Research On Three-dimensional Block Imaging Method Of Surface Nuclear Magnetic Resonance Based On Horizontal Smoothness Constraint

Magnetic Resonance Sounding(MRS)is the only geophysical method allowing direct,noninvasive groundwater investigations from the surface at present.However,for the groundwater in the complicated geological condition,SNMR has many problems on low operating efficiency,low imaging precision and also unstable imaging results.Therefore,development of new three-dimensional(3D)imaging method for complex geophysical settings has been the urgent demand.The study of the thesis includes 3D SNMR forward modeling with high precision,optimized Block inversion for SNMR based on prior information and 3D Block inversion method using horizontal smoothness constraint.The main research and achievements are as follows.1.For requirements of accuracy while accomplishing 3D SNMR forward modeling,Hammer integration algorithm was proposed based on unstructured non-uniform tetrahedral mesh.Firstly,finite element method was used to calculate the 3D magnetic field in the subsurface.Sensitivity kernel expression was deduced.With conventional mesh grid,the grid size was large with poor precision.On the other hand,the smaller grid size led to enormous calculation task with refined mesh grid.Therefore,Hammer integration was applied to calculate the kernel function with conventional mesh grid.Synthetic results including 3D kernel function and aquifer models show that with Hammer integration fewer grids,nodes and higher accuracy could be achieved.3D forward model could be calculated efficiently.2.An optimized Block inversion method was proposed in order to satisfy the high accuracy requirement of the aquifer boundary with SNMR in the thesis.The priori information is estimated from the derivative result of the Smooth inversion,which provides the smooth distribution of the water content at different depths.Based on the Tikhonov regularization inversion theory,the iterative scheme of Block inversion is deduced and solved.Block inversion provides the exact boundary and water content of the aquifer.The introduction of the priori information resolve the instability and low reliability compared with traditional Block inversion.A series of synthetic models with one or more aquifers under different blocks,noise level,depth and thickness were simulated.Results show that the accuracy of the aquifer boundary from Block inversion is superior to Smooth inversion and the anti-interference ability and stability is stronger.The optimized Block inversion method based on the priori information satisfies the requirement of inversion accuracy and stability.In addition,the method could provide more details for an aquifer with complex internal structure.3.When used to invert 3D SNMR data,the current fixed geometry inversion(FGI)method cannot clearly identify the boundaries of water-bearing structures and can easily cause artifacts.In the thesis,we propose a horizontal smoothness-constrained variable geometry inversion(VGI)method that allows the occurrence of sharp boundaries in the vertical direction and provides easier identification of the geometric boundaries of 3D water-bearing structures.The imaging quality in low-resolution areas is improved and the water content within water-bearing structures is more evenly distributed due to the use of 3D kernel function and horizontal smoothness constraints.4.In order to achieve optimum VGI results,suitable element size should be selected.Synthetic results show that an excessively large element size leads to relatively coarse imaging results,and an excessively small element size leads to relatively fluctuations in the imaging results.The real model cannot be reflected in either of these situations.In addition,the VGI model with the kernel function in a moderate element size has far fewer parameters than the FGI model as well as a low memory footprint when calculating the kernel functions and a high inversion calculation speed.Based on the study of the above contents,a set of 3D imaging methods for complicated geological condition is presented.The main innovation lies as follows.1.3D SNMR forward modeling based on Hammer integration was proposed for the first time.The Hammer integration algorithm is based on unstructured non-uniform tetrahedral mesh,with which fewer grids,nodes and higher accuracy could be achieved.Forward kernel function could be calculated efficiently.2.An optimized Block inversion method based on priori information was proposed in the thesis.The introduction of the priori information resolves the instability and low reliability compared with tradition Block inversion,with which exact boundary and water content of the aquifer is achieved.A series of synthetic models and field tests had taken place.Results show that with prior information the optimized Block inversion satisfies the requirement of inversion accuracy and stability.3.A horizontal smoothness-constrained VGI method was introduced.With the inversion method,imaging quality in low-resolution areas is improved and the water content within water-bearing structures is more evenly distributed.In addition,the VGI model with the kernel function in a suitable element size could achieve satisfactory 3D imaging results.The SNMR imaging method and hardware have been applied in engineering practice,like the groundwater detection for Shaoguo town in Changchun city,Sifangtuozi town in Baicheng country and Wenzhou Zeya Tunnel.A three-dimensional SNMR survey was also performed on an artificial lake in Germany.All inversion results are verified by geophysical data and methods such as electrical and drilling method,which proved the accuracy and reliability of the SNMR imaging method proposed by the thesis.