Mr. Rishabh Ray,
Mr. Shirshendu Maitra,
Ms. Anusri Mukhopadhyay,
- , Chief Executive Officer and Co-Founder, Aurora Technologies Private Limited, Pennsylvania, USA, ,
- , Head and Assistant Professor, Department of MCA, Thakur Institute of Management Studies Career Development and Research Mumbai, Maharashtra, India, ,
- , Assistant Professor, Department of MCA, Thakur Institute of Management Studies Career Development and Research Mumbai, Maharashtra, India, ,
Abstract
Large Language Models (LLMs) are emerging as transformative tools in materials science, offering human-like reasoning, zero-shot problem solving, and the ability to integrate fuzzy laboratory knowledge with structured data. This study extends and reinterprets the original systematic benchmark for using LLMs in sustainable concrete design, particularly for Alkali-Activated Concrete (AAC). We introduce an enhanced, multi-model framework combining LLM-based inverse design, Random Forest regression, Gaussian Process Regression (GPR), and a lightweight Artificial Neural Network (ANN). Additional figures and datasets are incorporated to strengthen reproducibility and provide deeper insight into model behavior. Results indicate that LLM-based design assistants can match or outperform classical data-driven approaches without requiring training data, while hybrid LLM–ML models significantly improve stability in predicting 28-day compressive strength. The proposed framework also demonstrates improved interpretability in identifying the influence of precursor materials, activator ratios, curing conditions, and supplementary binders on AAC performance. Comparative evaluation across multiple datasets highlights the adaptability of LLM-assisted optimization under limited-data conditions, which is a major challenge in sustainable construction materials research. Furthermore, uncertainty quantification using GPR reveals enhanced robustness and reliability of predictions when integrated with LLM-generated mix recommendations. The study emphasizes the growing role of explainable and collaborative AI systems in accelerating low-carbon concrete development and reducing experimental trial-and-error processes. Overall, the findings establish that the integration of generative AI with conventional machine learning can provide a scalable, efficient, and intelligent pathway for next-generation sustainable concrete mix design, supporting both academic research and industrial applications in environmentally responsible infrastructure development.
Keywords: Gen- AI, AAC, sustainable concrete, LLM
[This article belongs to Recent Trends in Civil Engineering & Technology ]
References
- Völker C, Rug T, Jablonka KM, Kruschwitz S. LLMs can design sustainable concrete – a systematic benchmark [Internet]. 2024. Available from: https://doi.org/10.21203/rs.3.rs-3913272/v1
- Patel P, Patel A. Sustainable green materials in construction. IOP Conf Ser Earth Environ Sci. 2021.
- Racaza O, Cabahug R. Coconut husk ash as partial cement substitute. Mindanao J Sci Technol. 2016.
- Claisse P. Civil engineering materials. Oxford: Butterworth-Heinemann; 2015.
- Li M, Zhang M, Chen Y. Cement production and air quality impacts. Sci Total Environ. 2020.
- Mohamad N, et al. Environmental impacts of cement manufacturing. Mater Today Proc. 2021.
- ACI Committee 211. Standard practice for selecting proportions for normal, heavyweight, and mass concrete. Farmington Hills (MI): American Concrete Institute; 1991.
- Kodur V. Behavior of concrete under elevated temperatures. ISRN Civ Eng. 2014.
- Gepulango G, Grio M. Compressive strength variation of concrete. Patubas J. 2003.
- IBM Corporation. Introduction to machine-learning algorithms [Internet]. 2024. Available from: https://www.ibm.com/
- Marcos C, et al. Predictive modeling for urban conditions using linear and nonlinear ML. In: Proceedings of IEEE International Conference on Recent Advances and Innovations in Engineering (ICRAIE); 2023.
- Silva D, Marcos C. Digital data collection in civil engineering. In: Proceedings of IEEE International Conference on Modeling, Simulation and Processing (ICMSP); 2023.
- Marcos C, Silva D. ANN-based prediction of structural behavior. In: Proceedings of IEEE Smart Computing and Materials Conference (SCM); 2022.
- Gao J, et al. Mix design of sustainable concrete using generative models. Constr Mater Sci. 2024.
- Tipu RK, et al. Optimizing sustainable blended concrete mixes using deep neural networks. Sci Rep. 2025.
- Erfani A, et al. Applications of multimodal large language models in construction [Internet]. SSRN. 2025. Available from: https://ssrn.com/
- Innovative concrete mix design incorporating recycled aggregates. Int J Civ Eng. 2025.
- Review paper on concrete mix design optimization using ML-based algorithm. J Prog Concr. 2024.
- Investigation study data to develop sustainable concrete mix using alternative aggregates. Nat Sci Data. 2023.
- Machine learning algorithms for optimizing mix design of recycled aggregate concrete. Migr Lett J. 2025.
Recent Trends in Civil Engineering & Technology
| Volume | 16 | |
| Issue | 02 | |
| Received | 10/02/2026 | |
| Accepted | 13/05/2026 | |
| Published | 14/05/2026 | |
| Publication Time | 93 Days |