Achieving effective control of thermal and mechanical distributions has been a long-standing goal, and metamaterials have emerged as a crucial tool for customizing functional structures to manipulate these physical fields. However, existing design paradigms do not apply to thermal-mechanical metamaterials that operate on thermal and mechanical fields simultaneously and independently. First, Due to the different geometric requirements imposed by the thermal and mechanical fields on the unit cells, there is a conflict between functional coupling and design coupling, which limits the design of thermal-mechanical metamaterials. Second, the fact that continuum mechanical equations do not remain invariant under general coordinate transformations hinders the application of conventional theories. Additionally, balancing minimal design costs, manufacturability, and optimal functionality remains a significant challenge. Here, we propose a global data-driven design method using Bayesian hyperparameter optimization. This method creates thermal-mechanical metamaterials from a large, pre-computed unit cell database. Our flexible method allows designing thermal-mechanical metamaterials with various functional combinations (e.g., cloaks, concentrators, and rotators) and shapes. Compared to traditional solutions, this approach balances manufacturability and functionality while offering unparalleled universality and low design costs. Experimental measurements validate the effectiveness of our method. Our approach can rapidly respond to new design scenarios and address design challenges related to the multi-physical effects.

中文翻译：

---

热机械多功能超材料的数据驱动设计


实现对热分布和机械分布的有效控制一直是一个长期的目标，超材料已成为定制功能结构以操纵这些物理场的关键工具。然而，现有的设计范式并不适用于同时独立地在热场和机械场上运行的热机械超材料。首先，由于热场和机械场对晶胞的几何要求不同，功能耦合和设计耦合之间存在冲突，这限制了热机械超材料的设计。其次，连续介质力学方程在一般坐标变换下不变的事实阻碍了传统理论的应用。此外，平衡最低设计成本、可制造性和最佳功能仍然是一项重大挑战。在这里，我们提出了一种使用贝叶斯超参数优化的全局数据驱动设计方法。这种方法从大型的预先计算的晶胞数据库创建热机械超材料。我们灵活的方法允许设计具有各种功能组合（例如，隐身器、集中器和旋转器）和形状的热机械超材料。与传统解决方案相比，这种方法平衡了可制造性和功能性，同时提供了无与伦比的通用性和低设计成本。实验测量验证了我们方法的有效性。我们的方法可以快速响应新的设计场景，并解决与多物理场效应相关的设计挑战。