硅酸盐学报

人工智能驱动可持续混凝土：方法、机遇、挑战

PDF下载

姜占 ¹ , 滕乐 ² , 童昕阳 ¹ , 程博远 ¹ , 元强 ³ , 龙武剑 ⁴ , 蒙维娜 ¹ , 保义 ¹

硅酸盐学报 | 水泥基材料基因与智能设计专题—综合评述 2026,54(3): 1016-1031

收起

硅酸盐学报 |水泥基材料基因与智能设计专题—综合评述 2026 , 54 (3) : 1016 -1031

人工智能驱动可持续混凝土：方法、机遇、挑战

全屏

姜占¹, 滕乐², 童昕阳¹, 程博远¹, 元强³, 龙武剑⁴, 蒙维娜¹, 保义¹

作者信息

^1．史蒂文森理工大学土木工程学院，新泽西霍博肯 07030，美国

^2．东南大学材料科学与工程学院重大基础设施工程材料全国重点实验室，南京 211189，中国

^3．中南大学土木工程学院，长沙 410075，中国

^4．深圳大学土木与交通工程学院，广东深圳 518060，中国

通讯作者:

保义（1987—），男，博士，副教授。

作者简介:

姜占（1996—），男，博士研究生。

Artificial Intelligence for Sustainable Concrete: Methods, Challenges and Opportunities

Zhan JIANG¹, Le TENG², Xinyang TONG¹, Boyuan CHENG¹, Qiang YUAN³, Wujian LONG⁴, Weina MENG¹, Yi BAO¹

Affiliations

^1.Department of Civil, Environmental and Ocean Engineering, Stevens Institute of Technology, Hoboken NJ 07030, USA

^2.School of Materials Science and Engineering, Southeast University, Nanjing 211189, China

^3.School of Civil Engineering, Central South University, Changsha 410075, China

^4.College of Civil and Transportation Engineering, Shenzhen University, Shenzhen 518060, Guangdong, China

Corresponding Author:

BAO Yi (1987-), male, Ph.D., Associate Professor. E-mail: yi.bao@stevens.edu

JIANG Zhan (1996-), male, Doctoral candidate. E-mail: zjiang43@stevens.edu

出版时间: 2026-03-20 doi: 10.14062/j.issn.0454-5648.20250765

文章导航

摘要

收起

人工智能（AI）技术为可持续混凝土研发开辟了新途径，能够在保障材料性能的同时，有效平衡成本、碳足迹和能耗等可持续性指标。本文系统回顾了面向可持续混凝土的AI技术，重点讨论了其可解释性和可信赖性提升方法。首先，总结了传统AI与先进AI方法在提升混凝土可持续性方面的应用现状；随后，分析了数据“质”与“量”、预测可靠性、多目标优化设计以及更广泛的社会性考量等当前面临的关键挑战；同时探讨了将AI与机理模型、全生命周期评估模型及信息物理系统相结合，以实现自动化与自适应混凝土生产的潜在路径；最后，提出了未来的研究方向，并规划了基于AI的可持续混凝土基础设施发展路径。旨在启发研究人员与工程师利用AI技术，共同推动绿色建筑环境的构建。

关键词

混凝土 / 网络物理系统 / 数字孪生 / 可解释型人工智能 / 多目标优化 / 高信任度机器学习 / 废物增值

Abstract

收起

Concrete is the most widely used construction material with an annual consumption of about 30 billion tons. However, the production of concrete creates significant environmental problems. Cement is a key binder in concrete, and its production requires burning limestone in high-temperature kilns, which causes approximately 8% of global carbon dioxide emissions. As urbanization and infrastructure construction continue to grow rapidly, the demand for cement and concrete is expected to be increased by 12%-23% by 2050. If we do not develop sustainable alternatives, this will affect a climate change. To address this issue, the industry is increasingly using supplementary cementitious materials, such as fly ash, slag, and calcined clay. In addition, using solid waste like mine tailings and demolition waste also becomes an important strategy. These methods can reduce the use of cement clinkers and keep waste out of landfills, providing both environmental and economic benefits. However, incorporating these diverse materials makes the design process much harder. Conventional design methods rely on experience and trial-and-error experiments. This process is time-consuming, consumes many resources, and cannot fully reflect the complex interactions between different materials. For instance, mixing multiple types of waste materials creates a high-dimensional design space that is hard to manage with simple, linear experiments.

Artificial Intelligence (AI) technology offers an effective way to solve these challenges. Machine learning models can accurately predict key properties of concrete, such as fresh properties, mechanical strength, and durability, based on the historical experimental data. This review systematically summarizes AI techniques for sustainable concrete. The research moves from simple data mapping to more intelligent systems. Initially, some classic models like artificial neural networks and support vector machines are used. These models act as a map between the ingredients (inputs) and the performance (outputs). They can be used for capturing non-linear relationships that conventional equations cannot obtain. For instance, models like XGBoost and Random Forest can predict compressive strength and chloride resistance with a high accuracy, even when using relatively small datasets. These models are often used in a design loop, i.e., predicting performance and using optimization algorithms to adjust the mixture, balancing high strength with lower cost and reduced carbon emissions.

However, conventional AI models have a major problem, i.e., they are often "black boxes". This means engineers cannot see how the model makes decisions, leading to a lack of trust in safety-critical projects. Also, these models depend heavily on data quality, and the existing datasets in concrete science are often small, inconsistent, or lack standard formatting. To fix this issue, the focus is shifting to advanced AI models. This review highlights three key improvements. First, data augmentation techniques, such as generative adversarial networks, are used to create synthetic data. This can solve the problem of small datasets via learning the patterns of real data and generating new, realistic samples. Second, explainable AI is used to open the "black box". Methods like SHAP analysis can show exactly how much each ingredient contributes to the final strength, making the results easier to understand and verify. Third, knowledge-guided learning combines data with physical laws. AI models can follow scientific principles via adding physical constraints (like diffusion laws) or using Knowledge Graphs, making them more reliable and scientifically correct compared to pure data-driven models.

Summary and Prospects

This review concludes that AI makes a great progress in predicting performance and optimizing mixtures, while there are still some challenges before it is widely used in real engineering. The main challenges include the lack of high-quality open datasets, the poor ability of models to work on new materials (generalization), and the need for better trust and interpretability. Future research needs to move from static predictions to autonomous systems.

In the future, an AI-driven autonomous system can be used for sustainable concrete. This involves building a cyber-physical system. In this system, sensors collect real-time data during concrete production. AI models can analyze the data and automatically adjust the mixture proportions to ensure consistent quality. This allows for adaptive manufacturing, reducing waste and ensuring that the concrete meets the design requirements. Furthermore, Digital Twin technology can extend this to the entire life of the structure. A digital twin is a virtual copy of the physical infrastructure. It can update itself using sensor data to monitor the health of the structure and calculate its carbon footprint in real time. This can favor to make better decisions about maintenance and repair, extending the service life of the infrastructure.

Finally, we need to focus on five priorities, i.e., creating open benchmark datasets for fair comparison; developing physics-informed AI that combines data with expert knowledge; advancing multi-objective optimization to balance performance, cost, and carbon; building cyber-physical systems and digital twins; and establishing standards and training programs. AI should not replace human experts. Instead, it should act as an intelligent assistant, helping engineers to design greener and more durable concrete. This human-AI collaboration is a key to achieving a carbon-neutral built environment in the future.

Key words

concrete / cyber-physical system / digital twin / explainable artificial intelligence / multi-objective optimization / trustworthy machine learning / waste valorization

引用本文

姜占, 滕乐, 童昕阳, 程博远, 元强, 龙武剑, 蒙维娜, 保义. 人工智能驱动可持续混凝土：方法、机遇、挑战. 硅酸盐学报, 2026 , 54 (3) : 1016 -1031 . DOI: 10.14062/j.issn.0454-5648.20250765

Zhan JIANG, Le TENG, Xinyang TONG, Boyuan CHENG, Qiang YUAN, Wujian LONG, Weina MENG, Yi BAO. Artificial Intelligence for Sustainable Concrete: Methods, Challenges and Opportunities[J]. Journal of the Chinese Ceramic Society, 2026 , 54 (3) : 1016 -1031 . DOI: 10.14062/j.issn.0454-5648.20250765

基金

收起

美国国家科学基金会项目(2046407)
美国科学院项目(SCON10001241)

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3期

PDF下载

引用本文

BibTeX

文章信息

doi: 10.14062/j.issn.0454-5648.20250765

接收时间：2025-10-20
首发时间：2026-10-27
出版时间：2026-03-20

补充材料

相关文章

文章信息

作者

出版历史

收稿日期：2025-10-20
修回日期：2025-11-11

基金

美国国家科学基金会项目(2046407)

美国科学院项目(SCON10001241)

作者信息

^1．史蒂文森理工大学土木工程学院，新泽西霍博肯 07030，美国

^2．东南大学材料科学与工程学院重大基础设施工程材料全国重点实验室，南京 211189，中国

^3．中南大学土木工程学院，长沙 410075，中国

^4．深圳大学土木与交通工程学院，广东深圳 518060，中国

通讯作者:

保义（1987—），男，博士，副教授。

参考文献

分享链接

https://castjournals.cast.org.cn/joweb/gsyxb/CN/10.14062/j.issn.0454-5648.20250765

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