Effects of loading mode on force-magnetic coupling and data-driven stress detection in rebars embedded in concrete

Abstract

Abstract

Accurate measurement of rebar stress within concrete structures under various loading conditions remains a significant challenge. This study investigates the characteristics of force-magnetic coupling effects in rebar under different loading conditions, focusing on the coupling effects between stress and magnetic fields. Twelve groups of reinforced concrete composite (RCC) specimens were loaded under axial tension and three-point bending. The magnetic composite parameter B V was introduced. An innovative integrated machine learning model was established to analyze the force-magnetic coupling patterns. Using 9346 sets of experimental data, four machine learning models have been trained to perform rebar stress detection. Research results indicate that the rebar diameter in concrete exhibits a positive correlation with external magnetic field intensity. During axial tension and three-point bending, the B V -σ fitting lines display opposite trends. Comparing four models, in the axial tension experiment, the stress testing data points from the RF and PSO-XGBoost models fall within ±10% error, with R² values both reaching 0.99. In the three-point bending experiment, the PSO-XGBoost model achieved the optimal R² value of 0.91, demonstrating excellent detection stability. SHAP value analysis indicates that B V and B Z are the critical factors for stress detection in axial tension and three-point bending loading, respectively. Under both loading conditions, the relative detection errors of the model are concentrated within the ranges (-0.164, 0.173) and (-0.951, 2.342), respectively. This indicates higher accuracy in rebar stress detecting under axial tensile loading.