Machine learning modeling demonstrates improved prediction of punching shear strength in reinforced concrete flat plates, highlighting safer structural engineering design.
Key Points
To develop an optimized machine learning model and a closed-form empirical equation using symbolic regression for predicting the punching shear strength of reinforced concrete flat plates.
Compiled a comprehensive database of 472 experimental reinforced concrete flat plate specimens without shear reinforcement.
Applied feature selection and Bayesian optimization to build and fine-tune predictive machine learning models.
Derived a symbolic regression closed-form equation and evaluated its accuracy, dispersion, and bias against existing building codes and prior machine learning models.
The proposed machine learning model and closed-form equation outperformed existing design code formulations in predicting punching shear capacity.
Validation on the experimental test dataset demonstrated that the developed models achieved higher predictive accuracy and lower dispersion and bias compared to prior predictive models.