Subsurface Joint Network Prediction From Outcrop Geostatistics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining the characteristics of joint networks below the surface in subsurface rock formations is challenging due to their inherent complexity and variability, which affects petroleum engineering and mining operations by influencing fluid flow and geomechanical responses.
Innovation Solution
A method using machine learning networks to analyze outcrop pavement images, determine geostatistical properties, and generate a synthetic joint network predictor to predict subsurface joint networks based on observed subsurface joints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning networks are used to analyze outcrop pavement images and generate synthetic joint network predictors, then prediction accuracy of subsurface joint networks is improved, but system complexity and computational requirements increase
Solution Approach 1:
The system divides the complex prediction task into multiple stages: (1) training phase using outcrop pavement images to build the synthetic joint network predictor, and (2) prediction phase using the trained predictor on subsurface data. This segmentation allows the complex ML model to be developed and validated separately from its application, reducing overall system complexity while maintaining high prediction accuracy.
Solution Approach 2:
The system performs preliminary action by training the synthetic joint network predictor using outcrop pavement images and geostatistical properties before actual subsurface prediction is needed. This pre-training phase creates a ready-to-use model that can quickly predict joint networks in subsurface layers without requiring complex real-time computations during drilling operations.
2Reliability
If comprehensive geostatistical properties are used to train the synthetic joint network predictor, then prediction reliability is improved, but data processing time and computational resources increase
Solution Approach 1:
The system extracts and processes comprehensive geostatistical properties from outcrop pavement images during the offline training phase, before actual subsurface prediction is needed. By pre-processing and storing these properties in a trained ML model, the system achieves high prediction reliability without requiring extensive data processing during time-critical drilling operations.
Solution Approach 2:
The system creates a synthetic joint network predictor that copies and encodes the statistical patterns and properties learned from extensive outcrop pavement image analysis. This copied knowledge allows rapid prediction of subsurface joint networks without re-processing the original comprehensive datasets, maintaining reliability while reducing processing time.
3Loss of information
If machine learning networks process large volumes of outcrop pavement images and subsurface data, then prediction completeness is improved, but computational energy consumption increases
Solution Approach 1:
The system segments data processing into offline training phase (processing large volumes of outcrop images to build comprehensive models) and online prediction phase (using the trained model for efficient subsurface prediction). This segmentation ensures prediction completeness is achieved during training while energy consumption is minimized during actual drilling operations.
Solution Approach 2:
The system creates a compressed representation (copy) of the comprehensive patterns learned from large volumes of outcrop pavement images. This synthetic predictor copy contains the essential information for complete predictions but requires minimal computational energy to operate during subsurface analysis, as it avoids re-processing the original large datasets.
Data Source
AI summary
Methods and systems for synthetic joint network prediction are disclosed. The method may include obtaining a plurality of outcrop pavement images of a plurality of joint networks and determining, using a first machine learning (ML) network, a plurality of detected joint networks using the plurality of outcrop pavement images. The method further includes determining, for each of the plurality of detected joint networks, a set of surface geostatistical properties and creating a database of surface joint properties including the plurality of sets of surface geostatistical properties. The method still further includes generating a synthetic joint network predictor by training a second ML network, using the database of surface joint properties, to produce a synthetic joint network. In addition, the method includes using the synthetic joint network predictor to predict a predicted joint network from a geostatistical description of observed subsurface joints.