The Study On Signal Enhancement And Interpretation Methods Of Full Tensor Gravity Gradient Data

The natural earth's gravity field is observed and studied in gravity exploration.Due to the inhomogeneous distribution of residual density from the surface to the deep part of the earth,gravity exploration has the advantages of being relatively economical and having larger exploration depth.Gravity gradient measurement observes the second order derivative of earth's gravitational potential and provides more detailed local information compared to gravity measurement.It has been widely used in recent years with the development of modern science and technology,materials science and test techniques.Currently,a couple of gravity gradiometry systems have been put into commercial applications and have played an important role in the regional geological survey,land and offshore oil and gas exploration.The methods of processing and interpretation are the key to the full tensor gravity gradiometry survey.Multiple interference fields are included in observed data and the anomaly is the superimposed total of the inhomogeneous distribution of underground density in the survey area.Therefore,the purpose of data processing is to remove various interferences,and then extract or enhance the anomalous response generated by the geological targets from the superimposed anomaly.The data interpretation techniques realize the reverse calculation from data to model.The geometric parameters and physical parameters of sources are predicted and estimated based on the anomaly.In this thesis,the study starts from forward calculation of geologic model,and the full tensor gravity gradient data is calculated for different models.Then the data is analyzed and combined with the distribution characteristics of noise,a new method for removing noise and enhancing the target signal is proposed.After that,anomaly analysis is carried out on the processed data.Based on the characteristics of geological targets reflected in the data,the fast imaging and three-dimensional inversion are applied to calculate the geometric and physical distribution of the targets.In light of the shortcomings of existing methods,improvements are made on the algorithms and constraints to improve the reliability of the interpretation results.Firstly,this thesis decribes the characteristics of gravity gradient data and the advantages over gravity data from the perspective of forward calculation and actual measurement,and briefly analyzes the measurement noise,which provides the foundation and basis for the research of the thesis.The analysis results show that each of the five independent components of the full tensor gravity gradient data reflects the distribution characteristics of the underground density along different directions.Since the derivative calculation is equivalent to a high-pass filter,the gradient anomaly contains more detailed information and higher horizontal resolution than the gravity anomaly,but more susceptible to high frequency noise.On the other hand,as the depth increases,gradient anomaly has a greater rate of decay than gravity anomaly and reflects less long-wavelength information of geological body,so it is difficult for gradient measurements to be applied into deep and regional tectonic interpretations.The data analysis results show that the gravity gradient measurement has a better target discovery rate in shallow-buried,medium-and small-scale geological targets.Based on the above analysis,the challenge in the signal enhancement of gravity gradient data is that the detailed information and high-frequency noise are difficult to be distinguished,and the weak signals caused by the small-scale geological targets are easily annihilated in the background and noise.For the problems exist in data processing,three types of commonly used filters are utilized for the filtering of model gravity gradient data.According to results of filtering effect evaluation methods,mean square error and recovery of anomaly shape,Gaussian filter has a preferable suppression effect on Gaussian noise as a spatial smoothing technique.On this basis,in the subsequent study,Gaussian filter is compared with the new filtering method proposed in this thesis.Wavelet is commonly used in geophysical data filtering,but there are a couple of problems in the conventional filtering scheme.The possible results of these problems are analyzed in this thesis.And in applications translation-invariant wavelet is proposed to decompose data contaminated with noise.Adaptive threshold is used for signal-to-noise separation according to noise level on each decomposition scale.And finally based on the characteristics of common soft thresholding and hard thresholding and their defects,a new mixed thresholding is constructed to process wavelet coefficients.The above improvement suppresses the pseudo-Gibbs phenomenon,retains more signal details,and avoids over smoothing of the filtering results,which leads to better results than Gaussian filter.For local features of anomaly,two feature enhancement methods are proposed.The direct downward continuation is unstable,and the depth of continuation is insufficient.In this thesis,the regularization term is added to the continuation equation,which can effectively improve the stability of the calculation and counteract the noise amplification effect.Additionally,for the linear geological structures reflected by gravity gradient data,feature enhancement is performed using directional filtering.Gaussian filter is added into the calculation formula to reduce noise interference effectively.And clearer,more continuous,higher resolution structure trend features are achieved.The fast imaging method is widely used in potential field data processing and has the advantages of being insensitive to noise,fast calculation,and small memory usage.It can perform real-time imaging with observed data.In this thesis,migration imaging and generalized linear inversion imaging are studied.To solve the problem of obvious skin effect and insufficient depth resolution in current applications,depth weighting function is introduced into imaging calculations.The functions used in DEXP imaging based on structural index and in the inversion are added to the imaging formula.Analyze the mechanism of the two weighting functions,and the latter one is improved to increase the a priori information on the depth of the source so as to improve the resolution in depth direction.However,when dealing with tensor data,migration imaging is only effective for T_xx,T_yy,and T_zz components.The generalized linear inversion imaging formula does not contain the resolution matrix and is simpler than migration imaging calculation.Model test proves that generalized linear inversion imaging constrained with improved depth weighting function has a good imaging effect on six tensor components.Different information about the distribution of sources that each component contains is exhibited effectively,providing more constraint information for inversion and interpretation,which can further improve the problem of multi-solution in inversion.Three-dimensional density inversion interprets the full-tensor gravity gradient data quantitatively,and the spatial location and residual density distribution of the source are obtained.However,a severe problem of multi-solution exist.Starting from the reducing multiplicity of solutions,the focusing inversion algorithm is used for target bodies with sharp boundry firstly.The residual density distribution with better convergence than smoothing inversion is obtained.Secondly,the existing and the calculated distribution range of the source along horizontal and vertical directions are taken as the geometric constraints,and the residual density range of the source is taken as the physical property constraints.The constraints are added to the inversion.Model test shows that the resolution of the inversion results is significantly improved along both horizontal and the vertical directions.Result with focusing source distribution,clear boundary location and reasonable residual density is obtained in real data application.The existing data is reasonably used and information contained in data is fully excavated in the above inversion method.Constraints are imposed on the inversion process,which effectively improves the inversion accuracy,and provides a reliable basis for geophysical modeling and geological interpretation.