Noise-Robust Time-Domain Multi-Scale Full Waveform Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Full waveform inversion (FWI) methods often fail to accurately capture key features of the subsurface due to noise contamination and limitations in handling low-frequency information.
Innovation Solution
The proposed method involves a noise-robust time-domain multi-scale full waveform inversion using convolved data, where a seismic velocity model is iteratively updated by convolving the observed seismic dataset with a wavelet of increasing frequency, thereby enhancing the resolution and accuracy of the subsurface model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional FWI methods are used to process seismic data, then the processing speed is maintained, but the accuracy and quality of subsurface features are degraded due to noise contamination and inability to recover low-frequency information
Solution Approach 1:
The patent applies preliminary low-pass filtering to the seismic data before initiating the full waveform inversion process. This preliminary action removes high-frequency noise components that would otherwise contaminate the inversion results, allowing the FWI to converge to an accurate solution without being affected by noise. The filtered data is then used in the iterative inversion process to recover subsurface properties with high precision.
Solution Approach 2:
The patent implements a multi-scale inversion approach that segments the frequency spectrum into different stages. The inversion process begins with low-frequency components to capture large-scale subsurface structures, then progressively incorporates higher-frequency components in subsequent iterations. This segmentation allows the method to recover both low-frequency information and high-resolution details while maintaining robustness against noise contamination.
2Manufacturing precision
If high-frequency components are emphasized to improve resolution, then detailed subsurface features are enhanced, but noise contamination increases and destabilizes the inversion process
Solution Approach 1:
The patent employs a periodic, multi-stage inversion process where the frequency content is systematically varied across iterations. Each iteration focuses on a specific frequency band, starting with low frequencies for stability and progressively moving to higher frequencies for resolution. This periodic modulation of frequency content allows the inversion to maintain stability while progressively improving resolution without being destabilized by noise.
Solution Approach 2:
Before performing the full waveform inversion with high-frequency components, the patent applies preliminary low-pass filtering to remove harmful high-frequency noise. This preliminary action prepares the data by eliminating noise components that would otherwise destabilize the inversion process when high-frequency details are emphasized, ensuring that the subsequent high-resolution inversion remains stable and reliable.
3Reliability
If low-pass filtering is applied to remove noise, then noise robustness is improved, but high-frequency details and resolution are lost
Solution Approach 1:
The patent segments the inversion process into multiple frequency stages. In the first stage, low-pass filtering is applied to remove noise and establish a stable, noise-robust inversion for large-scale structures. In subsequent stages, higher-frequency components are progressively introduced to recover fine details and improve resolution. This segmentation ensures that noise robustness is achieved in early stages while resolution is enhanced in later stages without compromising either objective.
Solution Approach 2:
The patent maintains continuous useful action by seamlessly transitioning from low-frequency to high-frequency inversion stages. The low-pass filtered data provides a stable foundation that preserves noise robustness, while the progressive addition of higher-frequency components in subsequent iterations continuously improves resolution. This continuous multi-scale process ensures that the benefits of noise filtering are maintained throughout the entire inversion process while progressively recovering high-frequency details.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach significantly improves the quality of FWI results by effectively filtering noise and recovering detailed subsurface features, including those missed by conventional FWI methods.
Implementation Method 1
forming a convolved seismic dataset based on a convolution of the wavelet with the observed seismic dataset
Data Source
AI summary
Systems and methods for noise-robust time-domain multi-scale full waveform inversion using convolved data are disclosed. The methods include obtaining, using a seismic acquisition system, an observed seismic dataset pertaining to a subsurface region of interest; obtaining, using a seismic processor, a seismic velocity model; and iteratively, using the seismic processor, until a stopping criterion is met: selecting a wavelet with a frequency parameter, wherein the frequency parameter increases with each iteration, forming a convolved seismic dataset based on a convolution of the wavelet with the observed seismic dataset, and updating, using a full waveform inversion, the seismic velocity model based, at least in part, on the convolved seismic dataset. The methods further include forming a seismic image of the subsurface region of interest using the updated seismic velocity model.