3D Seismic Interpolation via Machine Learning Matrix Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing interpolation methods for 3D seismic data fail to accurately reflect the three-dimensional characteristics of strata, leading to limitations in the accuracy of data values estimated by interpolation in seismic surveys.
Innovation Solution
A machine learning-based interpolation method is developed, which uses seismic ground truth data to interpolate missing data in 3D seismic surveys. The method involves specifying a partial matrix within the survey area, training artificial intelligence using survey data from odd-numbered rows as training data and even-numbered rows as correct answer data, and then using this AI to interpolate data for even-numbered rows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing interpolation methods are used for 3D seismic data, then the interpolation process can be completed, but the accuracy of data values estimated by interpolation is limited because the three-dimensional characteristics of the strata are not reflected
Solution Approach 1:
The patent divides the survey area into multiple unit areas arranged in a matrix form with rows and columns. The interpolation process is segmented into multiple steps: specifying a partial matrix, training AI using survey data from odd-numbered rows, transforming the partial matrix by alternating rows and columns, and generating interpolated data for even-numbered rows. This segmentation allows the complex 3D interpolation problem to be broken down into manageable computational steps while maintaining accuracy.
Solution Approach 2:
The patent transforms the partial matrix by alternating rows and columns through axis transformation. This dimensional manipulation enables the AI model to process and interpret the three-dimensional characteristics of strata more effectively. By transforming the matrix structure, the method captures spatial relationships in multiple dimensions, allowing the interpolated data to reflect the true 3D geometry of underground strata rather than treating it as a simple 2D surface.
2Quantity of substance
If hydrophones are densely arranged in the direction of travel of the probe, then data coverage in that direction is improved, but the spacing between streamers remains relatively sparse resulting in unsecured data for areas between streamers
Solution Approach 1:
The patent uses machine learning to create a copy or prediction of the missing data in areas between streamers. The AI model is trained on survey data from areas where data is available (odd-numbered rows) and then generates interpolated data for areas where data is missing (even-numbered rows). This copying approach allows the method to reconstruct the missing information without requiring additional physical measurements in the gaps between streamers.
Solution Approach 2:
The patent performs preliminary training of the AI model using survey data from odd-numbered rows before generating interpolated data for even-numbered rows. This preliminary action allows the system to learn the three-dimensional characteristics of the strata from available data and then apply this knowledge to predict missing values. By preparing the model in advance with training data, the method ensures accurate interpolation when processing the missing areas between streamers.
Data Source
AI summary
The present disclosure is an interpolation method of 3D seismic data based on machine learning. The present disclosure has been made to solve the limits and provides an interpolation method for three-dimensional seismic data, capable of performing interpolation for an area where data is not secured based on machine learning using seismic ground truth so that data estimated by interpolation reflects the three-dimensional characteristics of a stratum well.