请新用户注册赠送积分活动 874427
科研通“疑难数据库(出版商)”最低求助积分说明 804374