[{"data":1,"prerenderedAt":171},["ShallowReactive",2],{"content-query-VCmJWc1dLl":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"date":10,"cover":11,"type":12,"body":13,"_type":165,"_id":166,"_source":167,"_file":168,"_stem":169,"_extension":170},"/news/zh/2685","zh",false,"","登上流体力学顶刊！ 基于昇思MindSpore打造的AI+流体仿真大模型论文发表","中国商飞上海飞机设计研究院打造的二维超临界翼型流场快速预测方案相关研究论文被流体力学顶刊《Physics of Fluids》接收，该方案基于昇思MindSpore AI框架孵化了成果“东方.御风”，助力大飞机设计加速。","2023-07-31","https://obs-mindspore-file.obs.cn-north-4.myhuaweicloud.com/file/2023/08/02/286c89395f104efc9a45ef9cbc924bcd.png","news",{"type":14,"children":15,"toc":160},"root",[16,24,45,50,62,71,76,81,86,93,98,103,110,117,122,127,132,141,155],{"type":17,"tag":18,"props":19,"children":21},"element","h1",{"id":20},"登上流体力学顶刊-基于昇思mindspore打造的ai流体仿真大模型论文发表",[22],{"type":23,"value":8},"text",{"type":17,"tag":25,"props":26,"children":27},"p",{},[28,30,36,38,43],{"type":23,"value":29},"中国商飞上海飞机设计研究院打造的",{"type":17,"tag":31,"props":32,"children":33},"strong",{},[34],{"type":23,"value":35},"二维超临界翼型流场快速预测方案",{"type":23,"value":37},"相关研究论文被流体力学顶刊《Physics of Fluids》接收，该方案",{"type":17,"tag":31,"props":39,"children":40},{},[41],{"type":23,"value":42},"基于昇思MindSpore AI框架孵化了****成果“东方.御风”",{"type":23,"value":44},"，助力大飞机设计加速。",{"type":17,"tag":25,"props":46,"children":47},{},[48],{"type":23,"value":49},"该成果已由中国商飞首席科学家吴光辉院士在2022年9月举办的世界人工智能大会WAIC2022上发布，北京大学数学学院董彬老师团队参与指导孵化，相关代码和部分数据集已在开源社区Gitee的MindSpore Flow代码仓开源[1]。",{"type":17,"tag":25,"props":51,"children":52},{},[53,55,60],{"type":23,"value":54},"“东方.御风” 是利用昇腾AI澎湃算力打造的面向大型客机翼型流场高效高精度AI预测仿真模型，并在昇思MindSpore流体仿真套件的支持下，",{"type":17,"tag":31,"props":56,"children":57},{},[58],{"type":23,"value":59},"有效提高了对复杂流动的仿真能力，单次仿真时间相较传统方式缩短20倍以上",{"type":23,"value":61},"，大幅提高仿真效率的同时能精准预测，流场平均误差降低至万分之一量级，达到工业级标准。",{"type":17,"tag":25,"props":63,"children":64},{},[65],{"type":17,"tag":66,"props":67,"children":70},"img",{"alt":68,"src":69},"image.png","https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230802033052.31156039542057075517936947333060:50540801063838:2400:F764C8583CB2ED86C0C0C8CFAE6C13AE7B751E82D6FD2CAA1D4870135F81A64B.png",[],{"type":17,"tag":25,"props":72,"children":73},{},[74],{"type":23,"value":75},"“东方.御风”的表面压力系数预测结果及仿真效率对比",{"type":17,"tag":25,"props":77,"children":78},{},[79],{"type":23,"value":80},"在飞行器的设计中，机翼的阻力分布约占整体飞行阻力的52%，因此，机翼形状设计对飞机整体的飞行性能而言至关重要。为了实现超临界翼型的的AI高效高精度流场预测，需要克服如下的技术难点：翼型网格疏密不均，流动特征提取困难；不同气动参数或翼型形状发生改变时，流动特征变化明显；激波区域流场变化剧烈，预测困难。",{"type":17,"tag":25,"props":82,"children":83},{},[84],{"type":23,"value":85},"针对如上所述的技术难点，我们设计了基于AI模型的技术路径图[2]，构建不同流动状态下翼型几何及其对应流场的端到端映射，主要包含以下几个核心步骤：首先，设计AI数据高效转换工具，实现翼型流场复杂边界和非标数据的特征提取，如图数据预处理模块。先通过曲线坐标系网格转换程序实现规则化AI张量数据生成，再利用几何编码方式加强复杂几何边界特征的提取。其次，利用神经网络模型，实现不同流动状态下翼型构型和流场物理量的映射，如图ViT-based encoder-decoder所示；模型的输入为坐标转换后所生成的翼型几何信息和气动参数；模型的输出为转换后生成的流场物理量信息，如速度和压力。最后，利用多级小波变换损失函数训练网络的权重。对流场中突变高频信号进行进一步地分解学习，进而提升流场剧烈变化区域（如激波）的预测精度，如图loss function对应的模块。",{"type":17,"tag":25,"props":87,"children":88},{},[89],{"type":17,"tag":66,"props":90,"children":92},{"alt":68,"src":91},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230802033111.03305613419488941088962568379063:50540801063838:2400:7D39EB3F309BC5CCA3E67B76D1BAC4F8C14890E71CFCA93C12579260FC5662EB.png",[],{"type":17,"tag":25,"props":94,"children":95},{},[96],{"type":23,"value":97},"“东方.御风”模型的技术路径",{"type":17,"tag":25,"props":99,"children":100},{},[101],{"type":23,"value":102},"基于上述的方法策略和优化方案，“东方.御风”采用与壁面距离相关的几何编码技术和多级小波变换的损失函数，将全流场的预测误差降低了20%，激波附近区域的流场预测误差提升了50%。重构流场结果如下图所示：",{"type":17,"tag":25,"props":104,"children":105},{},[106],{"type":17,"tag":66,"props":107,"children":109},{"alt":68,"src":108},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230802033128.98913590366426957847301313767771:50540801063838:2400:9EA0C44F58802B8467583D56B62CE6A50DD9236DFE0C8142A5B79058A624FCB1.png",[],{"type":17,"tag":25,"props":111,"children":112},{},[113],{"type":17,"tag":66,"props":114,"children":116},{"alt":68,"src":115},"https://fileserver.developer.huaweicloud.com/FileServer/getFile/cmtybbs/e64/154/b38/90a1d5d431e64154b387b3660e356ff5.20230802033141.72980553186412879313217621202946:50540801063838:2400:B9116130315DFDC041CD2626FB75EFCEDB1786BB18379CEB02AD528D9DC297CB.png",[],{"type":17,"tag":25,"props":118,"children":119},{},[120],{"type":23,"value":121},"采用多级小波变换误差的优化效果",{"type":17,"tag":25,"props":123,"children":124},{},[125],{"type":23,"value":126},"此外，这个模型还可以作为2D翼型流场的预训练模型，对于全新的翼型和工况，进行快速微调泛化，消融实验表明，仅需1~5个流场数据的分钟级迁移学习，模型的推理预测精度就可达1e-4量级。",{"type":17,"tag":25,"props":128,"children":129},{},[130],{"type":23,"value":131},"“东方.御风”作为2D超临界翼型的研究起点，具备高精度和高效率等特点，为进一步三维翼型的研究提供了深厚的技术铺垫。未来，AI方法将进一步赋能流体力学、气象、海洋等基础研究，展现出极大的应用前景。",{"type":17,"tag":133,"props":134,"children":136},"h2",{"id":135},"参考文献",[137],{"type":17,"tag":31,"props":138,"children":139},{},[140],{"type":23,"value":135},{"type":17,"tag":25,"props":142,"children":143},{},[144,146],{"type":23,"value":145},"[1]",{"type":17,"tag":147,"props":148,"children":152},"a",{"href":149,"rel":150},"https://gitee.com/mindspore/mindscience/tree/master/MindFlow/applications/data%5C_driven/airfoil/2D%5C_steady",[151],"nofollow",[153],{"type":23,"value":154},"https://gitee.com/mindspore/mindscience/tree/master/MindFlow/applications/data\\_driven/airfoil/2D\\_steady",{"type":17,"tag":25,"props":156,"children":157},{},[158],{"type":23,"value":159},"[2]Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies. ZW Deng et al. Physics of Fluids 35, 075146 (2023)",{"title":7,"searchDepth":161,"depth":161,"links":162},4,[163],{"id":135,"depth":164,"text":135},2,"markdown","content:news:zh:2685.md","content","news/zh/2685.md","news/zh/2685","md",1776506070639]