Every phase. Every vessel. In real time.
Wrap electrodes around the outside of a vessel and the field between them already carries the answer — it bends differently through gas, liquid, and solid.
Nothing that touches the flow, or looks through it, can tell you what a sealed vessel is doing inside.
Inside a pressurised vessel, gas bubbles rise through slurry. Liquid slugs form and collapse in pipelines. Solids cycle through a fluidised bed. None of it can be measured by putting something into the flow to find out.
A probe pushed into the vessel changes the very flow it is trying to record. Light-based methods stop at the first opaque surface. Radiography cannot deliver a full volume fast enough to keep up with the process.
An array of electrodes placed on the outside wall sets up an electric field that passes straight through whatever is inside. That field bends differently depending on what it passes through — gas, liquid, and solid each hold their own dielectric permittivity (how strongly a material responds to an electric field). Measuring how the field bends across every electrode pair is enough to reveal how the phases are arranged inside.
Turning that raw capacitance data into an actual 3D picture is where 3D-NN-MOIRT (3D Neural Network Multi-criterion Optimization Image Reconstruction Technique — the algorithm that turns field readings into a volume image) does its work: rather than approximating the relationship between permittivity and capacitance as linear — the older, cruder approach — it solves the real non-linear relationship directly, so the resulting image matches what is physically there.
Underneath that reconstruction sits a multi-channel array (8 to 64 channels depending on the vessel), FPGA-based quadrature phase detection, and CMOS T-configuration (a switching layout that cancels out stray, unwanted capacitance picked up by the wiring itself) switching built specifically to keep the measurement clean enough to resolve 0.21–0.42 fF at four frames per second.
How a reading on the outside of the wall becomes a picture of what's moving inside.
Phenomena no one had actually seen happen, only inferred from what came out the other end.
The choking transition in a circulating fluidised bed — the moment a dilute stream of solids suddenly collapses into slugging flow — had never been watched as it happened inside the bed itself. It could only be detected indirectly, from pressure or flow changes elsewhere in the system.
A three-layer 12-channel cylindrical ECVT sensor was set around a 0.1 m ID circulating fluidised bed. Because the sensor reads the whole volume rather than a single point, it captured the actual solids holdup redistributing in real time as the transition occurred — the first time the choking phenomenon had been seen volumetrically rather than inferred.
The same sensor, wrapped around a gas-liquid bubble column running air through Norpar 15, picked up spiral bubble motion tracing through the full 3D volume — a trajectory no 2D imaging method could have followed, since it leaves the plane a flat image would be limited to.
The same sensing approach was then tested across a much wider range of vessels, from 1 inch up to 60 inches, including a 12-inch gas-solid fluidised bed with horizontal gas jets penetrating the flow, confirming that the underlying principle scales with vessel size rather than being limited to one geometry.
The same sensor architecture was later taken out of industrial vessels altogether and applied to soil water infiltration, on the reasoning that if the technique is really just reading permittivity distributions, there is nothing industrial-specific about it — a reasoning the results bore out.
Wherever a wall separates what you need to know from what you can see.
Each of these is a case where the flow could not be reached without disturbing it — so the measurement had to come from outside instead.
Circulating Fluidised Beds
The choking transition happens too fast and too internally for a probe to catch cleanly — the exterior sensor watches it unfold in the full volume instead.
Bubble Columns
Bubbles spiral out of the flat plane any 2D method would be confined to — the sensor follows them because it was never limited to a plane to begin with.
Three-Phase Systems
Because gas, liquid, and solid each carry their own permittivity, one exterior reading separates all three at once — no probe placed in any of them.
Gas-Solid Fluidised Beds
A horizontal gas jet developing inside a 12-inch bed is exactly the kind of internal event a wall-mounted point sensor would miss entirely.
Pipeline & Complex Geometry
Bends and T-junctions distort flow in ways a straight-pipe sensor can't follow — the volumetric reading doesn't care what shape the vessel takes.
Geophysical & Environmental
Soil water infiltration is still just a permittivity distribution moving through a volume — the same sensor reads it the same way, outside industry altogether.