Title: A comprehensive review of AI-driven techniques for modelling the mechanical and durability performance of high performance concrete

Authors: Gaurav P. Gohil; Indrajit N. Patel

Addresses: Department of Civil Engineering, Gujarat Technological University, Ahmedabad, Gujarat, 382424, India ' Charutar Vidya Mandal University, Vallabh Vidyanagar, 388120, Gujarat, India

Abstract: High-performance concrete (HPC) is widely adopted for its superior compressive, flexural, and durability characteristics, yet its complex mix design makes performance prediction challenging. This review synthesises recent advances in artificial intelligence (AI), machine learning (ML) and deep learning (DL) techniques applied to HPC property prediction. The analysed studies considered key input variables such as cement content, water-to-binder ratio, aggregate proportions, mineral admixtures (fly ash, silica fume, slag), superplasticisers, and curing conditions. Comparative analysis of different AI models revealed that support vector machine (SVM) and M5P model tree achieved the best predictive accuracy for compressive strength, with performance values up to R² = 0.9913 and RMSE = 2.158, significantly outperforming traditional regression methods (R² < 0.70). In durability predictions, gradient boosted regression trees (GBRT) and deep neural networks (DNN) demonstrated high reliability, reaching R² = 0.970 for carbonation depth and chloride penetration resistance. The novelty of this review lies in its integrated comparison of AI techniques for both mechanical and durability properties of HPC, highlighting the superiority of AI-driven approaches over conventional testing and pointing to future opportunities in hybrid models, explainable AI, and real-time monitoring to optimise next-generation HPC design.

Keywords: high performance concrete; machine learning; artificial intelligence; deep learning; durability; prediction.

DOI: 10.1504/IJSTRUCTE.2026.155427

International Journal of Structural Engineering, 2026 Vol.16 No.3, pp.285 - 317

Received: 20 Jun 2025
Accepted: 03 Dec 2025
Published online: 31 Jul 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article