Automated litho-fluid and facies classification in well-logs: the Rock Physics perspective

Abstract

While accurate litho-fluid and facies interpretation from wireline log data is critical for applications like joint facies and impedance inversion of seismic data, extracting this information manually is challenging due to the complexity and high dimensionality of the logs. Traditional clustering methods also struggle with litho-fluid type inference due to different depth trends in petrophysical rock properties due to compaction and diagenesis. We introduce a Rock Physics Machine Learning workflow that automates litho-fluid classification and property depth trend modeling to address these challenges. This workflow employs a maximum-likelihood approach, explicitly accounting for depth-related effects via Rock Physics models, to infer litho-fluid types from borehole data. It utilizes a robust Expectation-Maximization algorithm to associate each litho-fluid type with a specific Rock Physics model instance, constrained within physically reasonable bounds. The workflow directly outputs litho-fluid type proportions and type-specific Rock Physics models with associated uncertainties, providing essential prior information for seismic inversion.