Prediction the Chloride Ion Permeation Coefficient of Concrete Based on A Hybrid Intelligent Algorithm

Abstract

Chloride ion penetration resistance (CIPR) is a critical concern in engineering to ensure the long-term durability of concrete structures, accurately predicting concrete CIPR is essential for designing the appropriate mix ratio. The rapid chloride migration (RCM) test is the most commonly used experimental method, typically employed to measure CIPC. To efficiently and accurately predict the CIPR of concrete, a Bayesian Optimization (BO)-Light Gradient Boosting Machine (LGBM) model is developed. Through this research, it can be concluded that (1) BO can effectively search and optimize the hyperparameters in LGBM. Within 100 iterations, BO optimization can search the hyperparameters effectively and find the optimal solution quickly.(2) BO-LGBM has a strong predictive ability, and its prediction accuracy is superior than the other three prediction models. The outcomes indicate that the application of this model has important practical significance for predicting the CIPC of concrete, optimizing the design of the concrete mix ratio and improving the durability of concrete.