Abstract

Bond strength between fly ash-based concrete (FGPC) and rebars is a vital index for ensuring the reliability and safety of reinforced concrete members. Thus, prediction and analysis of ultimate bond strength between FGPC and rebars are performed based on machine learning data-driven methods in this paper. A latest database including 137 samples from the published papers in the last decade is established. Four machine learning algorithms that include Lasso regression, Support Vector Regression (SVR), Random Forest (RF) and Extreme Gradient Boosting (XGB) are adopt for the construction of prediction models, and model evaluations are carried out by using four metrics (R2, RMSE, MAPE and MRE). The comparison between machine learning models and empirical models based on experiments is finished. Besides that, the feature importance is analyzed based on RF and XGB models. The results show that the machine learning models have a better prediction performance than the empirical models, where the RF and XGB models are optimal estimators with R2 of over 0.84 on testing set. Lasso and SVR models possess stronger generalization ability than RF and XGB models. The relatively important influence factors are compressive strength of FGPC (fc), reactivity modulus of precursors (RM), the ratio of cover thickness to diameter of rebar (c/d) and the ratio of anchorage length to diameter of rebar (l/d). Increasing slag content, fc, RM, and c/d are beneficial to enhancement of bond strength. The recommended values of fc, RM, c/d and l/d are 50 MPa, 0.7, 4.5 and 5.0 respectively.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call