Introduction The magnetic method is widely used in geophysical exploration to map the subsurface structures, identifying the shallow/concealed mineral deposits and basement depth mapping of a sedimentary basin. The most commonly used interpretation methods for the estimation of source parameters (depth, location, and geometry) are Euler Deconvolution (Thompson 1982), enhanced local wavenumber (Thurston and Smith 1997), and Werner Deconvolution (Ku and Sharp 1983). A simple shape body (sphere, cylinder, sheet, and thin dyke) is assumed in these techniques, and a set of linear equations are solved to obtain the source parameters. The main disadvantage of these techniques is that they show erroneous solutions due to a lack of idea about the causative source bodies, noise, and improper choice of window sizes. Global optimization techniques have been used to overcome these problems for the past few decades to estimate the anomalous bodies' source-depth parameters. These methods include genetic algorithm, the differential evolution algorithm method (Balkaya et al., 2017), very fast simulated annealing algorithm (Biswas et al. 2015), ant colony and particle swarm optimization (PSO) methods (Srivastava et al., 2014). Nowadays, machine learning (ML) techniques are also widely used for the interpretation of geophysical data. These ML algorithms are automatically learned and create rules from data without giving a single rule, unlike earlier algorithms mentioned above. In the present study, particle swarm optimization (PSO) method and Artificial Neural Network, Random Forest Regression, and K-Nearest Neighbors Regression machine learning algorithms have been applied for interpretation of magnetic data of two field regions, namely Bankura Anomaly, India and Pima copper deposit, Arizona USA.
Paper
Full text
Inversion of Magnetic Anomaly using Machine Learning Regression Techniques along with PSO
Semantic Scholar · Geology · 2021
Abstract
Introduction The magnetic method is widely used in geophysical exploration to map the subsurface structures, identifying the shallow/concealed mineral deposits and basement depth mapping of a sedimentary basin. The most commonly used interpretation methods for the estimation of source parameters (depth, location, and geometry) are Euler Deconvolution (Thompson 1982), enhanced local wavenumber (Thurston and Smith 1997), and Werner Deconvolution (Ku and Sharp 1983). A simple shape body (sphere, cylinder, sheet, and thin dyke) is assumed in these techniques, and a set of linear equations are solved to obtain the source parameters. The main disadvantage of these techniques is that they show erroneous solutions due to a lack of idea about the causative source bodies, noise, and improper choice of window sizes. Global optimization techniques have been used to overcome these problems for the past few decades to estimate the anomalous bodies' source-depth parameters. These methods include genetic algorithm, the differential evolution algorithm method (Balkaya et al., 2017), very fast simulated annealing algorithm (Biswas et al. 2015), ant colony and particle swarm optimization (PSO) methods (Srivastava et al., 2014). Nowadays, machine learning (ML) techniques are also widely used for the interpretation of geophysical data. These ML algorithms are automatically learned and create rules from data without giving a single rule, unlike earlier algorithms mentioned above. In the present study, particle swarm optimization (PSO) method and Artificial Neural Network, Random Forest Regression, and K-Nearest Neighbors Regression machine learning algorithms have been applied for interpretation of magnetic data of two field regions, namely Bankura Anomaly, India and Pima copper deposit, Arizona USA.