| Geophysical exploration includes gravity,magnetic,electric and earthquake.Each method can only reflect one physical property of the underground material,and due to various factors,a separate geophysical inversion often has instability and non-uniqueness.However,with the economic development,the exploration targets become more and more complex,and the exploration depth continues to deepen,which places higher requirements on the accuracy of geophysical inversion.Therefore,high-accuracy and high-precision geophysical inversion methods need to be studied urgently.In this paper,the joint inversion between MT and seismic full waveform data based on cross-gradient structure constraints is studied.The joint inversion refers to the joint application of several geophysical observation datas in the process of inversion to establish a connection through the relationship between the rock physical properties or geometric structure of the geological body to invert the same underground geophysical model.Because the underground geophysical model we want to study is the same,it must fit any geophysical field observed on the ground.In order to simplify the calculation,we propose an alternate joint inversion strategy,that is,the cross gradient function is directly added to each separate inversion objective function,and the two inversion processes are carried out separately and alternately.Thus,the structural constraints between the resistivity model and the velocity model are realized,so that the physical property boundaries tend to be consistent.To achieve joint inversion,first of all,a good separate inversion is needed.The separate inversion includes two parts: forward and inversion.The seismic forward and inversion in this paper is carried out in the frequency domain,which can bettersimulate the attenuation of waves when propagating in underground media,and only need to select a few frequencies to get more ideal results,save calculation time.The forward simulation uses the 9-point finite difference method with higher accuracy to discretely obtain the wave field of the acoustic wave equation,and the inversion uses the full waveform inversion with high accuracy and high resolution.After that,we studied the modified PRP conjugate gradient method and applied it to the calculation of the update amount of the velocity model,this method converges faster and can avoid falling into a local minimum.Magnetotelluric forward modeling uses the finite element method,and the inversion uses the Gauss-Newton method that uses more comprehensive information to accelerate the convergence rate and improve the inversion accuracy.When constructing the objective function,model constraints are added to make the inversion result closer to the actual situation.Finally,through designing different models,the effect of joint inversion is shown.First,two sets of structurally consistent models are designed.The results show that,compared with the separate inversion,the joint inversion makes the velocity model and the resistivity model tend to be consistent on the physical property boundary,which improves the resolution of electromagnetic inversion and eliminates some slight false anomalies in seismic inversion.After that,a set of inconsistent models were designed.The results show that the cross-gradient function will not work,and false anomalies will not be introduced where one model has anomalies and the other model has no anomalies.Finally,we explored the influence of the weight coefficient selection in the objective function on the joint inversion results.The results show that when the weight coefficient is small,it exerts a slight constraint on the model structure,when the weight coefficient is too large,it will lead to over-constraint and introduce false exceptions.Therefore,in the joint inversion calculation,you can first conduct several experiments to find the appropriate range of weight coefficients,and find the weight coefficients within this range to make the inversion results better. |
PDF Full Text Request