Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

From activation energy to equivalent age: an effective machine-learning surrogate for fib-based strength prediction in concretes with supplementary cementitious materials

Aldawish, Abdulaziz, Kulasegaram, Sivakumar ORCID: https://orcid.org/0000-0002-9841-1339, Almutlaqah, Ayman and Alshahrani, Abdullah 2026. From activation energy to equivalent age: an effective machine-learning surrogate for fib-based strength prediction in concretes with supplementary cementitious materials. Case Studies in Construction Materials 25 , e06709. 10.1016/j.cscm.2026.e06709

[thumbnail of 1-s2.0-S2214509526009617-main.pdf] PDF - Published Version
Available under License Creative Commons Attribution.

Download (3MB)

Abstract

Maturity functions estimate concrete strength development from a recorded temperature history, with accuracy depending critically on the apparent activation energy. A recent extension of the fib Model Code maturity function allows this activation energy to vary with curing temperature and age, improving accuracy for concretes with supplementary cementitious materials, but its regression is calibrated on the mixes it predicts and its piecewise form is not portable to a different reference temperature. This paper presents a hybrid physics-based–machine-learning surrogate that retains the fib strength equation as the physics backbone and replaces the empirical activation-energy regression with an XGBoost model predicting the temperature-induced time-scaling factor log(te/t) directly. This target is physically equivalent to a variable activation energy; its advantage is not accuracy on the training domain, which a direct activation-energy surrogate matches, but that it is defined at every observation including the reference temperature and, in the test reported here, transfers from a 20 °C to a 30 °C reference temperature. The model was trained on 196 strength measurements from seven self-compacting concrete mixes covering Portland cement, silica fume, fly ash, ground granulated blast-furnace slag and rice husk ash, cured at 10–50 °C. Primary accuracy estimates use the 147 off-reference-temperature observations, excluding the 20 °C curves used for per-mix calibration. Under leave-one-mixture-out validation, the nested and fixed configurations both gave R2 = 0.943 and RMSE = 4.23 MPa; for the fixed configuration, the 95% mix-level bootstrap intervals were 0.899–0.969 for R2 and 3.28–5.21 MPa for RMSE. The modified fib gave R2 = 0.895 and RMSE = 5.74 MPa under the same validation scheme, compared with R2 = 0.922 and RMSE = 4.97 MPa for its published coefficients evaluated in-sample. Applied without retraining to independent literature data with a 30 °C reference temperature, spanning vibrated and self-compacting concretes (including one lightweight mix) and fly ash up to 45%, the model achieves R2 = 0.934 and RMSE = 5.43 MPa on the 119 off-reference observations, against 0.917 / 6.11 MPa for the original fib and 0.888 / 7.07 MPa for the modified fib; a surrogate trained on the activation-energy factor itself falls to 0.850 / 8.18 MPa on the same observations. SHAP attributions are consistent with the expected temperature and age dependence of hydration kinetics, and conformal intervals whose calibration residuals come from models fitted without the assessed mix give 94.6% pooled empirical coverage (nominal 95%; 81–100% by mix). Every mix, seen or unseen, still requires its reference-temperature strength–age curve, from which the two fib parameters are fitted; the model adds generalisation to unseen binders, demonstrated transfer from a 20 °C to a 30 °C reference temperature, and empirically calibrated uncertainty.

Item Type: Article
Date Type: Publication
Status: Published
Schools: Schools > Engineering
Publisher: Elsevier
ISSN: 2214-5095
Date of First Compliant Deposit: 17 November 2026
Date of Acceptance: 6 November 2026
Last Modified: 17 Nov 2026 11:00
URI: https://orca.cardiff.ac.uk/id/eprint/190994

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics