Study On Source Time Function In Rayleigh Waveform Inversion

In recent years,geophysical exploration offers a variety of methods to detect shallow surface structure.High frequency surface wave method is one of the mostrepresentative,which includes surface wave spectrum analysis technology and multichannel surface wave analysis technology.However,the limitations is that they are greatly influenced by the lateral inhomogeneous medium.So,they are not suitable for the actual complex field data processing.But the waveform inversion technology is not limited by the horizontal layered medium condition,which is applicable to the detection of underground horizontal non-uniform media.At present,it has attracted the attention of many scholars.The advantage of this method is that it uses all the waveform information contained in the seismic record.After decades of research and development,the waveform inversion technology has made great improvement and progress.In spite of this,the studies on the surface wave full-waveform inversion method are still inadequate.There are still many problems to be solved in the inversion process,such as the selection of initial model,the determination of seismic wavelet in inversion,and the determination of viscoelastic medium quality factor,etc.In the above context,in view of the difficulties in full-waveform surface wave inversion,this thesis proposes a Rayleigh waveform inversion technology based on source time function inversion in combination with the existing research results of our research group.Since in most cases the source time function used in field records is unknown,especially in the case of transient sources,the so-called "source time function" may vary from one shot to another.Therefore,it is impossible to use a proper source time function to simulate the seismic record in full waveform inversion.Therefore,this thesis adopts the method of establishing a wavelet correction filter.After using a general seismic wavelet for forward modeling and obtaining synthetic records,the wavelet correction filter is used for seismic wavelet inversion iteration to reduce the residual errors between the synthetic data and the measured data,which canimprove the final inversion accuracy.Because the effect of using wavelet correction filter is limited,the seismic wavelet extraction method is also adopted in this thesis.The wavelet is extracted directly from the field seismic record,and then is used in Rayleigh waveform inversion based on wavelet correction,which further improves the precision of waveform inversion.This process can be effectively applicable to the lateral non-uniform medium.In this thesis,viscoelastic model with increasingly velocity,void model and fault model are tested,and good inversion results are obtained,which verify the reliability of the proposed method.