Predicting Shear, Stiffness and Stirrup Strain Histories in Reinforced Concrete Beams Using Machine Learning

Abstract

Shear failures in reinforced concrete structures should be evaluated to ensure safety. Traditional evaluation methods for members with shear cracks include detailed analyses or expert opinion. This study uses machine learning to investigate how crack widths correlate with shear load, stiffness, and stirrup strain histories. A shear test database is assembled with 260 and 480 crack width measurements for rectangular slender beams with shear reinforcement ratios smaller and larger than the minimum required by ACI 318-19, respectively. Measured load-displacement relationships, stirrup strains, and crack widths were documented. Gaussian Process Regression (GPR), a machine learning method, is used to correlate crack width to shear history, stiffness, and stirrup strains considering beam design details. The three indicators (shear, stiffness, and stirrup strain histories) predicted based on test data given a crack width can be a rapid in-service performance evaluation tool for reinforced concrete beams with signs of shear distress.