Comparison of the Results of Linear Regression Models and Artificial Neural Network to Predict the Permeability of Sponge Concrete Pavement

Foam concrete pavement can be used as a suitable alternative to other pavements in light urban traffic. In order to develop the use of this type of pavement, recognizing its features is necessary and important. Since the permeability of spongy concrete pavement is the most important functional feature of this type of pavement, further understanding of this feature and how it is affected by the parameters of the mixing design is the aim of this study. For this purpose, suitable combinations of foam concrete samples with granulation and different water-to-cement ratios of 36 samples have been made and tested. The ratio of water to cement for making experimental samples in the range of 0.28 to 0.34 in combination with aggregates with maximum nominal size of 9.5 mm, 12.5 mm and 19.5 mm has been selected. In order to select the appropriate model for predicting permeability changes, a comparison has been made between artificial neural network techniques and linear regression. Using the data obtained from laboratory activities and examining the fit of the models, the optimal model is proposed. Comparison of models showed that linear regression produced closer results in predicting changes in the permeability of sponge concrete. The use of linear regression can reduce the number of test specimens to achieve the optimal mixing design for foam concrete.
.