Geobody Prediction with Guide Input Tiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional machine learning approaches in hydrocarbon exploration struggle to incorporate 'ground truth' data for geobodies into their models without requiring extensive retraining, which is costly in terms of computation time and does not guarantee improved predictions.
Innovation Solution
The implementation of a learning machine system that accepts post-stack seismic data and additional ground truth information through guide input tiles, allowing for iterative improvements in geobody predictions and enabling the use of labeled and unlabeled regions within the guide input tiles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional machine learning approaches are used to detect geobodies based on seismic data, then the model can operate with existing training data, but the model cannot incorporate new ground truth information without costly retraining that does not guarantee improved predictions
Solution Approach 1:
The system performs preliminary actions by pre-processing ground truth data into guide input tiles with labels indicating geobody presence or absence before the actual prediction process. This preparation allows the model to directly utilize new ground truth information during inference without requiring retraining, thereby resolving the contradiction between adaptability and time loss.
Solution Approach 2:
The patent introduces guide input tiles as an intermediary mechanism between ground truth data and the machine learning model. These tiles contain labeled regions that guide the model's predictions, enabling the incorporation of new ground truth information without direct model retraining. This intermediary structure allows flexible adaptation while avoiding the time-consuming retraining process.
2Measurement precision
If ground truth data is incorporated into the machine learning model, then prediction accuracy improves, but the system complexity increases due to the need for guide input tiles and labeling mechanisms
Solution Approach 1:
The system applies local quality by creating guide input tiles that contain labels only for specific regions where ground truth information is available, rather than requiring comprehensive labeling of entire seismic volumes. This localized approach improves prediction accuracy in targeted areas while minimizing the overall system complexity burden.
Solution Approach 2:
The patent implements partial action by allowing the system to function with incomplete ground truth labeling. Guide input tiles can contain labels for only certain regions of interest, and the model can still produce predictions for unlabeled regions based on seismic data alone. This partial labeling approach achieves improved accuracy where needed without the complexity of complete system redesign.
3Reliability
If extensive retraining is performed to incorporate new ground truth information, then the model may improve predictions, but the computational cost and time investment increase significantly
Solution Approach 1:
The system implements self-service by enabling the machine learning model to automatically incorporate new ground truth information through the guide input tile mechanism during the inference process itself, without requiring external retraining operations. This self-adjusting capability maintains prediction reliability while eliminating the energy-intensive retraining process.
Data Source
AI summary
Some implementations may include a method for detecting, by a learning machine, a geobody in a seismic volume. The method may include receiving a first seismic input tile representing first seismic data from the seismic volume; receiving a first guide input tile including first labels that indicate presence of the geobody in a respective region in the seismic volume or absence of the geobody in the respective region, and one or more unlabeled regions that make no indication about presence or absence of the geobody; and determining, based on the first seismic input tile and the first guide input tile, a first prediction about geobody presence or absence in the seismic volume.