Accurate detection of the oil–water interface in three-phase separators remains a critical challenge in upstream oil and gas operations, particularly under conditions characterized by unstable emulsion (“rag”) layers. These emulsions introduce complex dielectric gradients and density variations that compromise the performance of conventional measurement systems, leading to operational inefficiencies such as oil carry-over and water carry-under. While advanced sensing technologies - including Guided Wave Radar (GWR) and Radio Frequency (RF) Capacitance - offer improved capabilities over traditional methods, each exhibits limitations under varying process conditions. This study proposes a hybrid measurement framework that integrates GWR and RF capacitance sensing with an edge-based digital signal processing architecture. Central to the framework is a “Digital Bridge” algorithm, which employs confidence-weighted sensor fusion to dynamically adapt measurement outputs based on real-time signal integrity. The system was experimentally evaluated in a controlled three-phase separator environment under varying emulsion thicknesses. Results demonstrate that the proposed hybrid framework significantly outperforms standalone sensing technologies, achieving a mean deviation of ±5.2 mm, reduced variability, improved signal-to-noise ratio, and faster response time. A critical transition threshold (Confidence Score ≈ 0.38) was identified, marking the point at which measurement dominance shifts from radar to capacitance sensing. The system-maintained measurement stability across all tested conditions, effectively mitigating signal attenuation and noise associated with thick emulsions. The findings confirm that the integration of heterogeneous sensing technologies with adaptive digital intelligence provides a robust and scalable solution for multiphase interface detection. This approach offers significant potential for enhancing process reliability, improving hydrocarbon recovery, and supporting regulatory compliance in modern upstream operations.
Beverley Mack (Tue,) studied this question.