A data-driven simulation method for near-fault non-stationary ground motion

Abstract

The invention provides a near-fault non-stationary seismic oscillation simulation method based on data driving. Initially a near-fault non-stationary seismic oscillation record is decomposed into superposition of a series of mono-component signals by means of SWT, then Hilbert transform is conducted on each order of the mono-component signals, and simulation of a non-stationary seismic oscillation sample is achieved through introduction of random variables and by means of a signal reconstruction technology. By means of the method, a redundant wavelet base function is used, representation of original signals on the redundant base function can be seen as the average of representation of the signals on a series of base functions, wavelet coefficients and scale coefficients have equal length with the original signals, the translation invariance is ensured, not only can the oscillation effect of dyadic wavelet transform be well inhibited, but also disturbance of part of the wavelet coefficients cannot result in severe distortion of the signals, so that dependence of the reconstruction effect of the signals on single wavelet coefficient is reduced, thus the SWT has a better advantage over the aspects of seismic wave decomposition and reconstruction, and the precision of sample simulation is ensured.

Description

technical field [0001] The invention relates to the technical field of ground motion simulation, in particular to a data-driven near-fault non-stationary ground motion simulation method. Background technique [0002] Ground vibrations propagated by seismic waves are usually called ground motions. Near-fault ground motions have some characteristics that are significantly different from mid- and far-field ground motions. Near-fault ground motions contain high-frequency components and low-frequency pulse components. Because the pulse components have a large amplitude , The characteristics of short duration, so it will cause serious damage to the engineering structure. However, the number of measured near-fault ground motion records is very limited, so it is an urgent problem to propose an efficient simulation method for near-fault ground motion. [0003] At present, scholars at home and abroad often adopt the simulation method of superimposing the stochastic model reflecting t...

AI Technical Summary

Problems solved by technology

Empirical mode decomposition and overall empirical mode decomposition methods do not have a complete mathematical basis and largely depend on user experience; wavelet decomposition and wavelet packet decomposition methods do not have translation invariance, which will cause the loss of reconstruction information. The disturbance of some wavelet coefficients will also cause severe distortion of the signal, which is quite different from the actual signal, and the reconstructed simulation results are not accurate enough

Images