Deep Learning De-aliasing for Seismic Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods face difficulties in effectively identifying and removing aliasing from seismic data, which hampers accurate interpolation and data processing in hydrocarbon exploration, as traditional algorithms struggle to distinguish and correct aliased signals.
Innovation Solution
The implementation of deep learning algorithms using classification and segmentation processors, trained on both modelled and actual seismic data, to identify aliased signatures in the Fourier domain and generate masks for removing aliasing, thereby improving data accuracy and extending the effective Nyquist frequency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional algorithms are used to process seismic data, then device complexity is reduced, but measurement precision and reliability deteriorate due to inability to effectively identify and remove aliasing
Solution Approach 1:
The patent replaces traditional mechanical signal processing algorithms with deep learning-based classification and segmentation processors. These processors use neural networks to automatically identify aliased signals in the Fourier domain, substituting manual threshold-based methods with intelligent systems that learn optimal detection patterns from training data, thereby improving measurement precision while managing complexity through automated decision-making.
Solution Approach 2:
The patent introduces an intermediary deep learning processor that acts as a mediator between raw seismic data and final processed output. This intermediary component performs the complex task of aliasing identification and removal, generating masks that selectively filter aliased frequencies while preserving valid signals, thus resolving the contradiction by adding a specialized intermediate processing layer.
2Measurement precision
If deep learning algorithms are implemented to identify aliasing, then measurement precision improves, but device complexity and computational requirements increase
Solution Approach 1:
The patent segments the processing task into distinct functional modules: a classification processor that identifies aliased frequencies, a segmentation processor that creates frequency masks, and a final processing stage that applies the masks. This segmentation allows each component to specialize in a specific aspect of aliasing removal, improving overall precision while organizing complexity into manageable, modular units that can be independently optimized.
3Reliability
If traditional processing methods are used, then processing speed is maintained, but reliability deteriorates due to inaccurate aliasing removal
Solution Approach 1:
The patent implements preliminary action by pre-training the deep learning processors on large datasets of seismic signals with known aliasing patterns. This offline training phase prepares the classification and segmentation processors to rapidly recognize and correct aliasing in real-time applications. By performing the computationally intensive learning process beforehand, the system achieves high reliability during actual processing while minimizing time loss during operational use.
Data Source
AI summary
A method includes receiving modelled seismic data that is to be recognized by the at least one classification and/or segmentation processor. The modelled seismic data can be represented within a transform domain. The method includes generating an output via the at least one processor based on the received modelled seismic data. The method also includes comparing the output of the at least one processor with a desired output. The method also includes modifying the at least one processor so that the output of the processor corresponds to the desired output.