Estimating the primary crack spacing of reinforced concrete structures: Predictions by neural network versus the innovative strain compliance approach

Abstract

The paper presents a comparison of the novel strain compliance concept, proposed for predicting the crack spacing of reinforced concrete structures with neural network predictions. The concept represents an alternative way to accurately analyze the cracking behavior of reinforced concrete elements while maintaining compatibility of deformation behavior and ensuring mechanical soundness. A multiple run and surrogate data based approach was adopted to train and calibrate an artificial neural network for primary crack spacing prediction. The findings substantiate the experimental primary crack spacing data and reveal the performance of the strain compliance approach to be similar to the trained neural network.