Seismic Data Spectral Broadening via Non-Stationary Wavelet Deconvolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seismic data processing methods face challenges in achieving high-resolution imaging of subsurface structures due to bandwidth shrinkage caused by non-uniform attenuation, free-surface ghost waves, and seismic source limitations, particularly in accurately compensating for frequency-dependent attenuation and source signature effects, which results in noise amplification and inconsistent spectral enhancement across geological structures.
Innovation Solution
A pre-stack amplitude versus angle (AVA) compliant spectral broadening approach using non-stationary wavelet deconvolution and sparse inversion-based Q compensation, which enhances spectral balance and spatial coherence by modeling seismic data as a function of reflectivity and seismic source wavelet, incorporating structural conformity constraints and dip-dependent corrections to maintain signal integrity and reduce noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deghosting is applied to remove free-surface ghost waves, then ghost-induced bandwidth shrinkage is reversed, but frequency attenuation effects remain causing slanted spectra with peak frequency shifted towards low end
Solution Approach 1:
The patent applies frequency-dependent Q compensation to correct the spectral balance after deghosting. By adjusting the Q factor parameter across different frequency bands, the method compensates for frequency attenuation effects and restores the peak frequency to its correct position, transforming the slanted spectrum back to a balanced spectrum.
2Manufacturing precision
If amplitude Q compensation is applied to compensate for frequency-dependent attenuation, then absorption effects are corrected, but high frequency loss occurs due to stabilization requirements of regularization and difficulty in obtaining accurate Q models
Solution Approach 1:
The patent applies local Q modeling where the Q factor is determined separately for different spatial locations and frequency bands. This allows the compensation to be adapted to local subsurface conditions while preserving high frequency information. The method uses local analysis windows to estimate Q parameters that are specific to each region, avoiding the high frequency loss associated with global regularization approaches.
3Manufacturing precision
If deconvolution is applied to remove residual source signature and non-stationary Q effects, then resolution is enhanced, but noise amplification and structural non-conformity among neighbouring traces occur
Solution Approach 1:
The patent employs an iterative deconvolution process with feedback mechanisms that monitor noise levels and structural consistency across neighbouring traces. The method adjusts the deconvolution parameters based on feedback from previous iterations, suppressing noise amplification while maintaining resolution enhancement. The feedback loop ensures that structural conformity is preserved by comparing adjacent traces and correcting inconsistencies.
4Stability of the object's composition
If conventional L2 regularization is used in deconvolution, then total energy of deconvolved trace is constrained, but noise and structural non-conformity are not effectively reduced
Solution Approach 1:
The patent transitions from L2 regularization to L1 sparse regularization, changing the mathematical parameter that constrains the solution. L1 regularization promotes sparsity in the reflectivity series, which effectively reduces noise and enhances structural conformity among neighbouring traces while maintaining energy constraints. This parameter change allows the method to achieve both stability and noise reduction simultaneously.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Generating spectrally enhanced seismic data (100) expresses seismic data as a convolution of reflectivity and a seismic source wavelet (104). This seismic source wavelet varies over a sampling interval and defining a total amount of energy over the sampling interval. An enhanced seismic source wavelet that is a single-valued energy spike that yields the total amount of energy over the sampling interval is generated (112). In addition, the reflectivity is modified to preserve amplitude variation with angle (108). The reflectivity is convoluted with the enhanced seismic source wavelet (114) and residual energy is added to the convolution to generate the spectrally enhanced seismic data (116).