Bayesian Filtering of Two Phase Annular Flow with Solubility Mediated Mass Transfer and Vision Based Regime Identification

• Juan Camilo

  Universidad del Magdalena, Carrera 32 No. 22-08, Santa Marta, Colombia

  Author

• Andres Felipe

  Universidad de Nariño, Calle 18 No. 50-02 Ciudad Universitaria Torobajo, Pasto, Colombia

  Author

Gas influx into a drilling annulus can evolve rapidly as it migrates upward through changing pressure, temperature, and hydrodynamic conditions. The resulting two-phase flow regimes are not only a consequence of the influx rate and gas properties but also of transient frictional losses, pipe eccentricity, and phase change driven by solubility. Reliable monitoring therefore requires models that remain predictive across regime transitions while remaining consistent with limited measurements available in operational settings. This paper develops a regime-aware digital twin that couples a reduced one-dimensional two-fluid formulation, a solubility-mediated mass-transfer model, and a probabilistic computer-vision regime identifier, enabling joint estimation of kick size, gas distribution, and evolving flow pattern from pressure and imaging data. The technical contribution is a unified latent-regime closure that is differentiable, smoothly switches interfacial momentum and friction correlations, and is constrained by conservation laws while remaining compatible with learning-based observation models. The approach explicitly represents dissolution and exsolution through a saturation-tracking state variable and leverages temporal visual features to infer regime probabilities with calibrated uncertainty. The overall inference problem is cast as a constrained Bayesian filtering task, where the physics model provides a predictive prior and computer vision provides a high-bandwidth likelihood over regime and void-fraction proxies. The framework is designed to generalize across static and circulating conditions, to accommodate gas-specific solubility effects, and to support mitigation-oriented control by predicting near-term pressure evolution under candidate actuation trajectories.

• PDF

Copyright (c) 2025 International Journal of Scientific and Applied Computing Research

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.