力学研究中"大数据"的启示, 应用与挑战

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.