Seismic Data Multiple Removal via Sparse Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for processing ocean-bottom-cable and deep tow streamer seismic data struggle to effectively remove surface-related multiple reflections without assuming a specific subsurface velocity profile, especially in shallow water environments where additional data is required and wavefield separation is challenging.
Innovation Solution
A modified sparse inversion method that formulates a new model relating data to primary reflections, allowing for the reconstruction of missing near-offset data and removal of all surface-related multiples without wavefield separation, and improves computational efficiency by combining sources and receivers to reduce matrix size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If SRME or Amundsen inversion methods are used to remove surface-related multiples, then all surface-related multiples can be removed without assuming subsurface velocity profile, but additional data or wavefield separation is required which increases device complexity
Solution Approach 1:
The invention extracts and removes only the necessary components (surface-related multiples) from the seismic data without requiring additional data or complex wavefield separation. The sparse inversion method directly estimates primaries from the recorded wavefield, taking out the harmful multiples while preserving the desired primary reflections.
Solution Approach 2:
The method provides a universal solution that works for both streamer and OBC data types without requiring different processing workflows. It handles various water depths and subsurface velocity profiles uniformly, eliminating the need for additional streamer data that SRME would require for OBC data.
2Reliability
If SRME is applied to OBC data, then multiple removal is achieved, but additional streamer data recording is required which increases loss of time and resources
Solution Approach 1:
The method makes the OBC data self-sufficient for multiple removal without requiring additional streamer data acquisition. The sparse inversion algorithm uses only the recorded OBC wavefield to estimate and remove surface-related multiples, making the system self-service and eliminating unnecessary field operations.
3Reliability
If Amundsen inversion is used for OBC data, then all surface-related multiples are removed, but wavefield separation into upgoing and downgoing components is required which increases device complexity
Solution Approach 1:
The invention extracts the necessary information directly from the recorded pressure wavefield without requiring separation into upgoing and downgoing components. The sparse inversion method works directly with the total wavefield, taking out the primary signals and multiples through iterative optimization without complex wavefield decomposition.
4Ease of operation
If conventional methods are used for shallow water OBC data, then processing is simplified, but near-offset receivers are saturated by source signal which loses information
Solution Approach 1:
The sparse inversion method changes the processing parameters and approach to handle saturated near-offset data. By using iterative optimization with sparsity constraints, the method can recover and utilize near-offset information that would be lost in conventional processing, transforming the saturated data into useful primary reflections.
Data Source
AI summary
Method for correcting OBC or deep-towed seismic streamer data for surface-related multiple reflections. The measured pressure data, preferably after conditioning (71), are simulated using a forward model that includes a water propagation operator between source locations and receiver locations and a term representing primary impulse responses (72). Other terms include direct arrivals and source wavelets. Iterative optimization of an objective function is used to minimize the difference between measured and simulated data, updating the primary impulse response term and optionally the source wavelets term each iteration cycle (73). The converged primary impulses (74) are used to construct simulated multiples and direct arrivals (75), which can be subtracted from the measured data. Optionally the measured data might be blended during the forward simulation (72), to save computational costs in the forward simulation (72) and in the inversion (73).