%0 Journal Article
%T Modeling Compressive Strength of Roller Compacted Concrete Pavement Using Artificial Neural Network, ANFIS and Support Vector Machine
%J Journal of Transportation Infrastructure Engineering
%I Semnan University Press
%Z 2423-5350
%A emadi, mohammad
%A Madani, Seyed Hesam
%D 2017
%\ 11/22/2017
%V 3
%N 3
%P 55-78
%! Modeling Compressive Strength of Roller Compacted Concrete Pavement Using Artificial Neural Network, ANFIS and Support Vector Machine
%K Pavement
%K modeling
%K Artificial intelligence
%K Adaptive neuro fuzzy inference system
%R 10.22075/jtie.2017.1823.1176
%X Nowadays, roller compacted concrete (RCC) is used in building dams and roads, and in recent years, using RCC is extended because of its advantages such as short construction time, availability of required materials, appropriate performance in cold regions and high life span. Compressive strength of RCC is the most important mechanical property that its enhancement can improve RCC performance. The sensitivity of RCC to its ingredients has caused some problems in prediction of the compressive strength. Parameters such as cement content, water-cement ratio, the amount of replaced cementitious materials and coarse to fine aggregates ratio affect the compressive strength of RCC. In recent decades, modeling by artificial intelligence has found a special place in technical sciences and engineering, and prediction of the behavior of complex cases has become possible with the help of this method. In this study, constructed design mixes by the authors and design mixes made in other studies were collected. By considering concrete constituents and samples age as input variables, several artificial neural network (ANN), adaptive neuro fuzzy inference system (ANFIS) and support vector machine (SVM) models were prepared to predict the compressive strength of RCC. Comparison of the results indicated that ANN model has more ability in predicting the compressive strength of RCC than ANFIS and SVM models. Also, the predicted compressive strength by ANN and SVM models had the highest and lowest match with actual compressive strength, respectively. The correlation coefficient, root mean square error and mean absolute error of ANN model were 0.9717, 2.4859 and 2.1396, respectively. These values were 0.9566, 3.4013 and 3.0733 for SVM model, respectively.
%U https://jtie.semnan.ac.ir/article_2698_0292575b321ad054b3b5644f2c6c5ac0.pdf