A Novel Clustering Method for 2-D Low-Field NMR Spectra Working on Geological Fluid Parameter Estimation

Low-field nuclear magnetic resonance (LF-NMR) technology provides robust technical support for geophysical applications, including reservoir exploration and oil logging analysis. The inversion spectra of LF-NMR signals, particularly in 2-D form, reveal crucial geological information such as porosity, permeability, and other essential geologic parameters. However, the acquisition of geological parameters in geoscience relies on the scientific analysis, interpretation, and division of the inversion results, affecting the accuracy of the detection results. In LF-NMR, geophysical parameters are obtained by interpreting the fluid type and estimating the saturation of the inverted spectrum. Therefore, it is essential to develop corresponding qualitative and quantitative classification strategies, especially in cases involving fluid component aliasing. To address these geophysical issues comprehensively, we have created an improved fuzzy clustering algorithm using the local direction centrality (Fuzzy-CDC) based on 2-D LF-NMR parameter spectra to observe the distribution characteristics of saturation for samples with overlapping states. In addition, a fuzzy membership degree was introduced to enhance qualitative and quantitative abilities in assigning saturations to fluid components. To validate the clustering capability of this approach, we conducted simulation and actual experiments on four-phase fluids and water-gasoline two-phase fluids, respectively, comparing the results with conventional clustering methods. The improved method exhibits superior abilities and provides precise saturation estimation, yielding a relative error of 7.40% in simulation and 11.04% in actual experiments. In conclusion, our research significantly enhanced the analysis capabilities of 2-D LF-NMR geophysical parameters while demonstrating the potential for pore fluid assessment and component classification in field geological exploration.