Modelling of Velocity Pulse-Like Ground Motions Based on Adaptive Matching Algorithm

Due to the lack of a complete waveform parameter identification method aiming at the velocity pulse-like ground motions up to now, it is still difficult to identify the pulse-like records quantitatively, which hinders definitely collecting the wave shape parameters from ground motion database. In order to solve this problem, a framework for the parameters identification of pulse waveform is presented by combining the pulse quantitative identification and impulsive function analysis with an adaptive matching algorithm to realise the adaptive matching between pulse waveform and the analytic function. As the key part of the framework, the adaptive matching algorithm for parameterizing the pulse waveform is proposed by virtue of the empirical mode decomposition (EMD) in order to smooth the near-fault pulse-like earthquake records. Based on the pulse function given by Marverodis, and within the distribution range of the model parameters, a procedure for parameters adaptive matching is developed, characterized by the enumeration with several loops corresponding to the pulse frequency, pulse shape, and pulse oscillation characteristic respectively. In addition, the least square-based optimization method is applied to determine the final parameters. Using this method, the shape parameters of the main pulses existing in these pulse-like ground motions records were identified. In addition, the relationship between the peak ground velocity and the period of main pulse is drawn from the statistical results. According to the statistical and image analysis, compared with the existing method using the Dickinson model, the algorithm for identifying the main pulse period in this paper can reduce the classification mistake by about 67%. The better stability of near-fault records identification is meaningful to practical engineering applications in seismic design.