Estimation of CS of the HPC with hybrid RBF neural networks

This study used a radial function neural network (RBFNN) to create a novel system for calculating high-performance concrete's (HPC) compressive strength (CS) modified with fly ash and blast furnace slag. These admixtures could affect the mechanical and physical properties of HPC, and determining it definitely requires experimental efforts and costs. Herein, alternative methods such as machine learning algorithms named RBFNN could be useful to address these questions. The SSA (Salp swarm algorithm) and the artificial hummingbird algorithm (AHA) were utilized in this work to find optimal values of hyperparameters of the RBFNN approach that can be tuned. The suggested models were assessed utilizing a comprehensive dataset including 1030 data rows. Finally, the findings were compared to those documented in the literature. The findings of the calculations, which took into account evaluation metrics, depict that both hybrid SSA-RBFNN and AHA-RBFNN analysis might astonishingly perform good productivity during estimating, with R 2 values of 0. 8955 and 0.8608 for SSA-RBFNN and 0.8987 and 0.8643 for AHA-RBFNN, respectively, related to the test and train segments. In conclusion, the AHA-RBFNN model created for predicting the CS of HPC amended with BFS and FA could be identified as the proposed model to be applied in practical applications.