InterPore2022

An optimization-based method for upscaling carbonate volumetric image data

31 May 2022, 09:20

1h 10m

Poster Presentation (MS10) Advances in imaging porous media: techniques, software and case studies Poster

Speaker

Mr Anup Kumar Shahi (IITB-Monash Research Academy)

Description

Extraction of information from volumetric core data is highly dependent on the quality of the acquired images and processing. Segmentation of the 3D image helps separate the pore networks from the rock matrix using the process of binarization. The processing done by the user turns out to be subjective and there exists a trade-off between resolution and field of view for the features of interest. The need for a principled way of processing these images is of utmost requirement in the rock physics modelling pipeline.

So, in this paper, we develop a simple optimization-based routine to improve the resolution of the 3D volumetric images using laboratory-based data such as mercury porosimetry. Optimization algorithms work by minimizing the error between the observed and modelled data. Solutions obtained by such algorithms are a function of initialization and the cost function. Regularization is often required when dealing with nonlinear data such as the arbitrary shapes of the pores. It is known that the porosimetry derived capillary pressure vs saturation curve is indicative of the pore size distribution in a rock core sample. This curve can be exploited to enhance the upscaling process using multi-point statistics (MPS). Conventionally, MPS mandates a careful selection of kernel parameters for capturing the spatial variation in volumetric data. The MPS method works by sequentially populating a 3D grid to emulate the observed 3D image. However, finding optimum kernel parameters is crucial to capturing the spatial characteristics. Also, when dealing with multiple images, finding a single set of kernel parameters might not be a trivial task. We show that the selection of these kernel parameters can be enhanced using the pressure vs saturation curve of the MICP data when formulated as an optimization problem and minimizing for this curve.

We test the proposed methodology on carbonate rock sample data show the results on multiple 3D samples and evaluate the upscaling performance using statistical metrics. For the workflow, a low-resolution 3D volume image sample is acquired using the X-ray microtomography instrument which was then subjected to MICP simulations and pore-scale statistical analysis. The result of the application of such a method naturally adheres to the pore size distribution of the samples while giving the user the confidence of real laboratory-based data.