Blind Wavelet Extraction and Deconvolution in Seismic Data Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current seismic data processing techniques, particularly blind deconvolution methods like MCMC, face challenges such as time shift and scale ambiguity, resulting in non-unique wavelet and reflectivity pairs, and practical limitations that prevent accurate geophysical modeling, leading to poor wavelet extraction and deconvolution outcomes.
Innovation Solution
The method employs a Metropolis-Hastings sampling procedure to address time shift ambiguity and introduces a time domain filter to constrain wavelet frequencies, along with forming a super trace to conserve energy coherence, ensuring accurate wavelet extraction and deconvolution in the time domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If blind deconvolution using MCMC approach is applied to simultaneously estimate wavelet and reflectivity, then wavelet and reflectivity can be obtained at the same time, but time shift and scale ambiguity problems occur leading to non-unique solutions
Solution Approach 1:
The patent transforms the ambiguous wavelet and reflectivity estimation problem into a constrained optimization problem by changing parameters: introducing sparsity constraints on reflectivity, amplitude constraints on wavelet, and frequency band constraints. This transforms the non-unique MCMC sampling problem into a unique optimization solution with well-defined parameters and constraints.
Solution Approach 2:
The patent implements an iterative optimization algorithm that uses feedback from the convolution operation to continuously adjust wavelet and reflectivity estimates. The feedback loop compares the convolved result with the observed seismic data and refines the estimates until convergence, eliminating the ambiguity present in single-step MCMC approaches.
2Productivity
If MCMC method is used for blind deconvolution, then wavelet estimation is performed, but the extracted wavelet contains frequencies mostly out of the seismic input frequency band
Solution Approach 1:
The patent explicitly constrains the wavelet frequency spectrum to match the seismic input frequency band by introducing a frequency domain constraint in the optimization objective function. This parameter constraint ensures the extracted wavelet contains only frequencies present in the original seismic data, eliminating the frequency band mismatch problem.
3Ease of operation
If trace to trace operation is performed for deconvolution, then individual trace processing is achieved, but seismic events are broken and weakened due to multiple wavelets extracted from multi-channel traces
Solution Approach 1:
The patent merges all seismic traces into a single composite trace before wavelet extraction. This combining approach ensures a unique wavelet is extracted that represents the entire dataset, and this single wavelet is then applied consistently to all traces during deconvolution, preserving seismic event coherence across all channels.
Solution Approach 2:
The patent creates a universal wavelet solution that serves all traces simultaneously. Instead of extracting trace-specific wavelets, a single wavelet is extracted that functions for the entire seismic dataset, ensuring consistent deconvolution results across all channels and maintaining event continuity.
Data Source
Figure 1~5
Figure 3~10C
Figure 6~7
AI summary
Blind wavelet extraction and de-convolution is performed on seismic data to enable its practical usage in seismic processing and to provide quality control of data obtained in areas where data from wells are not available. The wavelet extraction and deconvolution are realized in the time domain by iteration, producing a mixed phase wavelet with minimal prior knowledge of the actual nature of the wavelet. As a result of the processing, the de-convolved seismic reflectivity is obtained simultaneously.