Application of artificial neural network for prediction of fracture energy of concrete

Abstract

The analysis of fracture parameters of concrete drawing the researchers attention and getting popular day by day. Every concrete structure undergoes crack formation, initiation and propagation phase, to understand the kind and severity of crack study the fracture mechanics is very much needed. Fracture energy (G_f) is one the main characteristic amongst numerous fracture parameters. The different parameters such as water to cement (w/c) ratio, compressive strength (fc), diameter of aggregates, testing age of specimens etc. play essential role in understanding the G_f. In present work, G_f of concrete is measured by replacing cement with nano TiO₂ (NT) at 1, 2, 3, and 4% of the concrete mix, as well as fly ash (FA) and ground granulated blast furnace slag (GGBS) at 10, 20, 30, and 40%. The G_f was investigated using the size effect technique (SEM), and the notched beams were subjected to a three-point bend test. According to the experimental results, the NT4FA40 mix had the maximum G_f, while mixtures including FA out performed than GGBS mixes. Furthermore, an attempt was made to anticipate G_f using the soft computing method in light of the current necessity. The G_f is predicted using an ANN. The literature database, which includes 193 fracture tests, was gathered from earlier research in addition to the data from the current experimental investigation. Furthermore, the formula proposed by Bažant and Becq-Giraudon was utilized to make predictions based on a number of characteristics, including compressive strength, maximum aggregate size, and w/c ratio. compressive strength, maximum aggregate size, and water to cement ratio are among the characteristics that are trained, verified, and tested for ANN. The ANN model developed using literature-based data, Bažant and Becq-Giraudon equation derived data and experimental data gives promising results with R values 0.999, 0.981, 0.984 respectively. The present study concludes, ANN model shows the excellent output for prediction of G_f.