A truss analogy and neural network for reinforced concrete deep beam analysis

Two simple neural networks are used to produce a generalised analysis of deep reinforced concrete beams based on a pin-jointed truss analogy. The first network carries out the structural analyses needed in the preparation of the training data required for the second neural net which then provides rapid analyses of beams described by four design parameters. When preparing the training data an iterative method is used to identify and remove concrete members cracked in tension. The output from the trained neural net consists of applied shear levels at first flexural and diagonal concrete cracking and at ultimate conditions.