PhaseVu AI: Turning Liquid Carryover into Actionable Insights
PhaseVu AI is a Machine Learning diagnostic for liquid carryover and mist breakthrough in gas processing. It combines live process data (flow, pressure, temperature, differential pressures and levels) with LineVu visual evidence from inside the gas pipeline to build a digital twin of normal separation behavior and provide root cause analysis.
The system then shows what is driving liquid carryover, flags abnormal deviations early and provides time-stamped visual proof for troubleshooting and reporting.
Detects liquid carryover and mist using visual evidence. Correlates process data with visual data.
Smarter alerts that adapt to real process behaviour.
Identifies variables driving carryover risk to guide proactive maintenance and tuning.
The presence of liquid in gas flows increases uncertainty on flow meters from around 1% to up to 10%.
Provides time-stamped video evidence aligned to historian tags for root cause analysis, disputes, and compliance reporting.
Model retrains on new operating data so thresholds and predictions remain aligned as the plant changes.
The feature importance chart below shows a root cause analysis from PhaseVu AI. In this example, of the 15 monitored process parameters, three have the greatest influence (Liquid level, Flow 1 & Temp 3) on the mist flow observed at the separator outlet. In this case, adjusting the liquid level threshold would improve performance.
PhaseVu AI delivers significant financial advantages by helping to detect and reduce liquid carryover into gas export lines, which can cost operators millions annually. Studies show that eliminating even small amounts of liquid carryover can protect revenues by up to $5 million per year on a 100 MMSCF/Day gas flow. In addition to revenue protection, this tool helps optimize pigging and disposal costs, minimizes unplanned shutdowns, and strengthens compliance with contractual and regulatory gas quality requirements, providing operators with both cost savings and peace of mind.
Demopassword1!
