基于仿真和数据驱动的先进结构材料设计

先进结构材料近年来受到材料和结构设计领域的广泛关注, 这些材料一般通过多个尺度的结构设计实现各种卓越的性能. 在早期的材料设计中, 有的基于设计者的丰富经验, 从天然拓扑结构中抽象出合理的数学力学模型; 有的基于生物系统的结构和功能特点提取出仿生力学模型. 然而, 仅依靠经验性的巧妙设计很难得到最优的设计方案, 通过反复迭代设计和试验来遍历设计空间也不切实际. 为此, 拓扑优化方法被成功应用于声子晶体、元胞材料等先进结构材料的优化设计中, 但现有的拓扑优化方法在实现精准的反向设计方面尚存挑战. 基于数据驱动的机器学习方法擅长建立数据空间多维变量复杂关系, 能够揭示传统力学研究方法难以发现的更深层次的力学机理和规律, 成为力学领域崭新的研究热点. 本文系统地回顾先进结构材料设计方法的发展历程, 对比阐述各种主流设计方法, 结合本课题组的相关工作介绍数值仿真和数据驱动在先进结构材料的智能化设计方面的应用现状, 并对该领域的未来研究趋势进行探讨和展望. 

Advanced structural material design based on simulation and data-driven method

Advanced structural materials have received extensive attention in the field of materials and structural design in recent years. These materials generally achieve excellent performances via structural design at multiple length-scales. In the early material design, some researchers created reasonable mathematical and mechanical models from the natural topologies; some researchers established bionic mechanical models based on the structural and functional characteristics of biological systems. Nevertheless, it isn't easy to obtain the optimal designs only based on ingenious design. To traverse the design space to search for the optimal design by trial and error is also not practical. For these reasons, the topology optimization method has been successfully applied to the design of advanced structural materials such as phononic crystals, cellular materials, etc. However, the existing topology optimization methods still have challenges in achieving accurate reverse designs. Data-driven methods can establish complex relationships of multi-dimensional variables, and they can reveal mechanical mechanisms and laws that are difficult to be discovered by traditional methods. Hence, this paper systematically reviews the development of advanced structural material design methods. Various mainstream design methods are compared and illustrated. The status of intelligent design of advanced structural materials based on data-driven methods is introduced. The prospect of this research area is discussed.