Reconstruction of porous media using ISOMAP-based MPS

摘要: Acquiring the structures of porous media is very important when predicting flow properties in media. However, direct measurements 3D microstructures with resolution microns or even nanometers are difficult to achieve due expensive cost high-precision equipment. Therefore, as a typical stochastic simulation method, multiple-point statistics (MPS) was used perform reconstruction based on real volume data scanned by micro-CT. Because ensemble patterns extracted from training image (TI) cannot be embedded into linear space, traditional MPS methods using dimensionality reduction, including filter-based and distance-based pattern simulation, not suitable deal nonlinear situation. A new method isometric mapping, which reduction proposed decrease redundant TIs so that subsequent can faster more accurate for Entropy theory introduced select proper size template balance CPU reconstructed quality. The comparisons between images target show structural characteristics our similar those data. This also has shown advantages quality over reduction.