Comparison of artificial neural networks and general linear model approaches for the analysis of abrasive wear of concrete

This study aims to determine the influence of metallic aggregate content, cement content and different loads applied on the abrasive wear of concrete by using artificial neural networks (ANN) and general linear model (GLM) approaches. For this purpose, experimental studies are made and suitable models based on experimental results are developed to estimate the abrasive wear of concrete. In these models, 60 data set was used. For training set, 48 data (80%) were randomly selected and the residual data (12 data, 20%) were selected as test set. Root mean square error (RMSE) and determination coefficient (R2) statistics are used as evaluation criteria of the ANN and GLM models and the experimental results are compared with these models. The comparison results indicate that the ANN models are superior to the GLM models in modeling of the influence metallic aggregate content, cement content and different loads applied on the abrasive wear of concrete.

► Hematite addition has increased abrasive wear of concrete.
► Developed models are reliable and accurate, and result in good agreement with experimental data.
► Parameters affecting wear of concrete have observed as hematite, cement and load.