Neural network based technique for preliminary design of reinforced concrete columns

Abstract

An Artificial Neural Network (ANN) based preliminary design methodology for reinforced concrete columns was developed by linking two neural networks with logical rules as per IS:456-2000. The first neural network is simulated for capturing interaction diagram for column subjected to axial load and uniaxial bending moments. The second neural network is developed to capture the interaction equation for column subjected to axial and biaxial moments. The sensitivity of the reinforcement in columns corresponding to different parameters such as depth to cover ratio, arrangement of reinforcement bars (number of bars in a row in horizontal and vertical direction), and grade of steel are captured in ANN. The trained ANNs automatically interpolate and extrapolate the results for given design parameters. In this study, the neural networks are used to replace the iterative processing steps involved in actual design computation.