硅酸盐学报

人工智能赋能功能陶瓷研发范式的破局逻辑与重塑路径

PDF下载

刘润林 ¹ , 景一洋 ¹ , 俞果 ¹ , 高小天 ¹ , 王建 ¹ , 沈忠慧 ¹ , 沈洋 ²

硅酸盐学报 | 综合评述 2026,54(7): 2352-2362

收起

硅酸盐学报 |综合评述 2026 , 54 (7) : 2352 -2362

人工智能赋能功能陶瓷研发范式的破局逻辑与重塑路径

全屏

刘润林¹, 景一洋¹, 俞果¹, 高小天¹, 王建¹, 沈忠慧¹, 沈洋²

作者信息

^1．材料复合新技术全国重点实验室，武汉理工大学材料与微电子学院，武汉 430070

^2．新型陶瓷材料全国重点实验室，清华大学材料学院，北京 100084

通讯作者:

沈忠慧（1991—），男，博士，教授；

沈洋（1980—），男，博士，教授。

作者简介:

刘润林（2001—），男，博士研究生。

Breakthrough Logic and Reshaping Pathways for AI-Enabled Functional Ceramics Research Paradigms

Runlin LIU¹, Yiyang JING¹, Guo YU¹, Xiaotian GAO¹, Jian WANG¹, Zhonghui SHEN¹, Yang SHEN²

Affiliations

^1.State Key Laboratory of Advanced Technology for Materials Synthesis and Processing, School of Materials and Microelectronics, Wuhan University of Technology, Wuhan 430070, China

^2.State Key Lab of New Ceramics and Fine Processing, School of Materials Science and Engineering, Tsinghua University, Beijing 100084, China

Corresponding Author:

SHEN Zhonghui (1991-), male, Ph.D., Professor; E-mail: zhshen@whut.edu.cn;

SHEN Yang (1980-), male, Ph.D., Professor. shyang_mse@mail.tsinghua.edu.cn

LIU Runlin (2001-), male, Doctoral candidate. E-mail: 294515@whut.edu.cn

出版时间: 2026-04-27 doi: 10.14062/j.issn.0454-5648.20260058

文章导航

摘要

收起

功能陶瓷具备优异的多功能耦合响应特性，是支撑高端智能技术发展的关键无机材料。然而，其性能设计与优化同时受多目标性能权衡、多尺度结构关联以及多物理场耦合效应的“三多”困境共同制约，使功能陶瓷材料设计空间呈现出高维、强约束与高度非线性的特征，导致传统依赖经验试错与正向迭代的研发模式效率不高、可迁移性不足。近年来，人工智能（AI）通过整合数据驱动方法与物理约束建模，推动材料研发由正向预测逐步向逆向设计与闭环优化转变，为破解功能陶瓷长期存在的“三多”困境提供了新的技术路径。基于此，本综述从数据、模型与场景应用3个核心层面，系统梳理了AI赋能功能陶瓷研发的研究进展与共性挑战，并进一步围绕逆向反演决策模型、跨尺度智能计算方法以及大语言模型驱动的协同研发平台3条主线，总结了AI赋能功能陶瓷研发的破局逻辑与内在方法框架。最后，面向功能陶瓷材料-工艺-器件一体化的智能研发需求，总结了当前的发展趋势与未来应用前景。

关键词

功能陶瓷 / 逆向设计 / 生成模型 / 智能计算 / 大模型

Abstract

收起

Functional ceramics are a class of key inorganic, non-metallic materials that can couple with and respond to electric, magnetic, mechanical, thermal and optical fields. They underpin a wide range of advanced technologies, including energy conversion, information storage and intelligent sensing. As high-end equipment and frontier science develop rapidly worldwide, the demands placed on the performance of functional ceramics are increasing, pushing materials design towards increasingly extreme compositional, structural and processing conditions. However, progress in this field has long been constrained by the intertwined challenges of multiple objectives, scales and fields. Firstly, the performance of materials and the feasibility of processing are often in strong competition with each other. This means that the design is inherently a problem of multi-objective trade-offs and optimization within a restricted feasible space. Secondly, macroscopic functional responses are regulated by factors spanning multiple length scales, including electronic structure, lattice characteristics, and microstructure. Feedback across these scales can greatly amplify the effect of local perturbations on properties. Thirdly, functional ceramics may undergo nonlinear and irreversible evolution under the combined action of high electric fields, elevated temperatures and mechanical loading, further complicating performance evaluation and controllable design. In recent years, AI for materials has emerged as an important paradigm for addressing complex design problems. Through the integration of machine learning and automation, substantial advances have been achieved in metals, polymers and composites, including property prediction, structure-property correlation analysis and inverse design. For functional ceramics, field coupling, defect sensitivity and process dependence make them a particularly challenging yet highly promising target for AI-enabled research. Against this backdrop, a systematic review of recent progress in AI-driven functional ceramics research, outlining how AI is reshaping the broader material design, is important for accelerating the intelligent design of functional ceramics.

AI-enabled research in functional ceramics can be understood through three connected layers: data, model and application. At the data level, the properties of functional ceramics are jointly governed by composition, microstructure and processing history, rendering the associated datasets intrinsically heterogeneous and complex. Composition is typically represented by stoichiometry and dopant species; microstructural information spans multiple hierarchical levels, from lattice distortion and domain structures to grain size and pore morphology; process data are strongly path dependent and require continuous descriptions of heating, cooling and atmosphere switching; and response data are expressed not only as scalar quantities, such as dielectric, piezoelectric and thermoelectric metrics, but also as spectra, loops and other unstructured responses. Accordingly, the foundation of AI for functional ceramics lies in the construction of high-quality, multimodal and traceable datasets, supported by high-throughput techniques, knowledge extraction and multi-fidelity data fusion to improve both data quantity and quality. At the model level, current studies are converging on an integrated framework comprising prediction, uncertainty, generation and decision models. Prediction models establish forward mappings among composition, processing, structure and properties; uncertainty models support out-of-distribution prediction and reliability assessment; generation models enable inverse searches conditioned on target properties; and decision models combine experimental cost and information gain to guide closed-loop optimization. Within this framework, robustness and interpretability have emerged as central requirements: models could generalize across different systems, processing routes and testing conditions, while physically informed descriptors and mechanistic correlations are needed to reveal the factors that govern functional responses. At the application level, AI has increasingly permeated the full design workflow, giving rise to surrogate modelling, demand-driven exploration, multi-objective optimization and closed-loop pipeline, and thereby creating a more targeted, iterative and efficient design paradigm.

AI is reshaping the design workflow of functional ceramics through three major directions: intelligent inverse design, cross-scale intelligent computation and large-language-model-driven research platforms. First, intelligent inverse design targets on-demand materials design by integrating multiple representations of composition, processing, microstructure and property with uncertainty quantification, physical feasibility constraints and active learning, thereby establishing a closed loop spanning data alignment, forward prediction, inverse design and experimental validation. This strategy can improve the hit rate, manufacturability and reproducibility of proposed candidates. Second, cross-scale intelligent computation leverages machine learning to bridge models at different scales, including electronic structure, atomic dynamics, mesoscale microstructural evolution and continuum responses. By retaining key physical constraints while reducing computational cost, such approaches enable density functional theory, machine-learning interatomic potentials, phase-field models and finite-element analysis to operate within a unified framework for materials design. Third, large-language-model-driven research platforms are evolving from tools for knowledge extraction and literature analysis into collaborative hubs that connect knowledge, computation, decision and experimental execution. By integrating information from papers, patents, laboratory records and process documentation, large language models can assist in building high quality datasets, extracting process pathways and failed experiment information, and coupling with agentic systems and automated experimental platforms, thus paving the way for self-driving closed-loop laboratories in functional ceramics research.

Summary and Prospects

As functional ceramics are increasingly required for more advanced device platforms and harsher service environments, the demand for even higher levels of performance continues to grow. At the core of this challenge lies the persistent coexistence of multi-objective trade-offs and multi-scale problems. Therefore, we systematically summarize the paradigm shifts and representative advances brought about by the introduction of AI into materials research. We outline the broader landscape of AI-enabled functional ceramics research from the perspectives of data, model and application, and further summarize the key strategies and transformative pathways for this field through three major directions: intelligent inverse design, cross-scale intelligent computation and large-language-model-driven research platforms. As data infrastructure, model capability and closed-loop deployment technologies continue to mature, artificial intelligence is poised to evolve from an auxiliary analytical tool into a systematic design framework that connects knowledge, computation, decision and experimental execution, thereby advancing functional ceramics research towards a more efficient and scalable paradigm.

Key words

functional ceramics / inverse design / generative model / intelligent computing / large model

引用本文

刘润林, 景一洋, 俞果, 高小天, 王建, 沈忠慧, 沈洋. 人工智能赋能功能陶瓷研发范式的破局逻辑与重塑路径. 硅酸盐学报, 2026 , 54 (7) : 2352 -2362 . DOI: 10.14062/j.issn.0454-5648.20260058

Runlin LIU, Yiyang JING, Guo YU, Xiaotian GAO, Jian WANG, Zhonghui SHEN, Yang SHEN. Breakthrough Logic and Reshaping Pathways for AI-Enabled Functional Ceramics Research Paradigms[J]. Journal of the Chinese Ceramic Society, 2026 , 54 (7) : 2352 -2362 . DOI: 10.14062/j.issn.0454-5648.20260058

基金

收起

国家自然科学基金青年科学基金项目（B类）(52422206)

参考文献

收起

文献

收起

参考文献引证文献

排序方式：

2026年第54卷第7期

PDF下载

引用本文

BibTeX

文章信息

doi: 10.14062/j.issn.0454-5648.20260058

接收时间：2026-02-02
首发时间：2026-10-22
出版时间：2026-04-27

补充材料

相关文章

文章信息

作者

出版历史

收稿日期：2026-02-02
修回日期：2026-03-02

基金

国家自然科学基金青年科学基金项目（B类）(52422206)

作者信息

^1．材料复合新技术全国重点实验室，武汉理工大学材料与微电子学院，武汉 430070

^2．新型陶瓷材料全国重点实验室，清华大学材料学院，北京 100084

通讯作者:

沈忠慧（1991—），男，博士，教授；

沈洋（1980—），男，博士，教授。

参考文献

分享链接

https://castjournals.cast.org.cn/joweb/gsyxb/CN/10.14062/j.issn.0454-5648.20260058

分享至

全文二维码

扫描看全文

引用本文

BibTeX

本文的引用情况

2种不同金属材料的力学参数

科 Family	属数 Number of genus	种数 Number of species	占总种数比例 Percentage of total species (%)	属 Genus	种数 Number of species	占总种数比例 Percentage of total species (%)
鹅膏菌科Amanitaceae	2	11	5.26	鹅膏菌属 Amanita	10	4.78
小菇科 Mycenaceae	2	12	5.74	丝盖伞属 Inocybe	5	2.39
多孔菌科 Polyporaceae	8	14	6.70	蜡蘑属 Laccaria	5	2.39
红菇科 Russulaceae	3	23	11.00	小皮伞属 Marasmius	6	2.87
				小菇属 Mycena	11	5.26
				光柄菇属 Pluteus	5	2.39
				红菇属 Russula	17	8.13
				栓菌属 Trametes	5	2.39

关闭全屏

BibTeX
EndNote
RefWorks
TxT