Table 1 A summary of recent studies on ML-based prediction of concrete mechanical characteristics.
S.No | Author | ML technique applied | Material | Target properties | Main findings |
|---|---|---|---|---|---|
1 | Yang et al.51 | Genetic algorithm–random forest (GA-RF), Single, Integrated, Neural network | Geopolymer concrete (GPC) | Compressive Strength at 20–1000 °C | The hybrid GA-RF model has the best accuracy. |
2 | Manan et al.40 | Gradient Boosting (GB) and Random Forest (RF) | Recycled concrete powder (RCP) as SCM | Compressive strength | The RF model’s R2 > 0.95 and life cycle assessment confirmed the sustainability benefits of RCP. |
3 | Manan et al.41 | XGBoost, SVR, ANN | Recycled Aggregate Concrete (RAC) | Compressive, flexural and tensile strength | All strength parameters were simultaneously accurately predicted by multi-output machine learning (R2 = 0.96). |
4 | Chen et al.46 | GEP and MEP | Waste foundry sand concrete | Split tensile strength and elastic modulus (E) | Both GEP and MEP correctly predicted E (R ≈ 0.99), with GEP outperforming MEP for STS. |
5 | Manan et al.42 | RF and MEP | Carbon nanotubes (CNT) reinforced concrete | Compressive strength | MEP’s higher predictive performance (R2 = 0.99) was confirmed through experimentation. |
6 | Farooq et al.50 | RF and GEP | High strength concrete | Compressive strength | RF achieved best accuracy (R2 = 0.96), outperforming GEP |
7 | Manan et al.52 | ANN, XGBoost, RF | FRP-reinforced concrete beams | Flexural strength | RF produced the best results (R2 = 0.98); complex analytical models were successfully replaced by ML. |
8 | Inqiad et al.47 | MEP and GEP | Fibre reinforced self compacting concrete (RF-SCC) | Compressive strength | GEP showed higher accuracy (OF = 0.029) and simpler equations than MEP. |
9 | Manan et al.53 | RF, ANN, LightGBM, | RCP-reinforced concrete and Steel fiber | Axial load–deflection response | High precision and comprehensible constitutive simulation for sustainable concrete design were attained using hybrid machine learning. |
10 | Khan et al.48 | MEP | Bentonite plastic concrete (BPC) | Compressive strength (CS), elastic modulus (E) and Slump, | High accuracy (R = 0.9999 slump, 0.9831 CS, 0.93 E). Water, cement, bentonite dominate slump; curing time and cement affect E. |
11 | Manan et al.54 | ANN, Genetic Algorithm (GA) | High performace concrete | Compressive and tensile strength | ANN-GA hybrid optimized SCM proportions, reducing embodied carbon and increasing strength. |
12 | Manan et al.54 | LCA with emergy-based assessment | Recycled concrete powder based concrete | - | Shown a 25% decrease in environmental energy; the usage of RCP promotes the circular economy |
