ACCELERATING FORMATION PETROPHYSICAL PROPERTY PREDICTION FROM LOG-DERIVED ROCK MODELS USING GPU
Publication Date: 2010-Dec-17
The IP.com Prior Art Database
The methods and systems described herein are directed at using graphics processing units (GPU) to accelerate the prediction of in-situ formation petrophysical properties from downhole logging measurements and/or geological information. The formation properties include at least those not measured directly from downhole logging measurements, such as absolute permeability, relative permeability, resistivity formation factor and index, capillary pressure, mechanical properties, and so on. In one aspect, computational methods used to derive these properties from an earth formation model are implemented on a GPU-based architecture to achieve dramatic computational performance gain as compared to those using just the central processing unit (CPU). The numerical methods implemented on the GPU include at least those calculating the properties mentioned above, such as the pore network model, lattice Boltzmann flow simulation, finite difference method and finite element method. In another aspect, the formation model provided to the GPU is a three-dimensional digital rock sample reconstructed from downhole logging measurements and/or geological information. The model represents the earth formation at a given depth, and accounts for different rock-forming diagenetic processes such as sedimentation, compaction, and precipitation or dissolution of carbonate and clay minerals. Algorithms related to model reconstruction are also implemented on a GPU platform to utilize the high computational performance of GPU.
Page 01 of 4
Record Number: 5006 Invention Infoview Form
A. INVENTION OVERVIEW
BHI File Number: Product Line: *
TLE4-51198-US D&E2 - Wireline
Date Started :08/04/2010
(PLG) Prod. Logging/DH Completion
Title of Invention:
Accelerating Formation Petrophysical Property Prediction From Log-Derived Rock Models Using GPU
Abstract of the Invention: Market Segment: Market
This invention is directed towards the use of modeling methods that enable the
construction of a representative formation rock model from downhole logging
measurements and geological information, and the use of graphics processing units (GPU)
to accelerate the calculation of rock properties from the model that are not measured by
the formation evaluation logging tools. The combination of rock modeling approaches and
GPU implementation allows to predict quickly (within one or several minutes) and
accurately petrophysical properties (e.g., permeability, relative permeability, resistivity
formation factor, and capillary pressure) directly from downhole logging measurements.
Product/Service Name (if known):
B. THE INVENTOR(S)
FirstName LastName MiddleName CostCenter EmployeeNumber WorkEmail Status
Guodong Jin 410300613 00128376 Guodong.Jin@bakerhughes.com Approved
Sergey Martakov 410300613 00117067 Sergey.Martakov@bakerhughes.com Approved
Non BHI Employee Inventor(s):
C. DETAILED DESCRIPTION OF THE INVENTION(S)
1. Describe the invention in detail. Describe the preferred manner of practicing the invention including listing the steps, method,
composition and operation of the invention. Point out features that are believed to be new and how the invention overcomes the
disadvantages of the old manner. Attach electronic documents such as sketches, drawings, photographs, logs, etc., as needed.
Non-electronic documents must be sent to the Legal Department representatives. * In-situ formation petrophysical properties are of
great importance in the petroleum industry not only to evaluate reservoir potential, but also to support drilling strategies in terms of borehole
stability, completion decisions related to sand production, as well as the potential for hydraulic fracturing and perforation. Some properties,
such as permeability, relative permeability, resistivity formation factor, and capillary pressure, are very difficult, if not impossible, to measure
directly using current sub-surface logging technology. Traditional laboratory measurements are often costly, time-consuming and require
large amounts of core materials that may not always be available at the time when drilling and completion decision have to be made.
Numerical techniques currently used, which rely on as an input the micro-CT images (require a core sample) or generated numerical rock
models (need information obtained on a thin section made either from a reservoir core plug, wall cores or drill cuttings), also demand large
computational resources (memory and time) and require days...