Abstract: Numerical Classification of Sandstone Pore Structure

J. C. Davis, F. W. Preston

Applicability of mathematical models to fluid flow in reservoirs continues to be impeded by lack of adequate methods for describing spatial variability of reservoir-rock properties such as porosity and permeability. Fluid-displacement processes are influenced by spatial variation in these properties at all scales of measurement. A new approach to this problem considers a porous medium to be a realization from a random process. The method generally is applicable at any scale, but has been applied initially to the classification of microscopic pore structure.

An optical microdensitometer is used to measure a succession of closely spaced points along traverses across a pore-grain image. This series of measurements constitutes a signal which characterizes the pore microstructure. Fourier coefficients extracted from this signal serve as numerical descriptors or parameters. These coefficients subsequently are used in numerical-pattern-recognition procedures to classify pore structures. Reservoir-rock samples which appear similar in a qualitative sense are grouped numerically by any of the classification techniques.

The Fourier power spectrum derived from many reservoir rocks can be emulated by a stochastic model based on an autoregressive process. The Fourier coefficients thus can be inverted to yield mathematical models of pore structure. These models in turn can be used to define the spatial variation of reservoir-rock properties which have engineering significance.

AAPG Search and Discovery Article #90972©1976 AAPG-SEPM Annual Convention and Exhibition, New Orleans, LA