力学研究中"大数据"的启示、应用与挑战  

Big Data in mechanical research:Potentials,applications and challenges

作　　者：杨强[1,2] 孟松鹤[1] 仲政 解维华[1] 郭早阳 金华[1] 张幸红[1] YANG Qiang;MENG Songhe;ZHONG Zheng;XIE Weihua;GUO Zaoyang;JIN Hua;ZHANG Xinghong(Harbin Institute of Technology,National Key Laboratory of Science and Technology for National Defense on Advanced Composites in Special Environments,Harbin 150001,China;Harbin Institute of Technology,Postdoctoral Research Center for Material Science and Engineering,Harbin 150001,China;Harbin Institute of Technology(Shenzhen),Shenzhen 518055,China)

机构地区：[1]哈尔滨工业大学特种环境复合材料国家级重点实验室,哈尔滨150001 [2]哈尔滨工业大学材料科学与工程博士后流动站,哈尔滨150001 [3]哈尔滨工业大学(深圳)理学院,深圳518055

出　　处：《力学进展》2020年第1期406-449,共44页Advances in Mechanics

基　　金：中国博士后科学基金资助项目(2018M641815)的支持

摘　　要：大数据在全世界发展迅猛,应用成效显著.大数据独特的思维和方法,为科学研究与探索提供了全新的范式.力学研究中,高时空分辨率、多参数同步观测与高精度、大规模模拟手段的发展,为力学大数据的发展提供了契机,大数据、机器智能方法的应用正呈现快速上升趋势.本文旨在分析大数据思维方法在力学研究中的应用,及其启示与挑战.首先从大数据资源、大数据科学及大数据技术3个层面分析了大数据的内涵及研究态势,概括了国内外政府及组织机构的大数据发展规划.而后对比分析了力学思维方法与大数据思维方法的特点,指出两者的本质区别在于数据使用方式的不同而带来的范式差异:大数据采用数据驱动模型替代力学中的偏微分方程组以描述问题,在复杂系统的分析、预测中优势显著.回顾了大数据方法在材料性能预测、材料本构建模、湍流建模、结构健康监测及试验力学等方面的最新研究进展,以及动态数据驱动与数字孪生等大数据驱动的建模模拟新范式.总结了大数据在力学研究中应用的3种方式,即驱动已有模型改进,挖掘复杂隐含的规律,以及替代已有的理论方法等.最后,建议以力学研究为主体和牵引,大数据与力学双驱动,推动大数据与力学交叉形成理论与方法突破、及学科发展新方向.Big Data has developed rapidly and made remarkable achievements.The unique thinking and methodology of Big Data provide a new paradigm for scientific research.High spatial-temporal resolution and multiple-parameter synchronous observation methods are offering opportunities for Big Data driven mechanics.Applications of Big Data or machine intelligence methods in mechanical researches have grown rapidly.This review is focused on the impact of Big Data method and its way of thinking on mechanical research and the corresponding challenges.First,the connotation and research situation of Big Data in three aspects,i.e.Big Data itself,Big Data science and Big Data technologies are discussed,and Big Data development plans of governments and organizations are summarized.Comparative analysis of the mechanical methodology and the Big Data methodology were carried out,with focuses on paradigm differences in utilization of data.Instead of partial differential equations used by mechanics,data driven models are used for mathematical descriptions of the underlying physical problem in the Big Data paradigm.The latter shows advantages in simulations and predictions of complex systems.Latest researches in material performance prediction,constitutive modeling,turbulence modeling,structural health monitoring,and experimental mechanics using Big Data,as well as Big Data driven new paradigm of modeling and simulation including Dynamic Data Driven Application System and Digital Twin are reviewed.Three kinds of Big Data driven mechanical researches are summarized,i.e.data driven improvement of existing methods,data mining for hidden and intrinsic physical laws,and data based new methodologies and theories for mechanics.A mechanics-centric approach employing both the prior knowledge of mechanics and advantages of Big Data methodology is recommended for a vision of breakthroughs along with new subject directions in mechanics.

关 键 词：力学 大数据 数据科学 复杂性 不确定性 

分 类 号：N39[自然科学总论] O3[理学—力学]