The real work has moved elsewhere. Almost everything genuinely new in flow measurement lives in firmware, diagnostics, and the network behind the device — not in the sensing element.
Coriolis meters can finally cope with gas in the line
Entrained gas has always been the weak point of Coriolis metering. When bubbles pass through the tubes, energy that should go into vibration goes into relative motion between liquid and gas instead. Drive power climbs to compensate — and because intrinsically safe installations cap available power, drive gain hits 100% with surprisingly little gas. After that, tube amplitude drops and both mass and density readings go soft.
The fix now shipping is to stop driving the tubes at a single resonance. Multi-frequency and dual-frequency drive schemes excite more than one frequency, letting the transmitter separate density effects from gas effects instead of lumping them together. Several platforms also offer Reynolds number compensation for high-viscosity service, and some straight-tube designs output viscosity directly.
You'll see 0.1% and 100:1 turndown quoted under multiphase conditions, though the figures most often cited trace back the better part of a decade. That's laboratory performance. It's a real advance, but if your application is a wellhead rather than a test loop, get data on your own fluid first.
Ultrasonic is improving, mostly in software
Clamp-on transit-time meters have benefited enormously from cheap signal processing. Better DSP and edge computation have cut noise sensitivity in electrically messy plants, and transducers now handle genuinely awkward temperatures and pressures — opening up installations a clamp-on would have been written off for a decade ago. There's also active research applying machine learning to ultrasonic calibration — good on paper, but not yet demonstrated across different pipe sizes, wall materials, and fluids.
Meters that check their own work
This is the change with the biggest day-to-day impact and the least noise around it.
Modern Coriolis transmitters measure flow tube stiffness and compare it against the factory baseline. If stiffness hasn't shifted, the calibration factor hasn't shifted. Manufacturers have deliberately corroded tubes in testing to characterize that relationship, so it's well understood.
The practical distinction is internal versus external verification. External means opening the transmitter, taking the meter out of service, sending a technician, and periodically recalibrating the verification tool itself — because that tool counts as test equipment. Internal verification skips all of it. For remote installations, and for multiphase meters that are expensive to pull, this shifts maintenance from calendar-driven to condition-driven. That's a budget line, not a feature bullet.
Virtual flow metering has grown up
Inferring flow rates from pressure and temperature isn't new, but it's gotten better. The driver is the cost of the alternative: measuring a single well conventionally means routing it to a shared test separator, waiting for transients to settle, and burning hours of disrupted operation.
Recent approaches blend physics with data — physics-informed neural networks trained against CFD, hybrid mechanistic schemes — rather than throwing pure regression at the problem. Sensible deployments treat it as a cross-check on physical meters, or as redundancy while one is down.
Ethernet-APL: real, and slower than you'd think
Ethernet-APL is a genuine step change at the physical layer. Two-wire, loop-powered, intrinsically safe, built on 10BASE-T1L, running up to 1,000 meters without repeaters and powering as many as 50 devices on a trunk-and-spur topology familiar to anyone who's worked with fieldbus.
Be realistic about timing. Five years after the specifications landed in June 2021, adoption is still mostly lighthouse projects, concentrated in Europe. The blockers aren't technical — they're brownfield economics, hazardous area approvals, workforce skills, and the fact that nobody rips out working instruments before end of life. If you do deploy it, segment the network properly from day one.
Valves that tell you what's wrong
Digital valve controllers have become edge analytics devices, flagging friction, air leaks, and calibration drift right at the valve. Vendor case studies put real numbers on it — one combined-cycle plant reportedly saved $68,000 in a single outage after upgrading controllers, plus around $33,500 a year in maintenance. Treat those as directional rather than proven; independent verification of predictive-maintenance ROI in this space is thin, and claims about predicting failures six months out are marketing until someone shows you field data.
Above the valve, reinforcement learning layered over existing APC is worth watching. Its critique of conventional MPC is fair: linearized models can't capture catalyst aging or fouling. But documented wins so far are narrow. It augments APC; it doesn't replace it.
Where the hard problems are
Hydrogen is a signal-to-noise problem. Gas density is about 0.09 kg/m³ at standard conditions and only reaches roughly 40 kg/m³ at 700 bar, so there's very little mass to work with. Dispensing happens at that pressure with pre-cooling to −40 °C, against embrittlement and permeation risk. Liquid hydrogen is harder still: below −200 °C the elastic moduli of 316 stainless go nonlinear, and a standard Pt100 won't read LH₂ temperature at all.
CO₂ for carbon capture is the bigger gap, and it isn't a device problem — it's a traceability problem. Until recently there was no facility anywhere offering SI-traceable calibration of dense phase CO₂ under realistic transport conditions. That's just starting to change: the UK's national flow measurement institute has built a world-first primary standard facility for liquid and supercritical CO₂, and a 2026 interlaboratory comparison brought five European labs together on gas phase. Capacity is still thin against the number of projects that will need it.
Impurities are the other half of the problem. In one test program, orifice meters running impure CO₂ showed around 1% error in gas phase but over 11% in supercritical conditions — though the authors attributed a large share of that to density measurement uncertainty rather than the meter. Separate gravimetric work on clean liquid CO₂ put Coriolis uncertainty near 0.11%. The two aren't a head-to-head comparison, but the direction of travel is consistent: Coriolis is the safer default. Fiscal accuracy under the EU ETS runs on the order of ±1.5% by mass depending on the applicable tier, so check which one governs your installation before assuming the meter spec covers you.
The short version
Specify for diagnostics, not just accuracy. In-situ verification will change your maintenance costs more than another decimal place ever will. Ethernet-APL is worth planning for and probably not worth retrofitting yet. And if you're heading into hydrogen or CCS work, budget for the uncertainty analysis, not just the meter.





