Optimizing ANN models with PSO for predicting short building seismic response

The present study aimed to optimize the artificial neural network (ANN) with one of the well-established optimization algorithms called particle swarm optimization (PSO) for the problem of ground response approximation in short structures. Various studies showed that ANN-based solutions are a reliable method for complex engineering problems. Predicting the ground surface respond to seismic loading is one of the engineering problems that still has not received any ANN solution. Therefore, this paper aimed to assess the application of hybrid PSO-based ANN models to the calculation of horizontal deflection of columns in short building after being subjected to a significant seismic loading (e.g., The Chi-Chi earthquake used as one of the input databases). To prepare both of the training and testing datasets, for the ANN and PSO-ANN network models, a series of finite element (FE) modeling were performed. The used FEM simulation database consists of 8324 training datasets and 2081 testing datasets that is equal to 80% and 20% of the whole database, respectively. The input includes Chi-Chi earthquake dynamic time (s), friction angle ( φ ), dilation angle ( ψ ), unit weight ( γ ), soil elastic modulus ( E ), Poisson’s ratio ( v ), structure axial stiffness (EA), and bending stiffness (EI) where the output was taken horizontal deflection of the columns at their highest level ( U x ). The result indicates higher reliability of the PSO-ANN model in estimating the ground response and horizontal deflection of structural columns in short structures after being subjected to earthquake loading.