A Finite Element Approach to Evaluate and Predict the Shear Capacity of Steel Fiber-Reinforced Concrete Beams

Abstract

Adding steel fibers to a concrete matrix enhances the shear capacity of reinforced concrete beams. A comprehensive understanding of this phenomenon is essential to evaluate engineering designs accurately. The shear capacity of Steel Fiber Reinforced Concrete (SFRC) beams is affected by many parameters, such as the ratio of the shear span to the effective depth of the SFRC beam, the compressive strength of concrete, the longitudinal reinforcement ratio, volume fraction, aspect ratio, and the type of fibers. Therefore, to cover the influence of these parameters on the shear capacity of SFRC beams, 91 beams from previous studies, divided into 10 groups, were considered in the current study. Two approaches have been used to predict the shear capacity of SFRC beams. The first approach used 7 predicting equations derived from previous studies and the second one used finite element analysis (ANSYS software) to simulate the 91 beams. Despite the many approaches to simulate the structure elements, there is no reliable approach able to simulate satisfactorily 91 SFRC beams as this study does. The log file of ANSYS software was used to simulate and calculate the shear strength capacity of the beams. The results show a reasonable agreement with the experimental tests. The extracted results were much closer and more realistic than those obtained by the predicting equations. Also, the χ factor (squared value of experimental shear capacity to the predicted shear capacity) of the ANSYS software results is 97%, while the closest proposed equation gives 91%.

Prediction of Concrete's Compressive Strength via Artificial Neural Network Trained on Synthetic Data

“…For example, several studies attempted to create models that accurately predict the properties of construction materials by training on data available in the literature [1][2][3][4]. In addition, many properties of fresh and hardened concrete of many different types have been studied and predicted using ML and deep learning models over the past two decades [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21]. Meanwhile, many studies attempted to increase the predictive power of ML models by introducing more complexities or training them with additional experimental data.…”