Intelligent prediction of tensile splitting strength of rubberised concrete containing brick powder: Unfinished path towards sustainability

Using waste materials when producing concrete has aroused much interest in many construction applications. In particular, brick powder (BP) and tire rubber (TR) applications in concrete have seen a revival of interest because their properties have opened the doors for sustainable construction, which could not be addressed previously. Although machine learning models have been extensively applied in concrete production, there is a shortage of studies on deep learning (DL) models for predicting the split tensile strength of concrete containing BP and TR. This gap is addressed using deep interest network (DIN) Transformer, feature tokenizer (FT) Transformer, convolutional neural network (CNN), neural oblivious decision ensemble (NODE) models applied to the tensile splitting strength prediction of concrete containing BP and TR. The study also integrates SHapley Additive exPlanation (SHAP), which relies on NODE model while model generalisations are captured using 5-fold cross-validation analysis. The findings demonstrate that NODE model achieves highest performance during training and testing phases, achieving the coefficient of determination of 0.996 for training dataset and 0.97 for testing dataset. Further predictive capabilities are provided through the error metrics which confirm the NODE model’s superior training precision (RMSE = 0.0349 MPa and MSE = 0.0012) whilst also maintaining outstanding testing behaviour (RMSE = 0.0652 MPa and MSE = 0.0043). In SHAP analysis, age substantially contributes to feature importance, while coarse aggregate and cement also demonstrate their considerable contributions. Cross-validation results demonstrate strong generalisation (gaps <0.02 for folds 1 and 2), good generalisation (gaps between 0.02 and 0.05 for folds 4 and 5) and moderate overfitting (gap between 0.05 and 0.1 for fold 3). Consequently, the solutions regarding the prediction of concrete with TR and BP align with the established sustainable solutions, both in situations of environmental cleanup and innovative concrete production.

Authors

Institutions

Publication Details

Journal
Next Materials
Published
2026-11-21
DOI
https://doi.org/10.1016/j.nxmate.2026.104096
Primary Topic
Innovative concrete reinforcement materials
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Intelligent prediction of tensile splitting strength of rubberised concrete containing brick powder: Unfinished path towards sustainability

Destine Mashava, Derrick Mirindi, David Sinkhonde, Tajebe Bezabih et al.
Next Materials
Innovative concrete reinforcement materials
article

Intelligent prediction of tensile splitting strength of rubberised concrete containing brick powder: Unfinished path towards sustainability

Destine Mashava, Derrick Mirindi, David Sinkhonde, Tajebe Bezabih, Frédéric Mirindi
article en

Abstract

Using waste materials when producing concrete has aroused much interest in many construction applications. In particular, brick powder (BP) and tire rubber (TR) applications in concrete have seen a revival of interest because their properties have opened the doors for sustainable construction, which could not be addressed previously. Although machine learning models have been extensively applied in concrete production, there is a shortage of studies on deep learning (DL) models for predicting the split tensile strength of concrete containing BP and TR. This gap is addressed using deep interest network (DIN) Transformer, feature tokenizer (FT) Transformer, convolutional neural network (CNN), neural oblivious decision ensemble (NODE) models applied to the tensile splitting strength prediction of concrete containing BP and TR. The study also integrates SHapley Additive exPlanation (SHAP), which relies on NODE model while model generalisations are captured using 5-fold cross-validation analysis. The findings demonstrate that NODE model achieves highest performance during training and testing phases, achieving the coefficient of determination of 0.996 for training dataset and 0.97 for testing dataset. Further predictive capabilities are provided through the error metrics which confirm the NODE model’s superior training precision (RMSE = 0.0349 MPa and MSE = 0.0012) whilst also maintaining outstanding testing behaviour (RMSE = 0.0652 MPa and MSE = 0.0043). In SHAP analysis, age substantially contributes to feature importance, while coarse aggregate and cement also demonstrate their considerable contributions. Cross-validation results demonstrate strong generalisation (gaps <0.02 for folds 1 and 2), good generalisation (gaps between 0.02 and 0.05 for folds 4 and 5) and moderate overfitting (gap between 0.05 and 0.1 for fold 3). Consequently, the solutions regarding the prediction of concrete with TR and BP align with the established sustainable solutions, both in situations of environmental cleanup and innovative concrete production.

Next MaterialsVol. 14
Jomo Kenyatta University of Agriculture and Technology (KE), African Economic Research Consortium (KE), Pan African University Institute for Basic Sciences, Technology and Innovation (KE)
Innovative concrete reinforcement materials
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon