Akademik Veri Yönetim Sistemi
COMPUTERS AND CONCRETE, cilt.34, sa.6, ss.1111, 2024 (SCI-Expanded)
Determination of the mechanical properties of mortar at elevated temperatures is of great importance for fire safety and long-term durability of structures in civil engineering. In this study, experiments were conducted to determine the properties of mortar using four different chemical admixtures. The aim was to predict the mechanical property changes induced by chemical admixtures on mortar at very high temperature. In order to reach the most realistic predictions, the data was analyzed by testing well-known machine learning methods. The Multilayer Perceptron (MLP) method showed the best performance and was compared with Ordinary Least Squares (OLS) regression, known as the benchmark for prediction. The results showed that MLP is a very suitable predictor. At 300 °C the tensile strength decreased to 1 MPa and at 600 °C all admixture ratios resulted in compressive strengths below 5 MPa, indicating severe degradation. The innovative aspect of this study is the successful prediction of the mechanical properties of mortar in fire scenarios, through applying machine learning methods that use the results of experiments up to 600 °C, and allowing the use of data obtained from experimentation that is not costly, time consuming and conducted at high temperatures. To produce data, the specimens were prepared in the laboratory using four specially selected types of chemicals in various proportions. This careful selection and varied ratio application represent another supportive innovation of the study. This study provides important and instructive results for both future research and practical applications in the field of building safety.