赵紫威,周凤翔,周杨理理,王灿,张佩,汪卫华.人工智能赋能材料创新的层次与体系布局[J].中国科学院院刊,2026,41(6):1115-1126.

人工智能赋能材料创新的层次与体系布局

Artificial intelligence-enabled materials innovation: Implementation levels and strategic layout

赵紫威^1,2
中国科学院东莞材料科学与技术研究所 东莞 523830;松山湖材料实验室 东莞 523830
ZHAO Ziwei^1,2
Dongguan Institute of Materials Science and Technology, Chinese Academy of Sciences, Dongguan 523830, China;Songshan Lake Materials Laboratory, Dongguan 523830, China
周凤翔^1,2*
中国科学院东莞材料科学与技术研究所 东莞 523830;松山湖材料实验室 东莞 523830
ZHOU Fengxiang^1,2*
Dongguan Institute of Materials Science and Technology, Chinese Academy of Sciences, Dongguan 523830, China;Songshan Lake Materials Laboratory, Dongguan 523830, China
周杨理理^1,2
中国科学院东莞材料科学与技术研究所 东莞 523830;松山湖材料实验室 东莞 523830
ZHOU Yanglili^1,2
Dongguan Institute of Materials Science and Technology, Chinese Academy of Sciences, Dongguan 523830, China;Songshan Lake Materials Laboratory, Dongguan 523830, China
王灿¹
中国科学院东莞材料科学与技术研究所 东莞 523830
WANG Can¹
Dongguan Institute of Materials Science and Technology, Chinese Academy of Sciences, Dongguan 523830, China
张佩¹
中国科学院东莞材料科学与技术研究所 东莞 523830
ZHANG Pei¹
Dongguan Institute of Materials Science and Technology, Chinese Academy of Sciences, Dongguan 523830, China
汪卫华^1*
中国科学院东莞材料科学与技术研究所 东莞 523830
WANG Weihua^1*
Dongguan Institute of Materials Science and Technology, Chinese Academy of Sciences, Dongguan 523830, China

         人工智能;材料创新;体系布局

        artificial intelligence;materials innovation;systematic layout

        材料创新长期面临设计空间大、工艺路径复杂、验证周期长和工程转化难等问题。传统材料创制和使役性能优化依赖经验积累、理论推演与实验试错的研发模式已难以适应关键材料快速突破需求。近年来，人工智能技术加快进入材料设计、制备、表征、评估、使役性能优化和应用反馈等环节，推动材料创新由经验驱动向数据驱动、由离散试错向闭环优化转变。文章在梳理材料科学研究范式演进和技术创新范式演进的基础上，提出人工智能赋能材料创新的“能力奠基—应用拓展—产业深化”3层实现框架。通过分析国内外发展布局态势、重点问题及体系布局方向，认为人工智能赋能材料创新正处于由概念验证向系统化应用过渡阶段，材料性质预测、候选筛选和文献知识抽取等前端环节发展较快，而复杂材料合成、实验验证、服役预测和中试放大等环节仍是制约人工智能赋能材料创新走向系统应用的关键短板。因此，建议未来应以关键材料任务为牵引，推动单点技术突破至创新链条贯通，形成覆盖设计、制备、验证、放大、服役和应用反馈的系统能力。

        Materials innovation has long been constrained by vast design spaces, complex processing routes, lengthy validation cycles, and difficulties in engineering translation. Traditional research and development models, which mainly rely on accumulated experience, theoretical deduction, and experimental trial and error, have become increasingly insufficient to meet the demand for rapid breakthroughs in critical materials. In recent years, artificial intelligence has been increasingly integrated into materials design, synthesis, and processing, characterization, evaluation, optimization, and application feedback, promoting the transformation of materials innovation from experience-driven exploration to data-driven development and from discrete trial and error to closed-loop optimization. Based on an analysis of the evolution of materials science research paradigms and technological innovation paradigms, this study proposes a three-level implementation framework for artificial intelligence-enabled materials innovation, namely, “capability foundation—application expansion—industrial deepening”. It further examines the development trends, major challenges, and system-level layout directions in this field. Overall, artificial intelligence-enabled materials innovation is currently in a transitional stage from proof of concept to systematic application. Front-end links such as materials property prediction, candidate screening, inverse design, and literature-based knowledge extraction have developed rapidly, whereas complex materials synthesis, experimental validation, service performance prediction, pilot-scale scale-up, and industrial process embedding remain major bottlenecks. In the future, it is necessary to promote coordinated development across the three implementation levels under the guidance of critical materials tasks, strengthen the innovation chain through synthesis validation, service performance prediction, and pilot-scale scale-up, and support large-scale expansion through platform-based and standardized systems.