Research On Seismic Traveltime Tomography And Waveform Inversion

With the development of the world economy, increasing demand for oil and gasresources call for higher standard of seismic exploration. The target of exploration graduallyshifts from simple structural reservoirs to complex lithologic reservoirs, subtle reservoirs,combination reservoirs or unconventional reservoirs with features of smaller geological scale,increased complexity, high heterogeneity and serious lateral variation for velocity. Currently,pre-stack depth migration is a key technology to solve seismic imaging problems of complexstructure, which is more sensitive to velocity than pre-stack time migration. The impact thatinaccurate velocity has on the pre-stack depth migration cannot be made up for by any precisemigration algorithms, so doing some research on velocity analysis and inversioncorresponding to pre-stack depth migration is very important. The strongly sensitivity ofdepth migration to velocity is a disadvantageous thing for migration itself, but on the contraryit is a good thing for velocity inversion. That is, we can take advantage of the strong coupledrelationship between velocity and depth for better inversion and ultimately provide ahigh-precision velocity field for pre-stack depth migration. The method mentioned above isbased on the fitting of image domain, also we can implement the velocity inversion by fittingin the data domain as well.Based on the above considerations, we have done research on traveltime tomography inimage domain and waveform inversion in data domain under the theoretical framework ofseismic inversion. Any kind of velocity analysis methods are not successful by one iteration,instead iterations needed to approach the real solution, this idea can be interpreted byinversion theory. Traveltime tomography methods use the best match of image domainbetween observed data and model data to get the velocity inversion, apply the idea of alternating iteration for migration and tomography, recover the high wave-number(velocityinterface) and low wave-number(velocity value) information of velocity field by migrationand tomography, which has high accuracy, stability and efficiency. While waveform inversionmethods use the best match of data domain, which has the highest precision of velocityinversion in theory, and the inversion process is carried out automatically free of humaninteraction in conventional velocity analysis. However, when seismic source wavelet isunknown and low wave-number information is inaccurate, it is easy to fall into the localminimum and with high cost of computing. Therefore, there is clear physical meaning fortraveltime tomography and waveform inversion under the framework of seismic inversion, wecan get the pre-stack velocity field by any of them, and also by joint method of them.The main researches in this paper can be summarized as follows:1. We have described the basic theory of seismic inversion, interpreted the significanceof traveltime tomography and waveform inversion viewed from probability and determinacyinversion, discussed the regularization problems and methods and expressed the point of viewthat regularization is necessary to make seismic inversion practical.2. Under the acoustic assumption we have derived the tomographic inversion equationand relationship between traveltime residual and depth deviation, proposed a traveltimetomography in image domain, developed a automatically fitting pick-up strategy for curvatureof ADCIGs, studied a damped regularization method in order to eliminate smearing effect ofray, presented detail things involved in process of implementation, and finally verified thecorrectness, effectiveness and practicality of this method through model and field datasetexamples.3. Under the framework of elastic vector wavefield, we have derived joint tomographyequations and relationship between traveltime residual and depth deviation for multi-wave,proposed a traveltime tomography in image domain using vector ADCIGs for PP-andPS-wave, selected the fast, flexible and ray-theory based method--Gaussian beam migrationto do migration and extract ADCIGs considering most computation are used up for migrationand ADCIGs extraction, and finally verified the correctness, effectiveness and practicality ofthis method through model and field dataset examples.4. On the basis of acoustic wave equation, we have expressed the theory of time-domain waveform inversion, derived the formula of updated gradient and updated step-length basedon least squares theory, provide objective functional formula with regularization and wellconstraint, established a set of CPU-based parallel algorithm flow, verified the correctness ofthis method by examples and proposed some reasonable recommendations making waveforminversion practical based on analysis of key influence factors.