首页|瞬态多相流场图神经网络时空预测方法研究

瞬态多相流场图神经网络时空预测方法研究

扫码查看
为实现对大型能源化工装备(如循环流化床)内瞬态多相流场的快速时空建模和预测,本文采用基于网格图神经网络的深度学习模型,针对循环流化床非结构化时变数值仿真数据,建立离散相体积分数的时空耦合预测器。该模型有效捕捉了反应器的时空多尺度特性,能高效地进行多相流场时空耦合动态预测,结果表明:速度远超传统数值仿真,加速比可接近 500。
Spatiotemporal prediction method for the transient multiphase flow field via graph neural network
This study aims to rapidly build a spatiotemporal model and predict transient multiphase flow fields within large-energy and chemical equipment(e.g.,circulating fluidized beds).Herein,a deep learning model based on the graph neural network was developed.It established a spatiotemporal predictor for discrete phase volume frac-tions with numerically simulated unstructured time-varying data of a circulating fluidized bed.The model success-fully captured the multiscale spatiotemporal features of the fluidized bed,achieving high-efficiency dynamic predic-tion of the spatiotemporal coupling of the multiphase flow field.The result showed that,compared with traditional numerical simulation,the data-driven model ran significantly faster,with a speedup ratio close to 500.

deep learninggraph neural networkcirculating fluidized bedunstructured meshtransient flow fieldmultiphase flowspatiotemporal predictionmultiscale feature

郝祎琛、谢心喻、丁家琦、谢蓉、王晓放、刘海涛

展开 >

大连理工大学 能源与动力学院,辽宁 大连 116024

深度学习 图神经网络 循环流化床 非结构化网格 瞬态流场 多相流 时空预测 多尺度特征

国家重点研发计划国家自然科学基金青年项目中央高校基本科研业务费国家自然科学基金面上项目辽宁省自然科学基金面上项目

2020YFA071440352005074DUT19RC3070523752312022-MS-135

2024

哈尔滨工程大学学报
哈尔滨工程大学

哈尔滨工程大学学报

CSTPCD北大核心
影响因子:0.655
ISSN:1006-7043
年,卷(期):2024.45(9)