High-frequency Surface-wave Waveform Inversion Based On Wavefield Separation

Urbanization is an inevitable trend in the modernization of human society.In the next 10 years,China's urbanization process will run at its highest speed in history.China's urbanization has great demand for underground space.The in-depth study and use of highfrequency surface wave methods will undoubtedly be crucial to promoting the rapid and sustainable development of our society before the impact and destruction of urbanization on the environment in China is not yet uncontrollable.Near surface geophysical methods face many theoretical and practical problems in the rational development,comprehensive utilization and effective protection of underground space.With the continuous development of near-surface geophysics methods,how to improve the accuracy of seismic inversion,accurate depiction of shallow underground structures has become an urgent problem to be solved in near-surface geophysics.The Multi-channel Analysis of Rayleigh wave and Love wave methods(MASW and MALW)are based on the inversion of the dispersion curve to obtain the subsurface Swave velocity structure.Since the dispersion curve forward modeling is based on the horizontal layered model assumption,the Rayleigh wave and Loew wave multi-channel analysis methods are only applicable to the horizontal layered model and the earth model of the lateral S-wave velocity gradual change.This is considered an important limitation in the development of the current high-frequency surface wave method.Another problem is the correct identification of the dispersion curve mode,such as the existence of leakage waves and "mode kissing".In the actual data processing,if there is no other data to limit the model,the correct identification of the dispersion curve mode is almost impossible.For complex earth models with soft interlayers,undulating surfaces,etc.,the extraction of the dispersion curve becomes very difficult.Therefore,the full waveform inversion(FWI)method,as a new research hotspot in the geophysical world,has attracted the attention of more and more mathematicians and geophysicists.It is a data fitting inversion method based on full wave field simulation technology.The seismic wave field observed by the surface is used to directly invert the parameters of the subsurface medium,such as the velocity of the longitudinal and transverse waves,the density,the attenuation factor,and the anisotropic parameters.Because it utilizes the motion and dynamics of the seismic wave field,it can more realistically reveal the structural details and reservoir information under complex geological backgrounds compared with the traditional seismic inversion method.Waveform inversion is not constrained by the layered model hypothesis.By directly fitting the measured seismic wave waveform to obtain the shallow surface shear wave velocity,it does not depend on the pick-up of the dispersion curve,and is suitable for most earth models,thus breaking the current high-frequency surface wave.In particular,the surface wave energy is strong,and the surface wave waveform inversion has great potential.This is demonstrated by the fact that the Love waveform inversion with fewer model parameters is present.Since Rayleigh wave is formed by superposition of P wave and SV(transverse wave vertical component)waves along free surface,Rayleigh wave is not only related to S wave velocity,but also related to P wave velocity,and more model parameters.The result of the Rayleigh wave waveform inversion is extremely dependent on the initial model,and the non-uniqueness is increased,so the practical application effect is limited.In order to make the results of the Rayleigh wave waveform inversion reasonable in geological sense,it is common practice to assume that the P wave velocity is known to improve the stability of the inversion process and reduce the non-uniqueness of the solution.This assumption is acceptable for models with small lateral P-wave velocity perturbations.However,in recent years,the practice of high-frequency surface waves has shown that this assumption does not hold for shallow surface velocity models with anomalous bodies(with large velocity differences).Although the S-wave velocity is the most important factor affecting Rayleigh wave,the P-wave velocity is relatively small,but for the actual high-frequency seismic data with serious influence on the body wave,it is assumed that the known P-wave velocity deviates far from the true value will cause the instability of the inversion process.Therefore,in the Rayleigh wave waveform inversion process,how to determine the relatively accurate P wave velocity remains to be solved.When there is no other data to constrain the Earth model,an effective means to improve the inversion stability of geophysical data and reduce the non-uniqueness of the solution is to reduce the model parameters(self-constraint).Research on Rayleigh wave shows that,in general,in addition to the S-wave velocity,the P-wave velocity is also one of the parameters affecting Rayleigh waves.Although the effect of P wave velocity on Rayleigh wave is only about 20% of S wave velocity,the influence in Rayleigh wave full waveform inversion cannot be ignored.If the P wave velocity is also used as an unknown parameter to participate in the Rayleigh wave full waveform inversion,the reliability of the inversion model cannot be guaranteed.Therefore,in practical applications,it is generally assumed that the P wave velocity is known.If we can retrieve the S-wave velocity structure before inverting Rayleigh wave,we use the body wave information contained in Rayleigh wave to extract the relatively accurate P-wave velocity,and then set it in the process of inverting Rayleigh wave to obtain the S-wave velocity structure.For known parameters,the stability of the inversion and the reliability of the inversion model will be effectively improved.This is especially true for Rayleigh wave full waveform inversion(more unknown variables).The core idea of this thesis is to divide the Rayleigh wave full waveform inversion into two steps.First,an effective and high-precision method for separating body waves and surface waves in vertical component seismic records is studied.The surface waves are separated from the seismic records,and the separated waveforms are retrieved by acoustic wave approximation full waveform inversion.Secondly,the P-wave velocity structure obtained by the inversion body wave is taken as a known parameter,and the PSV wave full waveform inversion is used to achieve the purpose of acquiring the S-wave velocity structure.Since the P-wave velocity is known in the P-SV wave full waveform inversion,the model is constrained,and the model parameters are reduced,thereby reducing the multi-solution of Rayleigh wave inversion and improving the stability of the inversion process.The main research contents of this thesis include:(1)The crosstalk influences between different parameters of elastic wave full waveform inversion(P-SV wave and SH wave)is analyzed.Since S wave occupies a major part in surface wave recording,the analysis results show that S wave model affects the inversion results of P wave and density.The multi-parameter elastic wave inversion can better recover the S-wave velocity structure,but the recovery of the P-wave velocity and density model is poor.However,although the inversion effect of P wave velocity is poor,P wave velocity is still an important reason for influencing the Rayleigh wave waveform inversion.In order to further prove the influence of P wave velocity on the inversion results,the accurate initial P wave velocity and the inaccurate initial P wave velocity are set respectively,and the S wave velocity model obtained by the inversion is compared.The result shows that the initial P-wave velocity is of great significance for obtaining a higher precision S-wave velocity model.At the same time,the accurate initial density model and the inaccurate initial density model are set respectively.The S-wave velocity model obtained by inversion is compared.The results show that compared with the P-wave velocity,the initial density model has less influences on the inverted S-wave velocity.Therefore,how to determine the relative accuracy of the initial P-wave velocity is important before the surface wave waveform is performed.(2)The high-resolution linear Radon transform is used to transform seismic data from the time-space domain to the frequency-velocity domain.In the frequency-velocity domain,due to the double difference between speed and frequency,the body wave and the surface wave are almost completely separated,and the high-precision body wave(or surface wave)is extracted by filtering.The body-wave recording is obtained by reverse transforming the frequency-velocity domain body wave into the time-space domain.This provides data for full waveform inversion based on acoustic wave equations.The separated body wave data contains almost no surface wave information.Therefore,the full waveform inversion result based on the acoustic wave equation can provide a more accurate P wave velocity for the Rayleigh wave full waveform inversion based on the PSV wave equation.The results of simulation and actual data show that the separation of body and surface waves in the frequency-velocity domain is better than the currently known methods by using the high-resolution linear Radon transform.The results of the simulation data show that the separated body waves and surface waves are almost the same as the theoretical data.(3)In the full waveform inversion of the acoustic waves,only the P wave velocity is used as the model parameter for the inversion of the body wave data after the face wave.Under the premise of no S-wave effect,the dependence of acoustic wave approximation full waveform inversion on the initial model is reduced,and the stability of the inversion process and the reliability of the result are improved.For the simulated tomographic model,by setting the relatively accurate P-wave velocity(acoustic wave full waveform inversion result),the Rayleigh wave waveform obtained after the separation is directly inverted to obtain the shallow S-wave velocity structure.The results show that when the P-wave velocity model obtained by the acoustic wave waveform inversion is used as the initial model,the obtained S-wave velocity structure has higher precision.(4)We analyze several main parameters that affect the inversion accuracy of surface wave waveform,such as the length of the array,the number of detectors,the number of shots,and the dominant frequency of the wavelet.Numerical experiments were carried out on different array lengths,different number of detectors,and number of shots.The effects of different parameters were analyzed by comparing the recovered S-wave velocity models.In this thesis,the characteristics of strong surface wave energy are used to obtain relatively accurate shallow surface P wave and high precision S wave velocity structure through the study of high frequency Rayleigh wave waveform inversion,which is a reasonable development,comprehensive utilization and effective protection of underground space.The results of this thesis aim to improve the accuracy of highfrequency surface wave method imaging on shallow surface complex media,broaden its application range,and improve its application value in shallow surface geophysics and geotechnical engineering.