Regression‐based models for the torsional stiffness reduction factor in reinforced concrete beams under combined torsion and shear

Abstract

This paper presents a regression‐based methodology to estimate the Torsional Stiffness Reduction Factor (TSRF) of reinforced concrete beams under combined torsion and shear. A numerical database of 6816 samples was generated using the Combined Action Softened Truss Model (CA‐STM) implemented in Python, considering design guidelines prescribed by ABNT NBR 6118:2023. Based on exploratory statistical analysis, two simplified regression models were developed to predict the TSRF: a simple linear model dependent on shear loading and a nonlinear model accounting for torsion–shear interaction. The proposed models were applied within an iterative structural analysis framework in a practical case study using the commercial design software TQS. Results show that torsional stiffness is highly sensitive to shear interaction, as expected, and may be either underestimated or overestimated by code prescriptions. The proposed methodology provides a practical tool to support a more refined estimation of torsional stiffness reduction in reinforced concrete members, accounting for shear force, which is often neglected.