Seismic Bandwidth Extension via Recurrent Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current geophysical prospecting methods for hydrocarbon management are limited by the narrow frequency bandwidth of seismic data, which restricts the accuracy of subsurface property inversion due to noise interference, particularly in ultra-low and high-frequency ranges, and existing solutions like ocean bottom nodes and ultra-shallow tow systems are costly and less used.
Innovation Solution
A computer-implemented method using a machine learning model, specifically a sequence-to-sequence architecture with recurrent neural networks, to identify correlations between frequency bands and predict frequency responses, enabling the extension of seismic data bandwidth without the need for expensive acquisition systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If towed streamer acquisition is used, then cost is reduced and ease of operation is improved, but frequency bandwidth is limited and measurement precision deteriorates
Solution Approach 1:
The patent creates synthetic copies of missing frequency components through machine learning. The system learns correlations between available frequency bands and generates synthetic representations of unavailable ultra-low and high frequencies, effectively copying the spectral characteristics needed for full bandwidth without physical acquisition of those frequencies
Solution Approach 2:
The patent replaces the mechanical/physical acquisition system (specialized ultra-shallow tow equipment or OBN nodes) with a computational approach. Instead of using complex hardware to capture full bandwidth, the system uses machine learning algorithms to synthesize missing frequency information from available data, substituting computational processing for physical measurement capabilities
2Measurement precision
If special acquisition systems like OBN or ultra-shallow tow are used, then frequency bandwidth is extended and measurement precision is improved, but cost increases significantly
Solution Approach 1:
The patent uses standard, widely-available towed streamer equipment instead of expensive specialized systems. The approach treats the limited-bandwidth data from conventional equipment as sufficient input, processing it through machine learning to achieve results comparable to expensive specialized acquisition without the high equipment costs
Solution Approach 2:
The system synthesizes the expensive specialized acquisition data through computational copying. By learning from examples or through physical principles, the machine learning model generates synthetic representations of what the specialized equipment would have measured, making the expensive data accessible through inexpensive conventional equipment
3Quantity of substance
If conventional towed streamer acquisition is used, then cost is reduced, but low and high frequency data quality deteriorates due to noise
Solution Approach 1:
The patent extracts useful signal information from the noisy conventional data by separating it from noise through machine learning. The system identifies and extracts correlation patterns that represent genuine subsurface signals while filtering out noise components, effectively isolating the useful information from the contaminated input data
Solution Approach 2:
The machine learning model acts as an intermediary processing layer between the noisy conventional data and the final high-quality output. This intermediary learns to transform the low-quality input by identifying patterns and relationships, mediating the transition from noisy conventional data to clean synthetic bandwidth extension
Data Source
AI summary
A methodology for extending bandwidth of geophysical data is disclosed. Geophysical data, obtained via a towed streamer, may have significant noise in a certain band (such as less than 4 Hz), rendering the data in the certain band unreliable. To remedy this, geophysical data, from a band that is reliable, may be extended to the certain band, resulting in bandwidth extension. One manner of bandwidth extension comprises using machine learning to generate a machine learning model. Specifically, because bandwidth may be viewed as a sequence, machine learning configured to identify sequences, such as recurrent neural networks, may be used to generate the machine learning model. In particular, machine learning may use a training dataset acquired via ocean bottom nodes in order to generate the machine learning model. After which, the machine learning model may be used to extend the bandwidth of a test dataset acquired via a towed streamer.