Prediction of compressive strength of fiber-reinforced concrete containing silica (SiO) based on metaheuristic optimization algorithms and machine learning techniques.

Concrete compressive strength (CS) is crucial for ensuring the safety, durability, and performance of structures. So, its precise simulation helps anticipate material behavior under various conditions. Despite a comprehensive experimental investigation of the impact of silica (SiO) on the CS of the fiber-reinforced concrete, its mathematical aspects were not well studied. So, this study integrates the ANFIS (adaptive neuro-fuzzy inference system) and ELM (extreme learning machine) machine learning models with three optimization algorithms, i.e., WCA (water cycle algorithm), PSO (particle swarm optimization), and GWO (grey wolf optimizer) to precisely estimate the CS of fiber-reinforced concrete (FRC) containing SiO. An experimental database comprising 228 datasets is used to develop the models, compare their accuracy, and select the best one. The database contains information on the volumetric percentage of fibers, sample age, amount of coarse/fine aggregates, water, cement, nano silica, and binder as independent features, while the compressive strength is the target variable. The sensitivity assessment approves that the training and generalization abilities of the ELM and ANFIS models for the CS prediction of FRC are improved by their integration with the GWO algorithm. The best model (i.e., ELM-GWO) predicts the testing datasets with the R (coefficient of determination), RMSE (root mean square error), SI (scatter index), RPD (relative percent deviation), and PMARE (percent mean absolute relative error) values of 0.9510, 3.985 MPa, 0.061, 0.8, and 5.421, respectively.