Inverting the structure-property map of truss metamaterials by deep learning.
Inverting the structure-property map of truss metamaterials by deep learning.
复制标题
DOI：
10.1073/pnas.2111505119
复制
发表时间：
2022-01-04
影响因子：
11.1
通讯作者：
Kochmann DM
中科院分区：
文献类型：
--
作者：
Bastek JH;Kumar S;Telgen B;Glaesener RN;Kochmann DM
十多年来，人们一直致力于利用周期性组合桁架超材料丰富的力学领域。然而，可制造单元格的巨大设计空间给逆向设计带来了挑战：如何有效地识别具有目标特性的复杂桁架？我们通过数据驱动的方法来回答这个问题，该方法立即（一旦训练，在几毫秒内）生成不是一个而是各种桁架单元格，其有效响应与给定的（完全各向异性）刚度张量密切匹配。此外，我们的框架可以在不同的单元格之间平滑过渡，从而设计出具有空间变化和局部优化特性的轻质结构，适用于从导波到人造骨的应用。受晶体学的启发，将桁架定期组装成建筑材料已经流行了十多年，并产生了无数具有有益机械性能的细胞结构。尽管桁架设计空间的成功和稳定丰富，但逆向设计仍然是一个挑战：虽然预测有效的桁架特性现在已经很常见，但有效识别具有均匀或空间变化目标特性的架构仍然是从轻量级结构到仿生植入物应用的障碍。为了克服这一差距，我们提出了一个深度学习框架，该框架将神经网络与强制物理约束相结合，以预测具有完全定制的各向异性刚度的桁架结构。在数以百万计的单元格上进行训练，它涵盖了拓扑结构不同的桁架晶格的巨大设计空间，并准确识别出与以前未见过的刚度响应匹配的结构。我们展示了应用于匹配临床刚度数据的患者特异性骨植入物，并讨论了扩展到具有局部最佳特性的空间梯度细胞结构。
More than a decade of research has been devoted to leveraging the rich mechanical playground of periodically assembled truss metamaterials. The enormous design space of manufacturable unit cells, however, has made the inverse design a challenge: How does one efficiently identify a complex truss that has given target properties? We answer this question by a data-driven method, which instantly (once trained, within milliseconds) generates not one but a variety of truss unit cells, whose effective response closely matches a given (fully anisotropic) stiffness tensor. Moreover, our framework to smoothly transition between different unit cells enables the design of lightweight structures with spatially varying, locally optimized properties, for applications from wave guiding to artificial bone. Inspired by crystallography, the periodic assembly of trusses into architected materials has enjoyed popularity for more than a decade and produced countless cellular structures with beneficial mechanical properties. Despite the successful and steady enrichment of the truss design space, the inverse design has remained a challenge: While predicting effective truss properties is now commonplace, efficiently identifying architectures that have homogeneous or spatially varying target properties has remained a roadblock to applications from lightweight structures to biomimetic implants. To overcome this gap, we propose a deep-learning framework, which combines neural networks with enforced physical constraints, to predict truss architectures with fully tailored anisotropic stiffness. Trained on millions of unit cells, it covers an enormous design space of topologically distinct truss lattices and accurately identifies architectures matching previously unseen stiffness responses. We demonstrate the application to patient-specific bone implants matching clinical stiffness data, and we discuss the extension to spatially graded cellular structures with locally optimal properties.
登录
查看更多内容
影响因子：
9.7
作者：
Kumar, Siddhant;Tan, Stephanie;Kochmann, Dennis M.
通讯作者：
Kochmann, Dennis M.
DOI：
10.1016/j.ijsolstr.2020.08.013
发表时间：
2020-12-01
影响因子：
3.6
作者：
Glaesener, Raphael N.;Traff, Erik A.;Kochmann, Dennis M.
通讯作者：
Kochmann, Dennis M.
影响因子：
13.3
作者：
Izard, Anna Guell;Bauer, Jens;Valdevit, Lorenzo
通讯作者：
Valdevit, Lorenzo
影响因子：
3.9
作者：
Kim, T;Zhao, CY;Hodson, HP
通讯作者：
Hodson, HP
影响因子：
8.4
作者：
Feng, Jiawei;Liu, Bo;Fu, Jianzhong
通讯作者：
Fu, Jianzhong