SEICMIC CAPACITY OF STRUCTURAL ELEMENTS USING NEURAL NETWORKS

This paper presents the applicability of neural networks trained on the compiled experimental database to predict the seismic capacity of reinforced concrete walls and columns. The best built network is used for prediction of the behavior of new elements. Use of neural networks enables dependence analysis of observed behavior on different variables and simplifies behavior prediction of building elements under seismic loadings. It could be used for comparison with other methods for performance prediction of critical horizontal load carrying elements. For the seismic capacity evaluation required input for walls and columns is: type of loading, dimension and type of cross section, material properties and reinforcement. They are fed to the neural network trained on the experimental database and as output variables we get prognosis of: shear strength, failure type, critical loads and displacements. The whole procedure, input data, optimized neural network model and output variables are implemented in one worksheet.