Seismic Trace Compression Using SVD Noise-Signal Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing seismic data processing technologies struggle with significant seismic background energy and sensor noise, leading to obscured wavefront features and high data storage and transmission requirements, which hinder offsite and real-time processing.
Innovation Solution
A method and system for seismic trace processing that separates signal and noise components using singular value decomposition and other machine learning techniques, allowing for data compression and efficient storage/transmission of seismic data while preserving important waveform features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If traditional filtering methods (bandpass or Wiener filtering) are applied to suppress noise, then some noise reduction is achieved, but the seismic background energy remains significant and wavefront features remain obscured
Solution Approach 1:
The patent segments the seismic data into signal components and noise components through singular value decomposition. By decomposing the data matrix into singular values and vectors, the method separates structured wavefront signals from unstructured noise, allowing selective processing of each component to improve both noise reduction and feature preservation
Solution Approach 2:
The patent changes parameters by applying different processing strategies to different singular value components. Small singular values (associated with noise) are suppressed or filtered differently than large singular values (associated with signals), enabling selective noise reduction while preserving wavefront features through parameter-based differentiation
2Object-affected harmful factors
If adaptive filtering methods are used to suppress noise effectively, then noise reduction improves, but the arriving waveforms are altered
Solution Approach 1:
The patent segments the seismic data into signal and noise subspaces through singular value decomposition. This segmentation allows the method to identify and preserve the structured wavefront signal components while removing unstructured noise, avoiding the waveform alteration problem of adaptive filtering by not applying uniform suppression across all components
Solution Approach 2:
The patent extracts the noise component from the seismic data by identifying it through small singular values and removing or suppressing only those components. This extraction approach eliminates noise while preserving the original waveform structure contained in the large singular value components, preventing information loss
3Loss of information
If full-resolution seismic data is stored and transmitted, then complete waveform information is preserved, but data storage and transmission resources are excessively consumed
Solution Approach 1:
The patent extracts only the essential signal components from the seismic data by identifying and retaining only the large singular values that contain wavefront information. This extraction eliminates redundant noise data while preserving critical waveform features, achieving significant data compression without information loss
Solution Approach 2:
The patent changes the data representation by transforming full-resolution seismic traces into a compressed singular value decomposition format. By storing only the significant singular values and their corresponding vectors rather than complete trace data, the method reduces data volume while maintaining waveform information integrity
4Quantity of substance
If data compression is applied to reduce storage and transmission requirements, then resource efficiency improves, but processing complexity increases
Solution Approach 1:
The patent segments the compression process into distinct mathematical steps (singular value decomposition, component identification, selective retention) that can be systematically implemented. This segmentation transforms a complex compression problem into manageable processing stages, reducing overall system complexity while achieving effective data reduction
Data Source
AI summary
In the field of seismic data collection and analysis, the problem of data compression is considered. It is desired to reduce the storage and transmission requirements of seismic data, such as associated with microseismic monitoring and processing, VSP (vertical seismic profile) surveys, and the like, for instance using a distributed acoustic sensor (DAS), which can generate in excess of 50 GB of data per hour on a survey that lasts days. A method and system for data compression that separates the data collected into additive signal and noise components, and compresses the estimated signal component for transmission, storage, and analysis, is described. The idea is that the signal component, which exhibits clear structure across the traces, may be accurately described with relatively few parameters, and therefore may be significantly compressed without loss of important information.