Does a Single ANN Properly Predict Pushover Response Parameters of Low-, Medium- and High-Rise Infilled RC Frames?

Artificial neural networks, ANNs, can predict the behavior of systems that have common main features. When the problem under consideration contains groups that involve different main features, different ANNs are needed to predict the behavior of each group separately. In this paper, the efficiency of a single ANN to predict the lateral behavior of two-span structures representing a mix of low-, medium- and high-rise buildings in Egypt was investigated. All buildings were first analyzed using nonlinear pushover analysis to obtain their capacity curves, failure loads and displacements. Obtained data were used for training different ANN models. The results indicated the efficiency of a single ANN to predict the behavior of a mix of all buildings under investigation with a confidence level of 99%. The successful network was further utilized to obtain another set of data that were merged with the original data and used to develop a design neural network. The obtained network showed a very good capability to predict design variables which can be a good tool for engineering practitioners. [ABSTRACT FROM AUTHOR]