Efficient metaheuristic-retrofitted techniques for concrete slump simulation

Abstract

Due to the benefits of the early prediction of concrete slump, introducing an efficient model for this purpose is of great importance. Considering this motivation, four strong metaheuristic algorithms, namely electromagnetic field optimization (EFO), water cycle algorithm (WCA), teaching-learning-based optimization (TLBO), and multi-tracker optimization algorithm (MTOA) are used to supervise a neural predictive system in analyzing the slump pattern. This supervision protects the network against computational issues like pre-mature convergence. The overall results (e.g., Pearson correlation indicator larger than 0.839 and 0.807 for the training and testing data, respectively) revealed the competency of the proposed models. However, investigating the rankings of the models pointed out the superiority of the WCA (MAEtrain = 3.3080 vs. 3.7821, 3.5782, and 3.6851; and MAEtest = 3.8443 vs. 4.0326, 4.1417, and 4.0871 obtained for the EFO, TLBO, and MTOA, respectively). Moreover, the high efficiency of the EFO in terms of model complexity and convergence rate, as well as the adequate accuracy of prediction, demonstrated the suitability of the corresponding ensemble. Therefore, the neural systems trained by these two algorithms (i.e., the WCA and EFO) are efficient slump evaluative models and can give an optimal design of the concrete mixture for any desirable slump.