Steel Rebar Yield-Point Prediction Model for Automated Quality Inspection in Tempcore Process

This paper conducts fundamental research to apply several widely used data-driven models to automate the quality inspection of Tempcore-processed steel rebar and design a prediction model for rebar mechanical properties by utilizing an optimal model. The steel rebar mechanical property prediction model has the potential to replace the conventional destructive test, thereby resulting in a reduction in test time and human resources and mitigating the risk of defective products. Subsequently conducting a comparative analysis of various machine learning-based regression models, the random forest (RF) model was chosen since it best predicted the mechanical properties of steel rebar. The performance of the RF model in predicting the yield point of rebar was determined to be 0.98 coefficient of determination, with a root mean square error (RMSE) of 12.59 [MPa]. Furthermore, the inside of the model was analyzed using the Local interpretable model-agnostic explanations method, with the aim of evaluating the rationality of the decision-making process of the model and assessing its reliability.