The Early Study for Nature-Inspired Algorithms in Geophysical Modeling Based on Gravity Method: Prototype Bat Algorithm in 3D Gravity Inversion

Abstract

The physical properties of the subsurface are a crucial consideration in various geoscientific analyses. Geophysical methods offer insights into the distribution of subsurface physical properties, with varying approximations based on these properties. One approach to developing a subsurface model is inverse modeling, a technique that converts data into models. However, new approaches are needed for gravity inversion methods to obtain models with reduced ambiguity. Stochastic inversion techniques are commonly used to avoid local optima that can occur in gravity data modeling due to the underdetermined nature of the model parameters. Nonetheless, a key constraint of this approach is selecting a range of model parameters that accurately represent the geological conditions. To address this issue, the Bat Algorithm is applied as a meta-heuristic method to stochastic inversion techniques. This method emulates the behavior of microbats in randomly searching for prey in the model space. Through simultaneous communication, all microbats aim to locate the best solution within a global search topology, thereby avoiding the entrapment of local optima. This study aims to develop a novel approach to 3D modeling of gravity data through inversion methods. A 3D grid of stacked rectangular prisms serves as the density model, with perturbation achieved through manipulation of the 3D array index as a density grid. Synthetic testing demonstrates a success rate of over 85% for single and multi-layer array models with a density body in the center of the mesh grid. The random distribution of density originates from the bat’s propagation of the Bat Algorithm in finding the best model. The suggested technique displays potential for expansion towards the modeling of gravity data with greater complexity, including its practical application in field data analyses.