Study On Theory And Interpretation Methods Of Multi-dimensional NMR Logging

Nuclear magnetic resonance (NMR) logging technology as a new loggingtechnology has been developed in the recent decade years, which has obviousadvantages in reservoir division, porosity calculation and fluid identification.However, given the increasing difficulty in conducting fluid identification due to theoverlap of one-dimensional NMR transverse relaxation time distribution (T2) incomplex fluid zones, the concept of two-dimensional NMR in spectroscopy isintroduced into the NMR logging by well logging experts, and then two-dimensionalNMR technology is set up by expanding the number of fluid characteristic parameterssuch as the longitudinal relaxation time (T1) or diffusion coefficient (D). Based ontwo-dimensional NMR technology, three-dimensional NMR technology is designed tocomprehensively consider fluid parameters—T1, T2and D. Therefore, it is of greatsignificance to carry out the two-dimensional and three-dimensional NMR theoreticaland methodology research for the purpose of the development of the theory, problemsolving in fluid identification, application and promotion of multidimensional NMRlogging.Based on the basic principle and relaxation mechanism of NMR logging, andconsidering the NMR properties of fluids, this paper introduced and conductedresearch on2D/3D NMR logging technology. The mainly points in this paper aimedat the studies of the methods of acquisition signal pretreatment, large sparse matrixinversion and interpretation-evaluation approaches. Firstly, the study of2D/3D NMRtheories and relaxation mechanism is carried out. Secondly, the completion of originalecho trains pretreatment is done through a series of techniques such as signal phasecorrection, coordinate rotation, signal superposition, echoes filtering, etc. In the aspectof inverse theories and methods, the filtering method, Hankel matrix reconstructionbased on SVD decomposition, is highlighted in the paper because of remarkableperformance in reducing the effect of noise for low SNR (signal to noise ratio) datainversion problem. And aimed at the problem of the large amount of multidimensionalNMR echo data, rapid inversion algorithms—SVD, Butler-Reeds-Dawson (BRD) andGlobal-Inversion (GI), which all based on data compression technique, are proposedin the paper to solve large sparse matrix inversion related issues. At last part, for illustrating the multi-dimensional NMR interpretation and application, two cylindricalrock samples from sandstone formation were conducted NMR measurement inlaboratory and two-dimensional constructed maps were used to analyze fluidcomposition. Additionally, we studied reservoir fluid characteristics, fluid evaluationmethod and application results with the example of2D NMR logging measured insome sand shale formation. As to3D NMR technology, the influence of acquisitionparameters in fluid identification was also analyzed by simulation results, whichunder different acquisition parameters and different SNR echoes.