{"title":"用于湍流子网格尺度解析的双尺度残差网络:先验分析","authors":"Omar Sallam, Mirjam Fürth","doi":"arxiv-2409.07605","DOIUrl":null,"url":null,"abstract":"This paper introduces generative Residual Networks (ResNet) as a surrogate\nMachine Learning (ML) tool for Large Eddy Simulation (LES) Sub Grid Scale (SGS)\nresolving. The study investigates the impact of incorporating Dual Scale\nResidual Blocks (DS-RB) within the ResNet architecture. Two LES SGS resolving\nmodels are proposed and tested for prior analysis test cases: a\nsuper-resolution model (SR-ResNet) and a SGS stress tensor inference model\n(SGS-ResNet). The SR-ResNet model task is to upscale LES solutions from coarse\nto finer grids by inferring unresolved SGS velocity fluctuations, exhibiting\nsuccess in preserving high-frequency velocity fluctuation information, and\naligning with higher-resolution LES solutions' energy spectrum. Furthermore,\nemploying DS-RB enhances prediction accuracy and precision of high-frequency\nvelocity fields compared to Single Scale Residual Blocks (SS-RB), evident in\nboth spatial and spectral domains. The SR-ResNet model is tested and trained on\nfiltered/downsampled 2-D LES planar jet injection problems at two Reynolds\nnumbers, two jet configurations, and two upscale ratios. In the case of SGS\nstress tensor inference, both SS-RB and DS-RB exhibit higher prediction\naccuracy over the Smagorinsky model with reference to the true DNS SGS stress\ntensor, with DS-RB-based SGS-ResNet showing stronger statistical alignment with\nDNS data. The SGS-ResNet model is tested on a filtered/downsampled 2-D DNS\nisotropic homogenous decay turbulence problem. The adoption of DS-RB incurs\nnotable increases in network size, training time, and forward inference time,\nwith the network size expanding by over tenfold, and training and forward\ninference times increasing by approximately 0.5 and 3 times, respectively.","PeriodicalId":501162,"journal":{"name":"arXiv - MATH - Numerical Analysis","volume":"17 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Dual scale Residual-Network for turbulent flow sub grid scale resolving: A prior analysis\",\"authors\":\"Omar Sallam, Mirjam Fürth\",\"doi\":\"arxiv-2409.07605\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"This paper introduces generative Residual Networks (ResNet) as a surrogate\\nMachine Learning (ML) tool for Large Eddy Simulation (LES) Sub Grid Scale (SGS)\\nresolving. The study investigates the impact of incorporating Dual Scale\\nResidual Blocks (DS-RB) within the ResNet architecture. Two LES SGS resolving\\nmodels are proposed and tested for prior analysis test cases: a\\nsuper-resolution model (SR-ResNet) and a SGS stress tensor inference model\\n(SGS-ResNet). The SR-ResNet model task is to upscale LES solutions from coarse\\nto finer grids by inferring unresolved SGS velocity fluctuations, exhibiting\\nsuccess in preserving high-frequency velocity fluctuation information, and\\naligning with higher-resolution LES solutions' energy spectrum. Furthermore,\\nemploying DS-RB enhances prediction accuracy and precision of high-frequency\\nvelocity fields compared to Single Scale Residual Blocks (SS-RB), evident in\\nboth spatial and spectral domains. The SR-ResNet model is tested and trained on\\nfiltered/downsampled 2-D LES planar jet injection problems at two Reynolds\\nnumbers, two jet configurations, and two upscale ratios. In the case of SGS\\nstress tensor inference, both SS-RB and DS-RB exhibit higher prediction\\naccuracy over the Smagorinsky model with reference to the true DNS SGS stress\\ntensor, with DS-RB-based SGS-ResNet showing stronger statistical alignment with\\nDNS data. The SGS-ResNet model is tested on a filtered/downsampled 2-D DNS\\nisotropic homogenous decay turbulence problem. The adoption of DS-RB incurs\\nnotable increases in network size, training time, and forward inference time,\\nwith the network size expanding by over tenfold, and training and forward\\ninference times increasing by approximately 0.5 and 3 times, respectively.\",\"PeriodicalId\":501162,\"journal\":{\"name\":\"arXiv - MATH - Numerical Analysis\",\"volume\":\"17 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-11\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - MATH - Numerical Analysis\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.07605\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - MATH - Numerical Analysis","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.07605","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
摘要
本文介绍了作为大涡模拟(LES)子网格尺度(SGS)解析的替代机器学习(ML)工具的生成残差网络(ResNet)。该研究探讨了在 ResNet 架构中加入双尺度残差块(DS-RB)的影响。针对先期分析测试案例,提出并测试了两种 LES SGS 解析模型:超解析模型(SR-ResNet)和 SGS 应力张量推理模型(SGS-ResNet)。SR-ResNet 模型的任务是通过推断未解决的 SGS 速度波动,将 LES 解决方案从粗网格提升到更精细的网格,成功保留了高频速度波动信息,并与更高分辨率 LES 解决方案的能谱相一致。此外,与单尺度残差块(SS-RB)相比,采用 DS-RB 提高了高频速度场的预测精度和准确性,这在空间和频谱域都很明显。SR-ResNet 模型在两种雷诺数、两种喷流配置和两种升尺度比的过滤/降采样 2-D LES 平面喷流注入问题上进行了测试和训练。在 SGS 应力张量推断方面,参照真实 DNS SGS 应变张量,SS-RB 和 DS-RB 都比 Smagorinsky 模型显示出更高的预测精度,基于 DS-RB 的 SGS-ResNet 与 DNS 数据显示出更强的统计一致性。SGS-ResNet 模型在滤波/降采样的 2-D DNS 各向同性均质衰减湍流问题上进行了测试。采用 DS-RB 后,网络规模、训练时间和前向推理时间显著增加,网络规模扩大了 10 倍以上,训练时间和前向推理时间分别增加了约 0.5 倍和 3 倍。
Dual scale Residual-Network for turbulent flow sub grid scale resolving: A prior analysis
This paper introduces generative Residual Networks (ResNet) as a surrogate
Machine Learning (ML) tool for Large Eddy Simulation (LES) Sub Grid Scale (SGS)
resolving. The study investigates the impact of incorporating Dual Scale
Residual Blocks (DS-RB) within the ResNet architecture. Two LES SGS resolving
models are proposed and tested for prior analysis test cases: a
super-resolution model (SR-ResNet) and a SGS stress tensor inference model
(SGS-ResNet). The SR-ResNet model task is to upscale LES solutions from coarse
to finer grids by inferring unresolved SGS velocity fluctuations, exhibiting
success in preserving high-frequency velocity fluctuation information, and
aligning with higher-resolution LES solutions' energy spectrum. Furthermore,
employing DS-RB enhances prediction accuracy and precision of high-frequency
velocity fields compared to Single Scale Residual Blocks (SS-RB), evident in
both spatial and spectral domains. The SR-ResNet model is tested and trained on
filtered/downsampled 2-D LES planar jet injection problems at two Reynolds
numbers, two jet configurations, and two upscale ratios. In the case of SGS
stress tensor inference, both SS-RB and DS-RB exhibit higher prediction
accuracy over the Smagorinsky model with reference to the true DNS SGS stress
tensor, with DS-RB-based SGS-ResNet showing stronger statistical alignment with
DNS data. The SGS-ResNet model is tested on a filtered/downsampled 2-D DNS
isotropic homogenous decay turbulence problem. The adoption of DS-RB incurs
notable increases in network size, training time, and forward inference time,
with the network size expanding by over tenfold, and training and forward
inference times increasing by approximately 0.5 and 3 times, respectively.