Chemistry-informed data-driven models for predicting strength of ambient-cured slag and fly ash geopolymer concrete

Machine learning-driven prediction of compressive strength in sustainable geopolymer concrete using artificial neural network and random forest models

ABSTRACT Geopolymer concrete is a sustainable substitute for ordinary Portland cement which minimizes carbon dioxide emissions and effectively utilizes the waste from industries. Proper predictive estimating compressive strength can assist in the mix deign optimization, structural reliability. The article presents a machine learning-based framework to predict the compressive strength of geopolymer concrete made with multiple industrial by-products as binders. This study investigated the subsequent strength of concrete when subjected to fly ash, ground granulated blast furnace slag, metakaolin, silica fume, and rice husk ash. A database was developed containing 243 experimentally prepared samples with different mix proportions. The study conducted the compressive strength prediction by implementing Artificial Neural Network (ANN) and Random Forest (RF) models in Python. The performance of model was analyzed through the coefficient of determination (R2) and mean absolute error (MAE) and root mean square error (RMSE). The RF model was found to be superior to the ANN model with R2 = 0.97, MAE = 1.9969, RMSE = 3.0586, which was an accurate result whereas ANN model was lower accurate R2 = 0.78. The results show that techniques using ensemble learning can capture complex non-linear relationships, reduce experimental efforts and assist in developing efficient and sustainable geopolymer concrete mix designs.

Experimental and ANN-based prediction of the mechanical properties of fly ash geopolymer concrete

A material-centered predictive framework for strength prediction in fly ash–based cementitious composites using an EAO-ELM hybrid model

This study proposes a material-centered predictive framework for fly ash (FA)–based cementitious composites by integrating mixture composition, early-age mechanical properties, and microstructural parameters to model the evolution of compressive strength (CS). A bio-inspired hybrid regression approach, the Enzyme Action Optimizer–enhanced Extreme Learning Machine (EAO-ELM), is introduced to capture the nonlinear relationships governing strength development. In the proposed hybrid model, the random assignment of input–hidden layer weights and biases—the principal source of prediction variance in conventional ELM—is replaced by a population-based global search that minimizes the training loss directly. This improves the conditioning of the hidden-layer output matrix and stabilizes the generalization capability of the model, which explains the performance advantage of EAO-ELM over conventional ELM-based and ensemble learning approaches. Beyond serving as a computational technique, the proposed model provides insight into material behavior and supports performance-driven mixture design. To the best of the authors’ knowledge, this work represents the first application of the EAO to cementitious materials. To systematically evaluate how material-related information influences predictive accuracy, three scenario-based information schemes were constructed. These scenarios enable a comprehensive assessment of the contribution of early-age mechanical properties to long-term strength estimation. Across all scenarios, the EAO-ELM model consistently outperformed classical and ensemble regression methods, achieving an improvement of approximately 36% over the strongest model (Random Forest) in the full-information case. Interpretability analyses using SHapley Additive exPlanations (SHAP) and Partial Dependence Plot further revealed that 7 d CS, cement content, and FA ratio exert the greatest influence on strength prediction, whereas water absorption and ultrasonic pulse velocity (UPV) play secondary roles. Overall, the proposed framework offers a accurate, interpretable, and adaptive means of understanding strength development in sustainable cementitious systems and provides subject to external validation on independent datasets, a promising decision-support aid for optimizing mixture design in construction materials engineering.

Predicting the compressive strength of sustainable concrete with recycled foundry sand using advanced soft computing methods.

Given the environmental challenges posed by the production and disposal of industrial waste, reusing such materials in the construction industry, especially for the development of sustainable concrete, offers an eco-friendly solution and cost reduction. This study investigates the use of waste foundry sand (WFS) as a partial replacement for fine aggregates in concrete. To accurately predict the compressive strength (fc) of WFS-containing concrete, a comparative modeling framework was employed by using one traditional statistical method, Response Surface Methodology (RSM), alongside two advanced soft computing techniques, namely Group Method of Data Handling (GMDH) and Gene Expression Programming (GEP). A dataset consisting of 397 laboratory samples, including various mix design parameters and curing ages as input variables, with fc as the output, was utilized to train and evaluate the models. The results indicate that the RSM model showed the best predictive performance. The fitted model achieved RMSE and MAE values of 4.289 MPa and 3.583 MPa, respectively. Under LOOCV validation, the corresponding errors were RMSECV = 5.40 MPa and MAECV = 4.19 MPa, indicating good generalization capability and stable prediction of compressive strength for concrete containing WFS. The correlation coefficient (R = 0.83) is reported as a secondary performance indicator, indicating a moderate level of agreement between predicted and experimental values. Additionally, sensitivity analysis of input variables indicated that the water-to-cement ratio and superplasticizer-to-cement ratio had the greatest impact on fc, while the WFS-to-cement ratio (WFS/C) and the WFS-to-fine aggregate ratio (WFS/FA) showed a relatively lower influence.

Bayesian-optimised machine learning for predicting aggressive environment resistance and service life of recycled aggregate geopolymer concrete

Developing reliable computational tools for durability and service-life assessment of concrete structures in aggressive environments is essential for advancing predictive modeling in structural engineering. This study introduces machine learning (ML)–based models for forecasting the sulfate and acid resistance of recycled aggregate geopolymer concrete (RGPC), produced with untreated and surface-treated recycled concrete aggregates (RCAs) through two mixing approaches. Three algorithms, i.e. Gaussian process regression (GPR), LSBoost ensemble, and Neural Network, were trained using nine input parameters related to material composition and exposure conditions, with durability indicators, namely mass loss rate (Kw) and compressive strength retention index (Kf), as outputs. A dataset of 336 experimentally tested RGPC specimens was used, applying Bayesian Optimisation for hyperparameter tuning and 5-fold cross-validation for generalisation. Among the models, the optimized GPR achieved the highest accuracy, confirmed by the lowest objective value. Feature importance analysis highlighted environmental cations, sulfate concentration, RCA replacement level, initial compressive strength, and exposure duration as the most influential factors governing degradation. The proposed Bayesian-optimized ML framework demonstrates a robust and generalizable method for predicting durability and service life of sustainable concretes, providing a valuable tool for simulation-driven design and durability-based performance assessment in mechanics and structural engineering.

Interpretable Machine Learning for One-Part Fly-Ash/Slag Geopolymer Strength Prediction: Toward Multifunctional Binder Design

Portland cement production accounts for roughly 8% of anthropogenic CO2 emissions, driving interest in low-carbon geopolymer binders. One-part (“just-add-water”) geopolymers, which replace hazardous liquid activators with a dry, pre-blended solid activator, are especially suited to field deployment where handling safety and logistics are decisive. However, their formulation space is combinatorially vast, and trial-and-error development cannot efficiently navigate it. This paper reviews one-part geopolymer science, presents a new comparative and interpretable ML analysis of a published 80-mixture one-part fly-ash/ground granulated blast-furnace slag (GGBS, hereafter slag) geopolymer dataset from twelve studies, and proposes an AI-assisted design framework. The ML demonstration targets 28-day compressive strength only. Under leave-one-source-out (LOSO) cross-validation—the appropriate test for a literature-pooled dataset—gradient-boosted trees achieved R2 = 0.61 (RMSE = 15.5 MPa; 95% bootstrap confidence interval on R2, 0.44–0.75), well above a linear baseline (0.36), suggesting that non-linear structure transfers across studies; a random split gives a higher but less reliable R2 = 0.90 on only 16 test mixtures. Because fly-ash and slag contents are near-perfectly anti-correlated (r=−0.99), we model the precursor axis as a single slag fraction descriptor; SHAP then identifies this precursor balance and the activator’s Na2O dosage as the dominant statistical predictors of strength in this dataset, an ordering consistent with known activation chemistry; causal confirmation of these associations awaits the experimental validation stage of the proposed framework. Demonstrated for strength only, at paste level, the framework offers a transferable route toward multifunctional low-carbon binders for protective and infrastructure applications; the multifunctional extensions are proposed, but not yet demonstrated.