Artificial neural network for structural behavior prediction of RC one-way slab strengthened by CFRP

Abstract

In this project, 6 Reinforced Concrete (RC) slabs with various length and thickness of carbon fiber reinforced polymer (CFRP) in comparison with the plain RC slab have been used to generate Artificial Neural Networks (ANNs) for structural behavior prediction. The slab dimension was 1800 × 400 × 120 mm and the length of the CFRP was 700, 1100 and 1500 mm in two different cross section area of 60 and 96 mm2. The results of this experimental work are noted in each testing process. The general regression neural network (GRNN) was the first practical approach that has applied for structural analysis prediction. The feed forward back-propagation (FFB) was the second method with a per regression method for data collection to increase the number of data for training, verifying and testing. The two used method had minimal error and maximum correlation coefficient. The amounts of MSE and RMSE in GRNN and FFB system were in the acceptable ranges. The correlation coefficient is closed to 1 for output data. Key words: Artificial neural networks, carbon fiber reinforced polymer, general regression neural network, feed forward back propagation, mean squared error, root mean squared error.