Using the Particle Swarm Optimization Algorithm for Identifying and Extracting the Prevalent Pulse of Near-Fault Ground Motions

With increasing the earthquake records over the past decades, it has been determined that ground motions in the vicinity of the causative fault (up to 15 km from the fault) can be significantly different from the motions away from the fault. These movements often include highdisplacement and velocity pulses with a significant structural damage potential and impose considerable seismic demand on the structure. Due to the devastating effects of such earthquakes, many engineers and seismologists have focused on the quantitative identification of pulse bearing records and simulation of the near-fault ground motions. Many simulation models extract the prevalent velocity pulse of the near-fault motions through fitting the displacement, velocity, acceleration time histories, and the corresponding elastic-response spectra obtained from the model and the actual record. So far, determination of the analytical models parameters has been accompanied by a manual trial and error process. Such trial-and-error-based process limits the ability of engineers to apply these relationships and investigate their effects on research and practical applications. In this study,a new approach is proposed for identifying and extracting the prevalent velocity pulse from earthquake records by optimization algorithms and the mentioned models. Particle Swarm Optimization (PSO) algorithm is used to simultaneously minimize the difference between the time history and the corresponding elastic-response spectra of the simulation model and those of actual, by defining a suitable objective function.The objective function in the optimization process is as a constrained function where the root-mean-square difference between the values of the pseudo-velocity elastic-response spectrum obtained from the pulse simulation model and its actual record is as the target function, and the root-mean-square difference between time histories of the corresponding velocity is as the constraint. With the objective function defined, the physical parameters of the simulation models are determined without the need for manual trial and error process. The optimization algorithm converts the manual trial and error in process of pulse extraction into the systematic trial and errors with the minimum analytic intervention and judgment. Then, by the proposed method and Hosseini Vaez mathematical model, a set of near-fault records have been simulated and stated in mathematical expression. The mentioned model includes harmonic and polynomial expressions, which is capable to simulate various pulses with a simpler form. Although this model simulates the long-period portion of near fault records, the parameters of model are determined based on a try and error method. The proposed method has been used to extract and simulate the prevalent pulses of near-fault records of Iran for the Tabas and Bam earthquakes. Comparing the results with other studies represents the efficiency of the proposed method for extracting the prevalent velocity pulse from near-fault records and expressing them as closed mathematical equations. The generated pulse history can be used to structural analysis and investigation of structures response to near-fault ground motions. Besides, since the synthetic near-fault ground motions are a combination of long-period-dependent velocity pulse and high-frequency independent seismic wave, the proposed approach can be used to generate time history of the long-period dependent velocity pulse.