AI Digital Rock Typing (AIDRT): A No-Code Physics-Augmented Machine-Learning Workflow for Carbonate Rock Typing and Static-To-Dynamic Property Prediction

Abstract

Abstract Carbonate rock typing requires connecting the pore-throat network (PorThN) and geological lithofacies description with porosity, permeability, capillary pressure, wettability, and relative permeability across multiple scales, from nano-, micro-, and macro-scales with core image and plug scales to log scale in feet, and finally to reservoir-model scales. This paper presents AI Digital Rock Typing (AIDRT), a no-code, physics-augmented workflow that builds that connection while keeping observations, predictions, calibrations, uncertainty, validation evidence, and expert approval distinct. AIDRT organizes image and tabular evidence through Machine Learning Image Processing and Recognition (MLIR), which identifies Controllable Measurable Volume (CMV) with descriptors as part of the Morphology Decoder process, and integrates 3D Carbonate Heterogeneous Permeability (kHeC), Permeability-linked Capillary Pressure Function (P-Function), wettability-sensitive relative-permeability analysis with pore-throat Deformation Factor (df), and the digital Rock Typing Carbonate Morphology chart (DRT-CAMO). The workflow registers sample scale and provenance, derives pore- and mineral-related evidence, evaluates the pore-throat network, estimates local and aggregated permeability, constructs capillary-pressure and flow-function hypotheses, and transfers candidate rock types toward core–log correlation. The AIDRT implementation completed all software tests and demonstrates traceable project artifacts, validation, and uncertainty views. These results verify the tested workflow behavior with the bundled synthetic table and image. Eqs. (1) - (9) remain the dependable AIDRT workflow formulation. AIDRT's principal contribution is therefore a reviewable evidence-to-reservoir-model chain in which petrophysicists, geologists, and reservoir engineers can examine the same rock-type decision, preserve its applicability limits, and retain accountable interconnectivity among porosity, permeability, lithology, lithofacies, capillary pressure, wettability, and relative permeability, to produce carbonate rock types and their reservoir static and dynamic petrophysical properties, all in one place.