Layered Multi-exponentisH Inversion Method Research Of Magnetic Resonance Sounding Data

Magnetic resonance sounding (MRS) is a widely concerned geophysical method thatallows direct determination of water distribution in the subsurface. This method throughartificial excitation electromagnetic fields generated by an antenna loop to formatmagnetic moment of hydrogen nuclei in groundwater using its paramagnetic, thenreceiving coil induced electromagnetic signals generated by precession of the magneticmoment. We can obtain information like water content, depth and thickness ofunderground aquifers after signal inversion. MRS method has advantages withuniqueness of solution, quantization and high efficiency and so on. Smooth inversion andblock inversion are two main methods for surface MRS at present, which can constructwater content and average effective pore size of aquifer through initial amplitudes andmean relaxation times, respectively. However, these inversion methods are limited inporous geological environment and in some case the interface of aquifers can not beportrayed clearly due to the loss of valid information during mono-exponential fitting. Inthis paper, we focused on multi-exponential characteristics of MRS signal, studied usingrelaxation time T2spectrum to describe underground pore size distribution, and then putforward the layered multi-exponential inversion method.The relaxation time and initial amplitude of MRS signal are directly interrelatedwith the water content and effective pore size in the subsurface. Traditional methoddescribe the attenuation of signal through the average relaxation time T2*, which is notconsistent with the actual complex and porous medium underground environment. Westudied multi-exponential decomposition method based on inequality constraint in thispaper, described the specific pore size through relaxation time T2spectrum. As therelaxation time affected by noise easily, we improved the T2spectrum extraction accuracyand anti-interference ability through the weighted correction. Then compared MRS signalfitted by mono-exponential type with multi-exponential type, found out that multi-exponential fitting can reflect the signal attenuation process better and restoreinitial amplitude more accurate. The study in this paper will have important referencevalue in water content inversion and rock pore size distribution assessment.We studied existing inversion methods on abroad include Time step inversion andQT inversion. These two methods have some defects. For time step inversion, first, thetotal water content obtained by superposition water content under each time point, easy tocause water content overestimation. Second, relaxation time T2depends on the results ofwater content inversion, which made the method lack of stability. For QT inversion, first,this method using smooth stratum mode, turned out stratum transition smoothly. This isnot conducive to ensure the specific position of the aquifer. Second, there are too manyinversion variables exist in the smooth type, the computation is very time-consuming.Considering all these inversion methods, we studied layered multi-exponential inversionmethod, which is found to solve inversion problem on the basis of MRS signal inherentmulti-exponential behavior, by taking the complete MRS signal datasets into account inthis paper. The multi-exponential MRS inversion of porous geological environment wasrealized through decomposing the relaxation time, deducing a new magnetic resonanceforward kernel function, combined with the aquifer layered mode and then build upTikhonov functional minimum equation. MRS all relaxation signal data taken into theinversion calculation, directly estimates the partial water content distribution and thusprovides the water content distributed with relaxation time and depth. Layered model innew inversion method given aquifer boundaries clearly, reduced the inversion variableseffectively and improved the efficiency of the inversion, which made this methodapplicable to the interpretation of complex geological environment. We establishedmulti-aquifer mono-exponential model and multi-aquifer multi-exponential model toverify the effectiveness of the new inversion method.Finally, two measurement examples are given in the paper. We used initial amplitudeinversion, QT inversion and layered multi-exponential inversion method to calculate thesetwo point data, obtained water content distribution, then estimated the water inflow, compared with drilling results of measuring point. Field test examples further verify thevalidity of new inversion method. The study in this paper provided a new thought andmethod for surface magnetic resonance sounding.