Stochastic reconstruction, 3D characterization and network modeling of chalk

Abstract

Systematic studies involving stochastic reconstruction, geometric and topological characterization, and network modeling of chalk, aiming at computation of petrophysical properties, are reported. The numerical chalk models are constructed exclusively from limited morphological information obtained from 2D backscatter scanning electron microscope images of the microstructure. Two different stochastic reconstruction methods are considered: conditioning and truncation of Gaussian random fields (GRF), and simulated annealing (SA). The potential of initializing the simulated annealing reconstruction with input generated using the Gaussian random fields method is evaluated and found to accelerate significantly the rate of convergence of simulated annealing reconstruction. This finding is important because the main advantage of simulated annealing method, namely its ability to impose a variety of reconstruction constraints, is usually compromised by its very slow rate of convergence. A detailed description of the chalk microstructure in the form of 3D volume data is essential for the prediction of petrophysical properties from first principles. The prediction of absolute permeability and formation factor directly from such information are considered first. The prediction of absolute permeability, formation factor and mercury–air capillary pressure curves are then considered using approximate network models constrained by information (pore- and throat-size distributions, coordination number) obtained from geometric and topological analysis of the reconstructed pore networks. Such information is extracted from the 3D volume data using morphological skeletonization and pore space partitioning methods. Very good agreement between the predicted and measured data is found for samples of North Sea chalk. On the basis of this study, it is concluded that (a) stochastic reconstruction from limited morphological information reproduces the essential features of pore geometry and connectivity of chalk, and (b) network modeling techniques can be used to predict petrophysical properties of chalk based on geometric and topological information of the stochastically reconstructed media.