Distributed Acoustic Sensing Full Waveform Inversion With Direct Strain Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed acoustic sensing data transformed to velocity data for full waveform inversion experiences significant noise degradation, leading to a reduced signal-to-noise ratio and inaccurate subsurface velocity models.
Innovation Solution
Directly utilize strain and strain-rate data within the full waveform inversion algorithm, employing reciprocal modeling to inject source functions that capture the directionality of sensor data, avoiding conversion to velocity data and maintaining a higher signal-to-noise ratio.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If strain data is transformed to velocity data for full waveform inversion, then compatibility with conventional geophone data workflows is improved, but signal-to-noise ratio deteriorates due to noise amplification during spatial deconvolution
Solution Approach 1:
Instead of transforming DAS strain data to velocity data (conventional approach), the patent inverts the approach by directly using strain data in the full waveform inversion algorithm. The sensitivity kernel is modified to compute derivatives with respect to strain rather than velocity, eliminating the need for spatial deconvolution that amplifies noise.
Solution Approach 2:
The patent changes the fundamental parameter being inverted from velocity to strain. This parameter change allows the inversion algorithm to work directly with DAS measurements without requiring transformation, thereby preserving signal-to-noise ratio while maintaining workflow compatibility through modified sensitivity calculations.
2Ease of operation
If spatial deconvolution is applied to transform strain data to velocity data, then conventional inversion workflows can be used, but noise is significantly amplified particularly for low-spatial frequencies
Solution Approach 1:
The patent extracts and removes the problematic spatial deconvolution step from the workflow. By formulating the inversion to work directly with strain data, the transformation step that causes noise amplification is eliminated entirely, while the essential functionality of full waveform inversion is preserved through modified sensitivity kernels.
3Measurement precision
If strain data is used directly in full waveform inversion without transformation, then signal-to-noise ratio is maintained, but compatibility with conventional geophone-based workflows is reduced
Solution Approach 1:
The patent creates a universal inversion framework that can handle both conventional velocity data and DAS strain data through a unified sensitivity kernel formulation. The algorithm is designed to be multi-functional, accepting different data types while maintaining optimal performance for each, thereby achieving both precision and compatibility.
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
Maintains a higher signal-to-noise ratio, resulting in more accurate subsurface velocity models through direct inversion of strain and strain-rate data, enhancing the precision of seismic imaging and earth-model building.
Implementation Method 1
Distributed acoustic sensing systems use fiber (e.g., optical fiber) that is sensitive to the strain or strain-rate parallel to the fiber. The sensitivity is related to the spatial derivative of the displacement (for the case of strain) or velocity (for the case of strain-rate).
Data Source
Figure 1
Figure 2
Figure 3(a)~3(d)
AI summary
A method implements the use of distributed acoustic sensing data with full waveform inversion. The method involves selecting a set of discrete locations along a fiber to act as a set of reciprocal source points, where the fiber provides sensor data sensitive to one or more of strain and strain-rate. The method further involves sorting the sensor data into a set of reciprocal source point gathers to generate sorted data. The method further involves modelling synthetic receiver data via reciprocity by injecting synthetic source data into a seismic model using the set of reciprocal source points. The method further involves updating the seismic model to reduce error between the sorted data and the synthetic receiver data.