A cracked housing comes back from a customer, a batch fails a leak test nobody can explain, or a first article does not match the drawing. The part goes out for a CT scan, the answer arrives days later, and by then the tool has run more shifts and the parts have shipped.
That delay, not the imaging, is what has kept CT at the edge of manufacturing quality work. The technique has always shown porosity, voids and internal wall thickness without cutting anything open. What it rarely does is deliver that while the process can still be corrected. Lumafield's industrial CT systems are built around that gap, pairing conveyor-integrated inspection of parts coming off the line with cloud-based analysis that reaches whoever acts on the result.
Which defect are you trying to catch?
Before comparing systems, name the failure. Porosity in a casting. Voids in a moulding. Solder joints on an assembled board. A seal seated wrong inside a sealed enclosure. All of them share one property: invisible from the outside and out of reach of a gauge, which is why the usual alternative is cutting a sample open.
Two questions follow, and they decide more than any specification comparison: how often does the defect appear, and how fast do you need to know? One part per thousand on a continuous line argues for inspecting everything; a defect that follows a tooling change argues for inspecting well at the right moment.
How fast does the answer have to arrive?
Sampling is inexpensive and adequate when a process is stable. It gets uncomfortable when a single escaped part is expensive to recall or dangerous to ship, because it only describes the parts you sampled. Inline inspection changes that arithmetic: a conveyor-integrated system scans hundreds of parts coming off the line, so internal quality is checked at production rate rather than on the few units pulled for destructive testing.
The honest test is what happens after the scan. Inline CT earns its cost when the result can change something inside the same shift: a tool corrected, a batch held, a supplier called. If the data lands in a report nobody opens until the quarter closes, that throughput is capacity you are paying for and not using.
Part size narrows the field faster than resolution does
Resolution dominates most CT conversations, but geometry settles them first: a cabinet can only inspect what fits inside it. Lumafield takes parts up to 1000 mm long and 600 mm in diameter, covering assemblies that will not go into a standard enclosure. The largest cabinet in the RX Solutions EasyTom line handles objects up to 720 mm long, and above that the options thin out.
The opposite mistake is just as common. If every part fits in your hand, a benchtop micro-CT gives finer detail in a fraction of the floor space.
Who actually needs to look at the scan?
Analysis in a browser sounds like a convenience feature and is really an organisational one: the people who need to see inside a part rarely run the scanner. The quality engineer is on the floor, the design engineer at another site, the supplier somewhere else again. When all three open the same scan without booking instrument time, an argument about a defect takes hours, not weeks.
When inline CT is the wrong investment
Two limits are quickly stated and routinely ignored. Elemental questions, an unknown inclusion or an alloy to confirm, belong to micro-XRF. External dimensions across a whole body belong to full-surface 3D scanning, which is faster and cheaper at that job. No CT system covers either.
Within CT, two boundaries deserve respect. A system built for throughput and for engineering teams to run is not the one for the finest achievable detail on a small, dense component; that case belongs to a configurable laboratory instrument, where microfocus and nanofocus sources sit on one platform. The part that cannot travel is the other, a structure already welded into place, and no cabinet solves it at any size. That work goes to X-ray inspection carried on a robot arm, which needs no enclosure.
One limit applies throughout: CT shows that a part is bad and where, not why the process produced it that way. That still takes an engineer who knows the tool, the material and the cycle.
What has to be agreed before the scanner arrives
Shielding, floor loading, power, where the conveyor ties into the line and what happens to a rejected part shape the project more than most specification debates. So does the accept and reject criterion: an internal defect needs an agreed threshold, and writing it is usually harder than acquiring the image.
One question a laboratory buyer never faces deserves an early answer. Once inspection sits in the line, the line depends on it, so agree what production does when the system is down for service or a scan is inconclusive. That rule is easier to write calmly than mid-shift with parts stacking up.
The trial worth asking for is not a clean picture of a good part. Send a run containing the defect you care about, then time how long the answer takes to reach the person who could have stopped the line. Merkel Technologies represents Lumafield in Israel and arranges demonstrations on customer parts, installation, training and local service.
The scan is only worth what the line does with it
Inline industrial CT is not a faster laboratory CT but a different proposition: less about maximum detail, more about seeing enough on every part, early enough to matter. That trade suits a manufacturer with a known internal defect, real volume and someone empowered to stop the line. It is poor value for a laboratory scanning a dozen chosen samples a month.
The useful question is not which system is more capable, but which decision you are trying to make faster, and whether anyone downstream is ready to act.