AI-Driven Slip Model Integration in Embedded Edge Devices for Accurate Multiphase Flow Measurement

Abstract This work presents a machine learning-based slip model designed for real-time integration into multiphase flow meters operating on embedded edge devices. Unlike traditional slip models constrained by fixed correlations, the proposed approach leverages physics-informed features and is trained on diverse flow loop datasets. The resulting model, deployed as a lightweight Java Archive (JAR) and interfaced with C++ via Java Native Interface (JNI), achieves sub-15 millisecond inference latency within limited hardware resources. Field and laboratory trials confirm the predictive accuracy and operational stability of this approach across a wide range of flow regimes, demonstrating its potential to enhance the fidelity and responsiveness of embedded multiphase metering systems.