Automatic Extraction of Surface Wave Dispersion Curves Using Unsupervised Learning and High‐Resolution Tau‐p Transform

Dispersion curves for surface wave recordings are required input for many surface wave inversion methods to image subsurface shear wave velocity distribution, while the conventional extraction of dispersion curves requires significant amount of human interaction. This step impedes efficiency enhance...

Full description

Saved in:
Bibliographic Details
Published in:Earth and space science (Hoboken, N.J.) N.J.), 2023-12, Vol.10 (12), p.n/a
Main Authors: Yao, Hai, Cao, Weiping, Huang, Xuri, Li, Luo, Wu, Bin
Format: Article
Language:English
Subjects:
Algorithms
automatic picking
Clustering
Decomposition
dispersion curve
Fourier transforms
high resolution dispersion spectrum
Machine learning
Neural networks
Principal components analysis
Signal to noise ratio
surface wave
unsupervised learning
Velocity
Velocity distribution
Wave analysis
Wave velocity
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Dispersion curves for surface wave recordings are required input for many surface wave inversion methods to image subsurface shear wave velocity distribution, while the conventional extraction of dispersion curves requires significant amount of human interaction. This step impedes efficiency enhancement of surface wave analysis and its automation. In this paper, we present an unsupervised learning scheme to achieve efficient automatic picking of dispersion curves of the fundamental mode for surface wave gathers. This scheme is composed of four major steps: computing a frequency velocity(f‐v) spectrum for the surface wave gather using a high‐resolution Tau‐p transform improved by the iteratively shrinkage Thresholding algorithm (ISTA) algorithm, generating clusters points along the dispersion energy in the f‐v spectrum via a weighted Kmeans algorithm, filtering these cluster points by principal component analysis (PCA) and Local Outlier Factor (LOF) algorithms to remove the erroneous clusters, and fitting the remaining clusters to form the dispersion curve. Tests with synthetic and field noisy surface wave recordings demonstrated the effectiveness of this approach and its potential in automatic processing of noisy surface wave data sets. Key Points An automatic scheme is developed to pick surface wave dispersion curves, and training data sets are not needed High resolution dispersion spectrum computation and simultaneous denoising of multiple curves enables reliable picking results A key technique for automatic processing of surface wave data to facilitate real‐time imaging and subsurface monitoring
ISSN:2333-5084
2333-5084
DOI:10.1029/2023EA003198