Özet:

In the field of Civil Engineering, machine learning methods have been used frequently in order to understand the mixture design and to predict some properties of building materials, especially concrete. Machine learning methods, which are very useful in this context, can be said to be innumerable. In this study, Gaussian Process Regression (GPR) and Support Vector Machines (SVM), which are two types of machine learning methods, were preferred to estimate the compressive strength of Self Compacting Concrete (SCC). The aim of the study is to prove the success of different machine learning methods in predicting concrete performance and thus to redound the usage of the methods, especially for concrete mix design. For this purpose, GPR and SVM models were developed with the experimental data set containing the SCC mix composition and properties. The performances of the developed models were compared both with each other and with another machine learning method, Artificial Neural Network, which has proven its success with numerous studies in the literature. As a result, it was revealed that the GPR and SVM models, trained and validated with an experimental dataset, were successful in predicting the compressive strength of SCC. In addition, GSR has been the most successful method in this problem. The correlation coefficients between the experimental data and the output of the GSR model were calculated as 0.9888 in the training state and 0.8648 in the testing state.

Anahtar Kelimeler: