Seismic Data Deblending via Self-Adapting Radon Interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional seismic data processing methods are inefficient in removing strong and high-density noise, particularly in simultaneous source acquisition, where noise is not fully random and affects signal prediction accuracy, especially in irregularly distributed data with moderate aliasing.
Innovation Solution
The method involves estimating signal-to-noise ratios and using data-domain weights to generate models of the signal and noise, employing Radon transforms with selective alpha-trimmed slant-stacks and self-adapting weights to improve signal extraction and noise attenuation in the presence of strong impulsive noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional frequency-space filtering is used to attenuate random noise, then noise attenuation is achieved for regularly distributed data, but the method becomes inefficient in the presence of high-density strong noise and irregularly distributed data with aliasing
Solution Approach 1:
The patent applies different processing strategies to different regions of the data based on local characteristics. Specifically, it identifies and handles irregularly distributed traces separately from regularly distributed ones, and applies different weighting schemes to different parts of the Radon transform domain based on local signal-to-noise ratios. This local differentiation allows the method to maintain reliability in difficult regions while achieving noise attenuation overall.
Solution Approach 2:
The patent dynamically adjusts parameters based on the data characteristics. It changes the weighting parameters in the Radon transform based on estimated signal-to-noise ratios, and adapts the interpolation parameters according to the local data distribution. This parameter adaptation enables the method to handle both regular and irregular data effectively, overcoming the limitation of conventional fixed-parameter filtering.
2Productivity
If simultaneous source acquisition is used to reduce survey time and cost, then productivity increases, but strong incoherent noise and cross-talk interference are introduced that compromise signal extraction
Solution Approach 1:
The patent extracts the signal from the blended simultaneous source data by transforming to the Radon domain where signal and noise separate. It then selectively removes the noise components while preserving the signal, effectively taking out the desired information from the contaminated mixed data. This extraction approach enables recovery of usable signals from simultaneous source acquisition despite the cross-talk interference.
Solution Approach 2:
The Radon transform serves as an intermediary that facilitates separation of signal and noise in simultaneous source data. By transforming the blended data into the Radon domain, the patent creates an intermediate representation where cross-talk from different sources appears as incoherent noise that can be filtered out, while the true signal maintains its coherence. This intermediary transformation enables the productivity benefit of simultaneous acquisition without the harmful cross-talk effects.
3Measurement precision
If data-domain weights based on signal-to-noise ratios are used to generate signal and noise models, then accurate signal prediction is achieved in the presence of strong noise, but device complexity increases
Solution Approach 1:
The patent implements a self-service approach where the data itself provides the information needed for weighting. The signal-to-noise ratio estimation is performed directly on the input data without requiring external reference data or complex manual calibration. The algorithm automatically adapts the weights based on the local characteristics of the data being processed, making the system self-adjusting and reducing operational complexity despite the sophisticated processing involved.
Data Source
Figure 1A~1B
Figure 2A~2B
Figure 3
AI summary
Methods (700) and devices (600) for seismic data processing estimate (720) signal-to-noise ratios of data in a spatio-temporal block of data, determine (730) data-domain weights associated to the data based on the estimated signal-to-noise ratios, and then generate (740) a model of the signal and/or a model of the noise using the data-domain weights.