Estimating the Concrete Ultimate Strength Using a Hybridized Neural Machine Learning

Abstract

Concrete is a highly regarded construction material due to many advantages such as versatility, durability, fire resistance, and strength. Hence, having a prediction of the compressive strength of concrete (CSC) can be highly beneficial. The new generation of machine learning models has provided capable solutions to concrete-related simulations. This paper deals with predicting the CSC using a novel metaheuristic search scheme, namely the slime mold algorithm (SMA). The SMA retrofits an artificial neural network (ANN) to predict the CSC by incorporating the effect of mixture ingredients and curing age. The optimal configuration of the algorithm trained the ANN by taking the information of 824 specimens. The measured root mean square error (RMSE = 7.3831) and the Pearson correlation coefficient (R = 0.8937) indicated the excellent capability of the SMA in the assigned task. The same accuracy indicators (i.e., the RMSE of 8.1321 and R = 0.8902) revealed the competency of the developed SMA-ANN in predicting the CSC for 206 stranger specimens. In addition, the used method outperformed two benchmark algorithms of Henry gas solubility optimization (HGSO) and Harris hawks optimization (HHO) in both training and testing phases. The findings of this research pointed out the applicability of the SMA-ANN as a new substitute to burdensome laboratory tests for CSC estimation. Moreover, the provided solution is compared to some previous studies, and it is shown that the SMA-ANN enjoys higher accuracy. Therefore, an explicit mathematical formula is developed from this model to provide a convenient CSC predictive formula.