Evaluation of parametric study of mixture proportions on the tensile strength regression-based model of concrete containing recycled construction waste

Restricting resources and maintaining environmental quality has inevitably required the recycling of materials. Concrete is one of the most attractive building materials for recycling. The waste concrete can be crushed again, used as aggregate in the manufacture of concrete. In this study, for the modeling of compressive strength of concrete containing recycled aggregate, two formula-based regression methods named multivariate adaptive regression splines (MARS) was used. Mixing data from the research background to create suggested models, 239 laboratory data was compiled to estimate tensile strength of recycled concrete for parametric investigation. Then, two scenarios based on volumetric/weighted and ratio variables were defined to determine the best input parameters for the model using mallow’s technique. Among which a scenario including volumetric/weighted content, based on three statistical error indices including correlation coefficient (R), Root mean square error (RMSE) and mean absolute error (MAE) were selected as the best scenario. In this study, to determine the best model for estimating the compressive strength of recycled concrete by means of error statistics, it was determined that the correlation coefficient (R) in the training stage for MARS 0.973 and 0.903. Also. The RMSE statistical value for the proposed model MARS this stage was 16.176. In the testing phase, the MARS method was better than the M5p method. The results of uncertainty analysis indicated that the proposed method had a predicted mean error with very little distance in the semester. In addition, among the proposed smart methods, the MARS base on volumetric/weighted content approach was chosen for parametric analysis.