Industrial
Every sealed vessel, every stockpile, every weld hides what matters most. This is the story of the instrument built to read through them anyway.
A vessel that's sealed for a reason. A stockpile too large to sample honestly. A weld you can't reopen just to check on it. Industrial engineering runs on structures and processes that are closed to the eye by design — and for just as long, the instruments meant to read them have been closed off too: point probes that disturb what they measure, radiography that stops at solid material, samples that are already history by the time they reach a lab.
The idea underneath everything on this page is simple to state and hard to build: wrap a ring of electrodes around the outside of something and read how an electric field bends as it passes through. Gas bends the field differently than liquid. A solid bends it differently than a void. Read those bends from enough angles and you can reconstruct, in three dimensions, what's moving or sitting inside — without ever touching it. The first time anyone proved that at scale was in 2007, watching bubbles spiral through a column of oil they were never meant to be seen inside. What follows is what happened when that same trick was pointed at a weld nobody wanted to cut open, a sealed process vessel, and a stockpile of coal.
ECVT Welins — The flaw you can't get to
A crack in a weld, a void in a laminate, a delamination under a pipeline coating — these are the failures that don't announce themselves. They sit inside solid material, waiting, and by the time they show up on the outside it's often too late to do anything but shut down. Every conventional way of finding them earlier comes with a cost: radiography needs two-sided access and a radiation permit, ultrasonics needs a coupled probe walked centimetre by centimetre. ECVT Welins needs neither. It reads the anomaly from a single external face — no drilling, no coating removal, no shutdown.
That single-face read looks straightforward until you consider what the signal actually looks like inside a solid. Fluids move fast enough that a coarse capacitance measurement still catches the physics; dense solids return differences orders of magnitude smaller — small enough that a void the width of a human hair produces a signal indistinguishable from ordinary electronic switching noise. Cancelling that noise couldn't be solved downstream, in software — it had to be solved in the hardware itself. T-configuration CMOS switching strips the parasitic capacitance out at the circuit level, before it ever reaches reconstruction, which is what makes sub-femtofarad resolution possible at all. Even the sensor's shape isn't an afterthought: a 2013 study comparing sensor geometries settled that flat structures like welds and laminates want a planar array, while curved pipes and vessels want one that conforms — a distinction ECVT Welins still respects for every surface it's asked to read. What comes back is a full 3D map: voids as near-zero permittivity, cracks as breaks in continuity, gradients where composition is quietly changing beneath a coating that looks, from the outside, perfectly fine.
ECVT Colins — A sample is already history
By the time a coal sample leaves a bunker, gets bagged, and comes back from the lab with a moisture and ash reading, the coal it describes has usually already been burned. Quality control built on sampling is quality control that's always looking backward — accurate about a shovelful from twenty minutes ago, silent about the eighty tonnes that moved since. ECVT Colins mounts the same exterior sensor architecture — planar or conforming, chosen per the geometry of the bunker wall, chute, or conveyor transfer point it's watching — and reads the coal continuously, in place, with nothing touching the stream.
What it's reading is dielectric contrast: moisture, ash, and carbon each bend an electric field differently, and at bulk-solid scale that contrast is faint enough to disappear into switching noise unless it's cancelled before reconstruction — the same sub-femtofarad, circuit-level cancellation that makes every instrument on this page work in solids at all. The output is a live volumetric map of moisture, density, and composition across the coal mass, streamed straight to plant quality control — not a number from a lab twenty minutes too late, but the read at the moment the coal actually moves.
View Material Inspection →ECVT Industrial — What a sealed vessel won't tell you
Open a fluidised bed reactor or a multiphase separator to look inside it and you've already ended the experiment — the process only behaves like itself while it's sealed and running. For decades that meant operators inferred what was happening in there from pressure drop, temperature, and outlet composition: the equivalent of judging a conversation by the sound through a wall. ECVT Industrial puts a multi-channel array — 8 to 64 channels depending on the vessel — around the outside instead, on vessels from a single inch to sixty across, including bends and T-junctions no straight-tube instrument could read.
The array alone doesn't produce an image; it produces a stream of capacitance measurements, cycled across every electrode pair through the same T-configuration CMOS switching that makes the whole platform's precision possible, at four full volumetric frames a second. Turning that stream into a picture is its own problem — one that took a genuinely different approach to solve. Rather than stacking flat 2D slices into a fake 3D volume, the reconstruction — a feed-forward neural network method called 3D-NN-MOIRT — solves the true non-linear relationship between measured capacitance and volumetric permittivity directly, which is what lets it resolve gas, liquid, and solid phases moving through the vessel in real time, without ever stopping the process to check.
View Process Imaging →