Method Study On The Crosshole Seismic Traveltime Tomography In Anisotropic Medium

With the continuous development of seismic tomography technology,the requirement of imaging resolution has been gradually improved.The main research works we have down in this thesis are aimed at how to improve the resolution of the seismic traveltime tomography in anisotropic medium.We discussed the fundamental theory about anisotropic medium and showed the expression of phase(group)velocity.Meanwhile,the curves of group velocity with respect to phase slowness angle and ray angle are plotted and analyzed with examples.And,we discussed the relationship between imaging resolution and the characteristics of sensitive kernel function of elastic(or Thomsen)parameters(the first order partial derivative of velocity with respect to each parameter)varying with different phase-slowness angles.The forward modelling method used here is a multistage modified shortest-path method,and the inversion method used here is the damped minimum constraint least squares with conjugate gradient approach and the subspace solver.Combined with direct and reflected arrival data,we realized the anisotropic traveltime tomography for simultaneously updating elastic(or Thomsen)parameters.To improve the imaging resolution in anisotropic traveltime tomography and overcome the multisolution problems of qSV wave inversion,we in this thesis mainly discussed four aspects to improve or overcome the above problems.Firstly,by combining later arrival information to increase the coverage of ray angle;Secondly,by applying weight to balance the trade-off for each elastic(or Thomsen)parameter updated;Thirdly,by normalizing the partial derivative matrix and applying the multistage inversion procedure;Fourthly,by using the subspace inversion solver.To verify the correctness and effectiveness of the above method techniques,we carried out synthetic traveltime inversion tests for 2D crosshole anisotropic medium.The results show that:(1)the inversion results by combined later arrivals are better than that of single arrival data;(2)the resolution of less sensitive parameters is improved after applying weighted averaged processing;(3)the inversion resolution is improved to a certain extent by the partial derivative normalization and multistage inversion strategy;(4)the inversion resolution of subspace method is not as good as expected,even worse than that of conjugate gradient method.