ML-Based Seismic Source Separation Parameterization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seismic data processing techniques face challenges in efficiently differentiating primary signals from interference and background noise, particularly in multi-source domains where strong interference noise can obscure weak coherent signals, leading to suboptimal source separation and requiring manual parameterization that is computationally intractable for large-scale datasets.
Innovation Solution
A method utilizing a sparsity-promoting transform domain combined with machine learning, specifically a convolutional neural network with classification, regression, and segmentation heads, to automatically select threshold values and enhance signal separation by identifying high and low signal-to-noise ratio areas, thereby reducing signal loss and stabilizing source separation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual parameterization is used for source separation, then processing accuracy can be controlled, but computational complexity and time requirements become intractable for large-scale datasets
Solution Approach 1:
The system uses machine learning models to automatically select threshold values and parameters for source separation, allowing the system to self-configure without manual intervention. The ML models process seismic data, identify signal characteristics, and determine optimal parameters autonomously, eliminating the need for manual parameterization while maintaining high separation accuracy.
Solution Approach 2:
The patent replaces manual mechanical parameter selection with automated machine learning-based parameter selection. Instead of requiring geoscientists to manually adjust parameters based on geological knowledge, the system uses trained ML models to automatically determine threshold values and separation parameters, substituting human expertise with automated intelligent systems.
2Productivity
If strong interference noise is present in multi-source domain, then acquisition efficiency is improved, but signal differentiation becomes difficult and separation quality deteriorates
Solution Approach 1:
The system dynamically adjusts threshold parameters and separation criteria based on the characteristics of interference noise and signal strength. The machine learning models analyze the seismic data to determine optimal parameter values for each specific case, allowing the system to adapt to varying noise conditions and maintain high differentiation accuracy even in challenging multi-source environments.
Solution Approach 2:
The machine learning models continuously evaluate the separation results and use this feedback to refine future parameter selections. The system learns from the characteristics of separated signals and interference patterns, adjusting its threshold values and separation strategies to better handle strong interference conditions while maintaining acquisition efficiency.
3Loss of time
If automated machine learning is applied to select threshold values, then processing time is reduced, but model complexity and computational requirements increase
Solution Approach 1:
The machine learning models are pre-trained on extensive datasets of seismic data with known characteristics and separation ground truth. This preliminary training allows the models to make accurate parameter selections quickly during actual processing, reducing the need for complex real-time computations while maintaining high accuracy in threshold value selection.
Solution Approach 2:
The system uses pre-trained machine learning models that have learned separation patterns from training data, effectively copying the expertise of manual parameterization into automated algorithms. The models replicate the decision-making capabilities of experienced geoscientists but execute them automatically and consistently, reducing processing time while managing computational complexity through efficient model architectures.
Data Source
AI summary
Systems and methods may use machine learning to automate the parameterization process for multi-stage iterative source separation. Seismic signals that are generated by a plurality of sources are received by a plurality of sensors within a field as a blended signal. An automated machine learning model that has been trained on blended and unblended signals determines if the incoming blended signal has a relatively high or low signal to noise ratio and then selects a threshold value based on the detected signal to noise ratio. The blended signal is then separated according to the source of the seismic data. A seismic image based on the separated seismic data is then generated which can then be used to adjust one or more control parameters in a machine or tool within the field.