Research On Seismic Wave Synthesis Method Based On Wavelet Transform

According to the code for seismic design of our country,multiple seismic waves are required to be used for seismic time history analysis of structures.However,the number of seismic records in the world is limited,and the strong earthquake records can not meet the needs of current seismic analysis.Therefore,the synthesis of artificial seismic waves is particularly important.Although there are many methods to synthesize artificial seismic waves,the main idea is to fit them according to the target spectrum.This synthesis method has a single form,and does not take into account the time distribution characteristics of seismic waves.The synthetic artificial seismic waves and natural seismic waves may have great differences in timefrequency characteristics.With the development of artificial intelligence technology,there are more and more technical means that can be used to extract the time-frequency characteristics of natural seismic waves.Before that,people did not notice that clustering analysis can extract the time-frequency characteristics of natural seismic waves better.In this thesis,a method of synthesizing artificial seismic wave by wavelet transform and cluster analysis is proposed.Firstly,the seismic signal is decomposed by wavelet transform,and then the time-frequency characteristics of a large number of natural seismic waves are classified by cluster analysis.Based on this classification,the wavelet coefficient matrix is recombined and the new seismic wave is obtained by inverse wavelet transform:（1）The natural seismic wave is transformed by wavelet transform and the corresponding wavelet coefficient diagram is drawn.By comparing the wavelet coefficient diagram with the seismic wave shape diagram,it is found that the time of the seismic wave peak is consistent with the time corresponding to the brightest place in the wavelet coefficient diagram.（2）The clustering analysis is used to classify the wavelet coefficients of natural seismic wave,and the corresponding wavelet coefficient database is constructed.Through the classification results,it is found that the wavelet coefficients of M ≥ 7 seismic wave can be well divided into seven categories by clustering analysis,and the wavelet coefficients of each category have obvious characteristics.（3）The trained neural network is used to match the target wavelet coefficient matrix.A group of wavelet coefficients are randomly selected from the corresponding database.According to the maximum and minimum value of the target wavelet coefficients,the randomly selected wavelet coefficients are enlarged or reduced accordingly,and a new wavelet coefficient matrix is obtained,A new ground motion acceleration data is obtained.（4）Firstly,ANSYS is used to establish the finite element model,natural seismic wave and artificial seismic wave are applied to the structure,and the dynamic response is calculated and compared.The comparison results show that the response of artificial seismic wave and natural seismic wave to the structure is basically the same.Then,through image recognition technology,the wavelet time-frequency maps of artificial seismic wave and natural seismic wave are classified and matched,and the recognition effect is analyzed.The recognition results show that the recognition accuracy of wavelet coefficient map is very high,which can better test the synthetic effect of artificial seismic wave.