Study on fatigue life of high-strength steel rebars joined by flash butt welding based on experimental and machine learning approaches

Abstract

The fatigue failure of steel rebars joined by flash butt welding is a contributing factor to structural damage in reinforced concrete structures. In this work, a series of fatigue tests based on steel rebars joined by flash butt welding were carried out, and a fatigue life database consisting of 155 specimens were obtained. Three distinct machine learning models are utilized to analyze and predict the fatigue life of the steel rebars joined by flash butt welding. Additionally, The SHapley Additive exPlanations (SHAP) method is employed to assess the significance and impact of each input parameter on the model predictions. The results demonstrate the superior performance of the ANN model, with an R2 of 0.865, MSE of 0.134, MRE of 0.872, and MAE of 0.171. The factors influencing the fatigue life of steel rebars joined by flash butt welding, ranked in descending order of significance, are stress amplitude, rebar diameter, stress ratio, loading frequency, and material strength. The experimental and predicted values for the 2 million fatigue stress amplitude differ by 4.64 MPa (approximately 2.99 %), demonstrating the robust generalization capabilities of the ANN model, enabling accurate predictions of the fatigue life of steel rebars joined by flash butt welding throughout the entire cycle. The proposed model can be used to assess the fatigue life and design fatigue stress amplitude of steel rebars joined by flash butt welding, offering a promising alternative to fatigue testing.