香港中文大學領導的大灣區研究團隊開發物理約束 AI 框架,設計超韌輕質超材料
From disorder to strength: CUHK-led GBA research team develops physics-constrained AI frame-work to design ultra-tough lightweight metamaterials
香港中文大學領導的研究團隊結合統計物理與 AI,開發受物理定律約束的設計框架,利用無序結構設計輕質超材料。團隊以高精度 3D 列印製作並測試材料,發現其能量吸收能力和斷裂韌性均有所提升。研究成果為輕質防護結構及汽車吸能部件的開發開拓新方向。
A research team led by the Department of Physics at The Chinese University of Hong Kong (CUHK), in collaboration with Sun Yat-sen University and Shenzhen University, has successfully harnessed irregular, “messy” structures to design lightweight metamaterials with enhanced energy absorption, improved impact resistance and exceptional fracture toughness. By integrating statistical physics with artificial intelligence (AI), this groundbreaking design framework challenges the conventional belief that advanced materials must rely on rigid, grid-like patterns. Published in the leading physics journal Physical Review X, this discovery opens new avenues for the development of lightweight protective structures, automotive impact-absorbing components and advanced engineering materials.
Physics-guided AI discovers high-performance disordered structures
For decades, the design of advanced materials has largely relied on ordered structures, where regular geometric patterns make it easier to predict and control material properties. However, many of nature’s toughest materials, including bone and spider silk, demonstrate that structural irregularity and disorder actually reduce vulnerability, distribute forces more evenly and introduce unique resilience mechanisms to resist failure. Nevertheless, identifying the single optimal randomised layout out of trillions of possibilities within a virtually infinite design space has long posed a monumental challenge for materials scientists.
Inspired by these natural systems, the CUHK-led research team constructed an AI framework constrained by the fundamental laws of physics. The system applies physical principles to automatically filter out mechanically unstable configurations, efficiently identifying optimised disordered structures that are both stable and highly flexible.
Unlocking unusual material behaviours with auxetic design
The framework is specifically programmed to search for “auxetic” layouts, materials designed to act like a swelling sponge, expanding and thickening when stretched rather than thinning out like a standard rubber band. By restricting the algorithm to physically viable parameters, the AI successfully navigates vast topological combinations without wasteful computation, zeroing in on structures with optimal mechanical performance.
Crucially, this AI framework is not a “black box”. Near an optimised structure, the team mapped the energy configurations to the “Ising model”, a classical physics model used to describe interacting magnetic spins. In this form, the influence of every single structural connection, as well as whether various connections cooperate to resist external pressure or offset one another, can be explicitly described and quantified, giving researchers a traceable blueprint of the material’s internal strength.
To validate the AI-designed structures, the optimised layouts were fabricated using high-precision 3D printing and subjected to a series of mechanical tests. Surprisingly, beyond achieving the targeted behaviours, the resulting materials exhibited unexpected “emergent” properties, including enhanced energy absorption and dramatically improved fracture toughness, far exceeding the initial performance targets.
Winding crack paths unlock superior fracture toughness
The secret to this exceptional toughness lies in how cracks travel through the material. In conventional ordered, grid-like materials, cracks typically propagate along short, straight paths, causing the structure to fail suddenly. In contrast, the randomised structures force propagating cracks to take winding, detour-like paths and trigger multiple tiny microcracks. These complex detours slow crack propagation, allowing the material to dissipate energy more effectively and helping to prevent catastrophic shattering.
Professor Xu Lei of CUHK’s Department of Physics said: “By connecting statistical physics, interpretable AI and practical materials engineering, this framework demonstrates that structural disorder is not a defect, but a powerful design tool. The resulting lightweight, high-performance metamaterials pave the way for vastly safer helmets and automotive bumpers, while the same physics-guided AI design approach may offer opportunities to customise the thermal, acoustic and optical properties of next-generation materials.”
The research was conducted by Professor Xu and PhD graduate Dr Jia Lingyu from CUHK, in collaboration with Associate Professor Shen Xiangying from the School of Science at Sun Yat-sen University and Assistant Professor Zhu Changliang from the College of Mechatronics and Control Engineering at Shenzhen University.
來源:中大新聞中心 · AI 研究及應用 · cpr.cuhk.edu.hk