1 Introduction
2 Research Significance
3 Methodology
3.1 Materials and Data Collection
Cement (kg/m3) | Water (W/C) | Coarse aggregates | Fine aggregates | Admixture (ml/100 kg) | Total height (m) | Casting rate (m/h) | Maximum pressure (kPa) | |
---|---|---|---|---|---|---|---|---|
Maximum | 500 | 1 | 2400 | 1200 | 10 | 17 | 71 | 186 |
Minimum | 182 | 0 | 305 | 379 | 0 | 0 | 0 | 2 |
Mean | 388 | 1 | 1008 | 799 | 1 | 3 | 7 | 48 |
Standard deviation | 82 | 0 | 459 | 182 | 2 | 3 | 15 | 37 |
Kurtosis | 0 | 3 | 4 | 0 | 7 | 7 | 14 | 3 |
Skewness | 0 | 2 | 2 | 0 | 3 | 2 | 4 | 2 |
3.1.1 Mix Composition of Concrete
3.1.2 Fresh Properties of Concrete
3.2 Support Vector Regression
3.3 Artificial Neural Networks
3.4 Optimization Algorithms
3.4.1 Genetic Algorithm
3.4.2 Salp Swarm Algorithm
3.4.3 Grasshopper Optimization Algorithm
3.5 Proposed Models
3.5.1 Modified ANN
3.5.2 Modified SVR
4 Results and Discussion
4.1 Normalization
4.2 Comparison of the Models’ Performances
Models | MAE (MPa) | MBE (MPa) | RMSE (MPa) | MAPE (%) | SI | R |
---|---|---|---|---|---|---|
ANNGA | 3.01 | − 0.28 | 6.31 | 12.62 | 0.13 | 0.984 |
ANNSSA | 4.08 | − 0.32 | 7.24 | 18.22 | 0.15 | 0.979 |
ANNGOA | 3.60 | 0.25 | 6.2 | 14.66 | 0.14 | 0.983 |
SVRSSA | 2.60 | − 0.06 | 6.16 | 8.20 | 0.13 | 0.985 |
SVRGOA | 3.05 | − 0.07 | 6.12 | 11.59 | 0.13 | 0.985 |
4.3 Validation with Mathematical Modelling
5 Conclusion
-
Following ACI 347-04 Guide to Formwork for Concrete, it can be said that the lateral pressure exerted concrete in the real-world application did not record a value above 200 kPa. However, this mostly depends on the height of casting, casting rate, the constant value of gravity and the density of the materials; therefore, it could be expected to increase with increasing these parameters.
-
Generally, SVR-based models have better performances compared to ANN-based models, although all models have the same correlation coefficient approximately.
-
Based on MAE, MBE, and MAPE, SVRSSA is the most accurate model followed by SVRGOA closely. Nevertheless, SVRGOA has lower RMSE compared to SVRSSA.
-
All machine-learning-based models have a high correlation coefficient, which indicates the great correlation between experimental and predicted lateral pressure. Therefore, all of them can be used to estimate the lateral pressure of concrete.
-
Even ANNSSA which is the least accurate model has an acceptable performance with the RMSE value of 7.24 and MAE value of 4.08.