3D Mechanical Earth Modeling via Neural Network Data Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current three-dimensional mechanical earth modeling techniques face challenges in accurately characterizing rock properties with high lateral and vertical resolution, especially in fields with anisotropic rock properties and complex structural elements, leading to issues like rock failure, subsidence, and wellbore problems.
Innovation Solution
The integration of seismic and wellbore data using neural networking to correlate rock properties, allowing for the generation of high-resolution rock property cubes that honor structural elements and provide accurate rock deformation and strength properties, even in areas with limited wellbore information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional 3-D rock property modeling is used, then the modeling process is simpler, but the lateral and vertical resolution of rock property characterization is insufficient
Solution Approach 1:
The patent combines seismic data processing with wellbore data integration to create a unified 3-D rock property modeling system. The neural network correlates rock properties across both data sources, merging their complementary strengths to achieve high-resolution characterization that neither method could accomplish alone.
Solution Approach 2:
The patent introduces neural networks as an intermediary computational tool that bridges seismic and wellbore data. The neural network learns complex relationships between the two data types and uses them to predict rock properties at high resolution, acting as a mediator that transforms raw data into accurate property predictions.
2Reliability
If high-resolution rock property characterization is achieved, then rock failure and wellbore problems are reduced, but the amount of data processing and computational resources required increases
Solution Approach 1:
The patent performs preliminary data processing and neural network training on available seismic and wellbore data before actual 3-D modeling. By pre-processing the data and establishing correlation models in advance, the system reduces computational requirements during the actual rock property prediction phase while maintaining high accuracy.
3Area of stationary object
If seismic data alone is used for modeling, then the coverage area is larger, but the vertical resolution and accuracy in specific zones is reduced
Solution Approach 1:
The patent applies local quality by using wellbore data to enhance vertical resolution in specific zones where wells are located, while maintaining broader seismic data coverage for overall field characterization. The neural network integrates these different quality levels of data to produce a unified high-resolution model throughout the entire field.
Data Source
AI summary
A technique includes receiving a first dataset that is indicative of seismic data acquired in a seismic survey of a field of wells and receiving a second dataset that is indicative of wellbore data acquired in a wellbore survey conducted in at least one of the wells. The technique includes determining a mechanical earth model for the field based at least in part on the seismic data and the wellbore data.