Electrical resistivity prediction model for basalt fibre reinforced concrete: hybrid machine learning model and experimental validation

Abstract

The application of basalt fibre reinforced concrete (BFRC) is crucial for reducing carbon emissions, enhancing structural performance, and extending service life. Electrical resistivity (ER), a non-destructive testing indicator, can be used to evaluate parameters such as compressive strength and chloride ion permeability of concrete. Therefore, this study examines BFRC-ER from three perspectives: the applicability of existing ER prediction models, hybrid machine learning modelling, and experimental validation. The findings indicate that the predicted values of the existing nine models have a poor correlation with actual values, limiting their practical application. The prairie dog–optimised XGBoost (PDO–XGBoost) model developed in this study exhibited closer alignment between predicted and actual values. It boasted smaller mean and standard deviation (μ = 0.0508 kΩ·cm, σ = 3.409) of model error distribution, along with superior performance evaluation metrics (MAE = 2.165, MAPE = 0.243, RMSE = 3.410 MSE = 11.625, and R² = 0.984). Analysing the contribution of each input feature to BFRC-ER revealed that saturation, age, and water–binder ratio are the three significant influencing factors. Moreover, this study developed a graphical user interface (GUI) for BFRC-ER, enabling the visualisation of BFRC-ER predictions. Subsequently, BFRC with varying mix proportions was prepared, and BFRC-ER was tested using the two-electrode method. The comparison between actual values and GUI predictions showed errors below 7.5%, highlighting the accuracy of the predictions. This research achieves high-accuracy predictions of BFRC-ER, laying the foundation for optimising BFRC mix proportions and evaluating concrete performance.