A NEURAL NETWORK-BASED METHOD FOR PREDICTING SEISMIC RESPONSE OF STRUCTURES

Simulating and predicting the realistic behaviour of structures subjected to seismic excitations has been one of the most cited topics in structural issues over recent years. Due to the fact that the aleatoric variability and epistemic uncertainty in earthquake records can result in a significant variability in seismic structural response, the precise assessment of the maximum lateral displacement of structures requires solving numerous differential equations for a large number of earthquakes. However, the analytical procedure is time consuming and computationally intensive.On the other hand, data-driven modelling techniques including Artificial Neural Networks (ANNs) can serve as simplified mathematical models of human brain, which have been widely utilized in reducing the computational burden of numerous complex analytical problems. In fact, supervised ANNs can be considered as parallel computing systems capable of input-output mapping aiming at simulating the desired target by only using the information learned from past experiences.Hence, the current paper investigates the novel idea of the application of Wavelet Neural Networks (WNNs) in rapid and precise seismic response assessment of structures using ground motion indices. To achieve this objective, a wavelet activation function is considered as the nonlinear estimator of a classic multi-layer feed forward neural network to predict the maximum lateral displacement of a simplified lumped mass model subjected to real unidirectional seismic excitations. Sensitivity analyses are also conducted to determine the relative importance of the seismic parameters in characterizing damage potential of the studied structural system. Based on the obtained results, displacement- and energy-related parameters have stronger correlation with displacement demands of long-period structures, compared to acceleration- and velocity-related variables.

کلیدواژه ها:

Seismic Response of Structures ، Ground Motion Indices ، Data-Driven Modeling ، Seismic Response of Structures ، Ground Motion Indices ، Data-Driven Modeling