Evaluation of neural network model for compressive strength of the steel fiber reinforced concrete using break-off method

Abstract

In the present paper break-off test as a partially-destructive method is used for durability evaluation of steel fiber reinforced concrete. In recent years, utilizations of steel fibers have been known as an appropriate solution method for sudden fracture of concrete. In order to provide a comprehensive statistical database, 24 mixtures are designed with various cement content (400, 450, and 500 Kg/m3), maximum aggregate size (12.5, 25 mm), steel fibre volume fractions (0, 0.33, 0.67, 1 %), and the constant water/cement ratio of 0.4 for all mixtures. Hence, influencing factors of steel fiber reinforced concrete characteristics and break-off test results are evaluated. The investigations show that the volume fraction of steel fibers and its features significantly affect the results of break-off test. Furthermore, in this study conventional numerical neural networks are developed for predicting the compressive strength of concrete with various mixtures and ages. ANN is sophisticatedly capable of being trained from the existent data and extending their behavior on a new dataset. This ability introduced ANN as an apt tool for modeling the complex mechanisms and systems in engineering applications. Statistical indices are used to compare the efficiency and accuracy of models. The result of this study has confirmed the accuracy of artificial neural network models in determination of the compressive strength of concrete.

Keywords

• break-off test
• concrete strength
• steel fiber
• partially-destructive test
• artificial neural network