Backpropagation Model Iterative Training for Subsurface Feature Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional backpropagation-enabled processes for identifying subsurface features from seismic data suffer from human error and bias in field-acquired data interpretation, leading to time-consuming labeling and high false positive rates.
Innovation Solution
A method that iteratively trains a backpropagation-enabled model using a combination of initial training data and computed inferences, enforcing consistency by updating the training data with target data and associated inferences, thereby reducing false positives and improving inference quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If field-acquired seismic data is used for training backpropagation-enabled processes, then the model can be trained with real-world data, but human error and bias are introduced leading to high false positive rates
Solution Approach 1:
The patent implements a feedback mechanism where the backpropagation-enabled process continuously refines its predictions by comparing initial inferences against updated training data. The system uses feedback from consistency checks and iterative retraining to correct false positives and improve identification accuracy.
Solution Approach 2:
The patent creates synthetic seismic data that copies the characteristics of real field-acquired data while eliminating human error and bias. These synthetic datasets are generated by applying geological models and forward modeling techniques to create realistic seismic signals without the contamination of human interpretation errors.
2Measurement precision
If manual labeling of field-acquired data is performed, then the training data can be annotated, but the process is time-consuming and resource-intensive
Solution Approach 1:
The patent implements self-service through automated labeling using the backpropagation-enabled process itself. The system generates its own training labels by performing consistency checks and iterative retraining, eliminating the need for manual annotation while maintaining high labeling accuracy.
Solution Approach 2:
The patent uses synthetic seismic data that can be automatically labeled without human intervention. The synthetic data is generated with known ground truths from geological models, providing ready-to-use training labels that eliminate time-consuming manual annotation processes.
3Productivity
If conventional training methods are used with initial training data, then the model can be trained initially, but the model generates false positives and requires iterative improvement
Solution Approach 1:
The patent implements continuous improvement through iterative retraining cycles. The backpropagation-enabled process continuously refines its predictions by repeatedly training on updated datasets that incorporate both initial training data and newly generated synthetic data, ensuring continuous improvement in accuracy without sacrificing productivity.
Solution Approach 2:
The patent changes key parameters in the training process by transitioning from static initial training data to dynamic datasets that are continuously updated. The system modifies training parameters including learning rates, data augmentation factors, and consistency thresholds to optimize both speed and accuracy throughout iterative training cycles.
Data Source
AI summary
A method for improving a backpropagation-enabled process for identifying subsurface features from seismic data involves a model that has been trained with an initial set of training data. A target data set is used to compute a set of initial inferences on the target data set that are combined with the initial training data to define updated training data. The model is trained with the updated training data. Updated inferences on the target data set are then computed. A set of further-updated training data is defined by combining at least a portion of the initial set of training data and at least a portion of the target data and associated updated inferences. The set of further-updated training data is used to train the model. Further-updated inferences on the target data set are then computed and used to identify the occurrence of a user-selected subsurface feature in the target data set.