Prediction of Mechanical Properties of Steel Fiber-reinforced Concrete Using CNN

The performance of Steel Fiber-reinforced Concrete (SFRC) is superior to that of conventional concrete. Due to its intricacy and limited available data, the development of a strength prediction model for SFRC is very difficult. To prevail over this constraint, research was carried out to build a deep-learning algorithm for the prediction of flexural, split tensile and compressive strengths of SFRC. To accomplish this, a dataset was created by accumulating SFRC strengths through an extensive literature survey. Initially, the deep features of fine aggregate-cement ratio, coarse aggregate-cement ratio, water-cement ratio, fly ash-cement ratio, super plasticizer-cement ratio, length, diameter and dosage of fiber are learned through a convolutional neural network. Then, softmax regression was used to develop a prediction model. The prediction model is trained and tested using 89 datasets with various mix ratios. From the results, we can conclude that the deep-learning-based prediction model exhibits greater accuracy, greater efficiency and greater generalization capacity compared to those of the conventional neural network model.