An artificial neural network model for preliminary design of reinforced concrete beam-column
Additional Document Info
The dimensions of a beam-column cross-section and the area of reinforcing steel required to support a specific combination of axial load and moment can be established by using the column design interaction curves, where an interaction curve represents all possible combinations of axial load and moment that produce failure of the cross-section. The bending resistance of an axially loaded column about a particular skewed axis due to biaxial moments can be determined through iterations and lengthy calculations. These extensive calculations are multiplied when optimization of the reinforcing steel or column cross-section is required. This paper investigated the suitability of an Artificial Neural Network (ANN) for modeling a preliminary design of reinforced concrete beam-column. Neural computing is a relatively new field of artificial intelligence (AI), which tries to mimic the structure and operation of biological neural systems, such as the human brain, by creating an ANN on a computer. An ANN back-propagation multi-layered model was developed to design a beam-column, which predicted column cross-section for a given set of inputs, which were concrete compressive strength, column types (Tied and Spiral), reinforcing steel ratio (0.01 - 0.08), factored axial load, P u , and moment, M u . In the present research, several different ANN back-propagation trial models with different layers/slabs connections, weights and activation functions were trained. The presented back-propagation multi-layered neural net with logistic activation function, "Rotation" for pattern selection, and "TurboProp" for weight updates was the best one among all other trials, which converged very rapidly to reach an excellent statistical performance. The trained ANN back-propagation model was tested with several actual design data, and a comparative evaluation between the ANN model predictions and the actual designs was presented.