NMR Fluid Typing Using Independent Component Analysis Applied to Water-Oil-displacement Laboratory Data

摘要

In low field NMR relaxometry for hydrogen,the T1 or T2 spectra of core samples are mixtures of the individual contributions of the saturating phases filling the pore space,either water or hydrocarbon. Fluid typing(identification and quantification of each phase)becomes challenging,however,when individual spectra overlap within a certain relaxation time window. A traditional fluid typing approach,based on the diffusion effect,is complex and expensive. It requires magnetic field gradient,multiple radio frequencies,and a special fit-for-purpose inversion software. This approach applies to T2 only;T1 is not affected by diffusion relaxation mechanisms. Independent component analysis(ICA)has been tested as a means of solving this problem. ICA is a blind source separation(BSS)technique that yields statistical independent T2 spectral source components of the present fluids,without previous knowledge of either relaxation mechanisms or fluid properties. The technique can also be applied to T1. In a water-oil drainage experiment,beginning with 100% oil saturation,we measured the T2 spectra of the samples at five water saturation stages,to nearly 97%,and used these spectra as input for the BSS-ICA processing. As output,we obtained the statistical independent T2 spectra for oil and water and a mixing matrix containing the individual weights that define the measured T2 spectra at any acquisition step. The results show close correlation between the oil-T2 ICA-component and the T2 spectrum of the 100% oil saturated sample. The ICA-derived water saturations show a good match with reported experimental values. The water-T2 ICA-component shows a bimodal behavior,reflecting the pore size distributions for T2 shorter than 400ms and the bulk water for T2 values greater than 1000ms. The bulk relaxation is caused by the presence of remaining oil layers that prevent a magnetic interaction between water molecules and the pore surface.