Noeradi, John Adler; Pongga Dikdya Wardaya; Lilik Hendrajaya; Bagus Endar (2012) Porosity and Mineral Fraction Estimation of Carbonate Rock with an Integrated Neural Network / Image Processing Technique.
academeia_60-473-2-PB.pdf
Download (1MB)
Abstract
Porosity and mineral fraction information are crucial in reservoir characterization, however the exact value of these parameters is difficult to measure. We propose a new method for estimating the porosity and mineral fraction of carbonate rock from thin section images using an integrated neural network/image processing technique. Neural networks were built and trained to classify porosity and minerals of carbonate (calcite and dolomite) based on their color after chemical treatment. Pixel values of these colors were attributed with a target code value and represented in a 2D image (matrix) from which a simple image processing pixel filtering and counting algorithm was employed to calculate each fraction. Computation time was less than 40 seconds and classification error was less than 2%. This method may be useful as a cost-effective alternative for estimating 2D-porosity and mineral fraction for thin section images of rock. Unlike porosimetry or X-ray diffraction (XRD) measurements, this method does not require liquid injection at the coreplug scale.
Porosity and mineral fraction information are crucial in reservoir characterization, however the exact value of these parameters is difficult to measure. We propose a new method for estimating the porosity and mineral fraction of carbonate rock from thin section images using an integrated neural network/image processing technique. Neural networks were built and trained to classify porosity and minerals of carbonate (calcite and dolomite) based on their color after chemical treatment. Pixel values of these colors were attributed with a target code value and represented in a 2D image (matrix) from which a simple image processing pixel filtering and counting algorithm was employed to calculate each fraction. Computation time was less than 40 seconds and classification error was less than 2%. This method may be useful as a cost-effective alternative for estimating 2D-porosity and mineral fraction for thin section images of rock. Unlike porosimetry or X-ray diffraction (XRD) measurements, this method does not require liquid injection at the coreplug scale.
| Item Type: | Article |
|---|---|
| Additional Information: | Migrasi dari GDL. Identifier: jbptunikompp-gdl-johnadlerp-32893. Folder: Member / john.adler007@gmail.com |
| Uncontrolled Keywords: | Backpropagation; Lavenberg-Marquadt; mean square error; convergence |
| Divisions: | ?? Member_GDL ?? |
| Depositing User: | Robot Migration |
| Date Deposited: | 05 Nov 2026 11:00 |
| Last Modified: | 05 Nov 2026 11:00 |
| URI: | https://elibrary.unikom.ac.id/id/eprint/43194 |
