DEVELOPING AN ARTIFICIAL NEURAL NETWORK MODEL THAT PREDICTS THE FUNDAMENTAL PERIOD OF STEEL STRUCTURES USING A LARGE DATASET

Abstract

The fundamental period of structures is an important parameter to consider when designing structures in seismic-prone areas. Currently, the formulae available in the international literature and design codes fail to capture the true dynamic behaviour of structures, especially when they are founded on soft soils. It is necessary to develop more accurate models for predicting the fundamental period while taking into account the soil-structure interaction (SSI) effect. For the needs of this research, a dataset of 49,154 models (98,308 numerical results) was created for developing a predictive model for calculating the fundamental period of steel structures. The SSI phenomenon was also considered with structures modelled with a soil domain with varying depths. The model used herein is an Artificial Neural Network (ANN). The ANN model was able to predict the fundamental period with a correlation of 99.99% and a mean absolute percentage error (MAPE) of 0.7%.