Performance of ANN, Random Forest and XGBoost methods in predicting the flexural properties of wood beams reinforced with carbon-FRP

Wooden material can be used in different areas due to its various positive properties. Glued Laminated wooden elements (glulam) are wood composite materials widely used especially in the construction industry. Carbon fiber-reinforced polymers (FRP) are widely used to increase the bearing capacity values of glulam beams and improve their overall load-displacement behavior. This study was carried out in two stages. In the first stage, the bending properties of glulam timbers of different sizes with wide spans reinforced with carbon fiber-reinforced polymers were experimentally examined. In the next stage, the obtained data were predicted with three different machine learning techniques (ANN, Random Forest and XGBoost). As a result of the study, it was determined that as the section dimensions increased, the bending properties increased, and the reinforcement was effective by approximately 22%. All three different prediction techniques used could make predictions with high accuracy. However, it was determined that the best prediction was made with Random Forest (R2: 0.9892). Therefore, the bending properties of reinforced beams of different sizes can be predicted with machine learning (ML) techniques.