|
Application of ensemble machine learning techniques in civil engineering materials |
|---|---|
| รหัสดีโอไอ | |
| Title | Application of ensemble machine learning techniques in civil engineering materials |
| Creator | Irwan Afriadi |
| Contributor | Chanachai Thongchom, Advisor |
| Publisher | Thammasat University |
| Publication Year | 2567 |
| Keyword | FRP, Machine learning, Random forest, XGBoost, CatBoost, AdaBoost |
| Abstract | The deterioration of reinforced concrete (RC) structures is primarily caused by the corrosion of steel reinforcement, significantly increasing maintenance and repair costs, while fire exposure further accelerates structural damage. Asa result, fiber-reinforced polymer (FRP) is a promising non-corrosive substitute for conventional steel reinforcement in RC construction. This study applies ensemble machine learning techniques, including XGBoost, AdaBoost, CatBoost, and Random Forest, to predict critical material properties in civil engineering. The models were used to estimate ultimate strength in FRP-concrete pull-out tests, ultimate shear stress at the FRP-steel interface, and residual steel strength properties post-fire exposure. In rank analysis, XGBoost is considered as the best technique in forecasting ultimate capacity of concrete-FRP pull-out test. In other hand, Adaboost is identified as the best-performing model in forecasting the ultimate shear capacity of FRP-steel test and steel residual post-fire mechanical properties test. Analysis of the feature importance revealed that the most influential parameters is bond length (lb) for concrete-FRP pull-out test, CFRP thickness (tC) for FRP-steel test, and temperature for post-fire steel properties. The findings highlight the effectiveness of machine learning in predicting material behavior. |
