Predicting strength in polypropylene fiber reinforced rubberized concrete using symbolic regression AI techniques

Polypropylene fiber-reinforced rubberized concrete (PP-FRC) is emerging as a promising material for sustainable construction due to its improved mechanical behavior and environmental benefits, particularly through the inclusion of recycled crumb rubber and polypropylene fibers. However, estimating i...

Full description

Saved in:
Bibliographic Details
Published in:Case Studies in Construction Materials Vol. 23; p. e05024
Main Authors: Alawi Al-Naghi, Ahmed A., Aamir, Kinza, Amin, Muhammad Nasir, Iftikhar, Bawar, Mehmood, Kashif, Qadir, Muhammad Tahir
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.12.2025
Elsevier
Subjects:
Gene expression programming
Multi-expression programming
Polypropylene fiber
Rubberized concrete
Sensitivity analysis
SHAP analysis
Rubberized concrete
Sensitivity analysis
Multi-expression programming
SHAP analysis
Polypropylene fiber
Gene expression programming
ISSN:2214-5095, 2214-5095
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Polypropylene fiber-reinforced rubberized concrete (PP-FRC) is emerging as a promising material for sustainable construction due to its improved mechanical behavior and environmental benefits, particularly through the inclusion of recycled crumb rubber and polypropylene fibers. However, estimating its compressive strength (CS) accurately remains a challenge, primarily because of the nonlinear and interdependent nature of its constituent materials. Traditional machine learning methods have been explored for this purpose, yet many suffer from limited transparency and generalizability, making them less practical for engineering applications. In this study, two symbolic regression techniques, Gene Expression Programming (GEP) and Multi-Expression Programming (MEP), are applied to generate interpretable and precise equations for CS prediction of PP-FRC. An experimental dataset containing nine key input variables was used to train and validate the models. Among the two, the MEP model demonstrated stronger predictive capability, achieving a coefficient of determination of 0.90 and a mean absolute error of 3.83 MPa. Analysis using sensitivity techniques and Shapley Additive exPlanations values further identified superplasticizer, cement, and crumb rubber as the most critical variables influencing strength outcomes. This research contributes to the field by offering transparent, data-driven tools that aid in optimizing mix design while reducing the need for extensive laboratory trials. Overall, the models developed in this work support efficient material selection and promote environmentally responsible construction practices.
ISSN:2214-5095
2214-5095
DOI:10.1016/j.cscm.2025.e05024