Depth-to-basement Estimation of Basin Relief Using the BP Network

Abstract

Nonlinear gravity inversion is a popular method for determining basin bottom relief and delineating basin configuration. However, traditional gravity inversion presents certain challenges, including the complexity and time demand of calculating and transforming large matrices, as well as instability and non-uniqueness caused by the inherently ill-posed nature of inversion problems. Over the past decade, deep learning, a subset of machine learning, has seen successful applications in geophysical interpretation and exploration. In this study, we propose an innovative method for estimating two-dimensional (2D) depth-to-basement using a BP network structure. This structure leverages the Leaky Rectified Linear Unit (ReLU) as an activation function, yielding more realistic geophysical models.