Seismic ML Training Using Reflectivity Images From Real Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The quality of neural network-based seismic data processing is limited by the quality of training data, as simulated data often differs significantly from acquired data, leading to suboptimal processing efficiency and accuracy.
Innovation Solution
Employing real seismic data preprocessing to generate training output data, including steps like denoising, deblending, deghosting, and demultiple processing, followed by interpolation and regularization to create a reflectivity image, which is then used to train neural networks to enhance seismic data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simulated data is used for training neural networks, then training data generation is easier and faster, but the quality and accuracy of seismic data processing deteriorates due to significant differences between simulated and acquired data
Solution Approach 1:
The patent creates a pipeline that copies and processes real seismic data through denoising, deblending, deghosting, and demultiple processing to generate training data. This approach uses actual acquired data rather than simulated data, ensuring the training data accurately represents real seismic conditions while maintaining processing efficiency through automated pipelines.
Solution Approach 2:
The patent performs preliminary processing actions on real seismic data before using it for training. By pre-processing the data through denoising, deblending, deghosting, and demultiple operations, the system prepares high-quality training data that accurately represents real seismic conditions, resolving the contradiction between data generation efficiency and processing accuracy.
2Measurement precision
If complex seismic processing steps are applied to real data, then processing accuracy improves, but computational time and complexity increase
Solution Approach 1:
The patent performs complex processing steps (denoising, deblending, deghosting, demultiple) as preliminary actions during the training data generation phase. By completing these computationally intensive operations beforehand on a subset of data, the system prepares high-quality training data that enables faster and more accurate processing during actual seismic analysis.
Solution Approach 2:
The patent creates a pipeline that copies and processes real seismic data through multiple processing steps to generate training data. This approach uses actual acquired data rather than simulated data, ensuring the training data accurately represents real seismic conditions while maintaining processing efficiency through automated pipelines.
Data Source
AI summary
Methods of seismic data processing employ neural networks and use a reflectivity image based on the acquired seismic data to generate output training datasets. The neural networks thus trained are used for generating production datasets, without ghosts, source effects, multiples and/or populating a predetermined set of bins in inline-crossline plane for a set of offset classes.