Machine-Learning-Based Convolution Method for Fast Forward Modeling of Induction Log

Abstract

We built a convolution model using machine learning (ML) to calculate induction log responses for one-dimensional (1D) earth models. Compared to analytical forward modeling, the convolution model is extremely fast. ML-based convolution finds accurate induction tool responses from an earth model with layer resistivity and layer boundaries. For a unit induction tool 2C40, the 101-point, 50-ft window convolution model works satisfactorily for a well deviation angle of 60.