Portable Surface Roughness Measurement on Any Part

Walk through most machine shops or MRO hangars and you will still see it. An inspector runs a thumbnail across a blended edge, pauses, and decides whether it feels right. If the call is close, out comes a comparator plate and the surface is judged against a row of reference finishes, by eye and by feel.

It is a skilled judgement. It also produces no number, and no record of how anyone reached it.

Rolls-Royce gave GelSight sensors to its service engineers to replace exactly that check, with one executive saying he could put the device in the hands of a child. The fit is a natural one. GelSight started life at MIT as an attempt to give a camera a sense of touch. Press it onto a surface and it does what the fingernail does, then turns the feel into a 3D map with Ra and Sa attached.

So the better question is how to carry the measurement to the part, instead of finding a lab instrument the part will fit.

How does GelSight measure surface roughness?

Portable surface roughness measurement means taking parameters such as Ra and Sa directly from a component, wherever it happens to be, with a handheld instrument. GelSight does it with a pad of clear elastomer just under 2 cm across, coated on its contact face with a thin reflective skin.

Press the sensor on and that skin settles into every peak and valley beneath it. Six LEDs inside the unit light the skin from different angles while a camera records it, and the software builds a 3D height map. On the GelSight Mobile Series 2 the capture takes 100 ms.

One map gives you Ra, Rz and the other profile parameters, plus Sa and the areal set from ISO 25178-2. Our piece on surface texture vs roughness explains what each family tells you.

Why does it work on any surface?

Because the camera never sees the part. It only sees the inside of the gel.

That removes the problems optical instruments wrestle with every day. Polished titanium throws light straight back at the lens and clear polycarbonate lets it through. Black carbon fibre swallows it, while a painted panel with orange peel scatters it everywhere. Under a GelSight pad, all four look like the same uniform grey skin. Hangar lighting, or Gulf sunshine through an open roller door, changes nothing.

Every point in the map is a measured point, too. On steep flanks and glinting edges, optical methods can return gaps that software fills by interpolation. GelSight has nothing to fill.

What does it take to get a roughness number?

Think about what each method asks of you first.

A portable stylus tester wants a stable seat and a straight run for its skid. You align the traverse across the lay, hold still while a diamond tip drags across 4 mm or more of surface, and trust that the line you picked was typical. On a radius or a blade root, getting the skid to sit square is half the job.

A microscope or optical profiler wants the part on its stage, fixtured so the right area faces the objective, level and in focus. For a turbine blade or the inside face of a flange, the fixture can take longer than the measurement.

Replicas are the usual answer for bores and tight corners. Degrease, apply the compound, let it harden, then wait at least another 30 minutes before easing it out without cracking it. Mount it flat and run a stylus over the copy. MTU Aero Engines used to send impressions like these to its lab while engines in final assembly sat on hold. Its assembly teams have used GelSight since 2019.

With GelSight you put the sensor on the area and press. The gel wraps onto curves, edges and radii, and the capture is done before a stylus would be seated. The GelSight Modulus takes a 90-degree lens for bores, and where no sensor can physically reach, the GelSight Max measures from a replica.

Nothing is dragged across the surface. A soft pad goes on and comes off, so finished, coated and soft parts leave exactly as they arrived.

How does GelSight compare with the alternatives?

Optical profilers have their own territory. For sub-micron texture, film thickness or form on precision surfaces, a Sensofar 3D optical profiler is the instrument, and our comparison of confocal, interferometry and focus variation shows which technique suits which surface. For fast Ra and Sa wherever the part sits, use GelSight.

Can one bad reading fail the part?

Under the new roughness standard, yes. By default.

ISO 21920 has replaced ISO 4287, ISO 4288 and ISO 1302. Most of the change is housekeeping: Rz is now called maximum height, sampling lengths have become section lengths, and Ra is calculated over the full evaluation length instead of averaged across five pieces.

One change has teeth. Unless the drawing says otherwise, acceptance now follows the maximum rule, so every measured value must sit inside tolerance. The old 16% rule, which let a minority of readings stray, now has to be called up explicitly.

Consider what that does to a single stylus trace. You have measured one line, in one direction. If a patch of chatter sits 2 mm to the side, the part passes or fails on where you happened to put the skid.

An area scan takes the luck out of it. GelSight shows the whole patch under the pad, so you find the worst of the surface and run the Ra profile straight through it, along the lay or across it, from the same capture. Guidance on adopting ISO 21920 already points engineers towards 3D methods under ISO 25178 for surfaces that are awkward to profile.

Why can Ra alone hide the problem?

Ra is an average height. It says nothing about shape.

A ground face, a reamed bore and two milled surfaces can all read Ra 0.4 µm and behave nothing alike in service. One is all sharp peaks that chew through a seal. Another is full of deep valleys that hold lubricant well, or hold processing fluid long enough to flash-rust. Steel can sit comfortably inside its Ra specification and still corrode.

Areal parameters see what Ra misses. Ssk tells you whether a surface is dominated by peaks or valleys, Sv gives the deepest valley, and Sdr measures how much real surface a coating has to grip. These used to be laboratory numbers. With GelSight they come from the same press as the Ra value, so a sealing face, a pre-paint surface or an AM part is judged on how it will behave, not only on how rough it averages.

Where is portable roughness measurement used?

Picture one inspector's shift: a blend repair on a compressor blade, the flange face of a skid-mounted pump, an AM bracket fresh off the build plate, a painted access panel and the inside of a bore. That used to mean four or five different methods. With a handheld sensor it is the same press, five times.

• Aviation MRO: blend repairs, landing gear and airframe skins, on the wing or in the shop

• Aerospace and defence: first-off checks on structural parts too big for any lab stage

• Additive manufacturing: as-built and post-processed surfaces, where roughness shifts with build orientation

• Oil and gas: sealing and flange faces on installed equipment

• Tool and die: mould surfaces checked in the press

• Coatings and composites: painted, anodised and composite surfaces a stylus could mark

Which GelSight model fits your work?

All three run the same GelSight Mobile software with the full profile and areal parameter sets.

• GelSight Mobile Series 2: the widest field of view, 17.0 × 14.2 mm, in a 400 g unit. The everyday choice for general roughness work.

• GelSight Max: the highest-resolution handheld system, measuring roughness from 0.2 to 20 µm with traceable accuracy to 0.3 µm.

• GelSight Modulus: swappable lenses built for access, with HR lenses covering 0.2 to 10 µm, including the 90-degree L2090HR for bores.

Our GelSight range page has the detail on each.

What should you prepare before you call us?

• The roughness callout from the drawing, including any cut-off or acceptance rule

• Photographs of the area and the space around it

• How many measurements you need per part or per shift

• How results are reported today

• Your current method, and roughly how long it takes

How AIET helps

AIET Group supplies and supports GelSight Mobile Series 2, GelSight Max and GelSight Modulus for customers in the UAE, Saudi Arabia, across the GCC, Nigeria and Kenya. We start with your drawing and your part. After an application review, we measure your own components, on site or from samples, so you see Ra and Sa on your surface before you commit. AIET Group then handles supply and operator training, and stays on hand locally once the system is in daily use.

Speak to AIET Group

Send us the roughness callout from your drawing, a few photographs of the area, the material, and how you measure it today. Add a sample or an offcut with the same finish if you can spare one. From that, AIET Group will recommend the right GelSight model and show you live results on your own surface. Contact the AIET Group sales team for a site visit and to test your parts live.

Frequently asked questions

How do you measure surface roughness without moving the part?

You take a portable instrument to the part. A handheld GelSight sensor is pressed onto the surface, captures a 3D height map in a fraction of a second and reports Ra and Sa on the spot. Castings, installed pipework and aircraft structures are measured where they sit, with no fixturing or replica.

Can a fingernail test replace a roughness measurement?

No. A fingernail test gives an opinion, and a drawing tolerance needs a value. It cannot produce Ra or Sa, it varies between inspectors, and it leaves nothing for an audit. GelSight was developed to digitise touch, so it keeps the speed of the fingernail check and adds a measured, documented result.

What is the difference between Ra and Sa?

Ra is the average roughness height along a single line, while Sa is the average roughness height across an area. Ra is defined in ISO 21920-2 and Sa in ISO 25178-2. GelSight measures an area, so it reports Sa directly and extracts Ra in any direction from the same capture.

Can GelSight measure roughness on shiny or transparent surfaces?

Yes, GelSight measures polished, transparent, black and painted surfaces exactly as it measures matt ones. The camera images the inside of a gel pad that has taken the shape of the surface, never the part itself, so reflectivity and ambient light have no effect.

Does GelSight scratch or mark the part?

No, GelSight is non-destructive. A soft elastomer pad is pressed onto the surface and lifted away, so nothing is dragged across it. It is safe on finished, polished, painted and coated surfaces, and on soft materials a diamond stylus could mark.

Is GelSight better than a stylus roughness tester?

For shop-floor and field work, GelSight gives you more data for less effort. A stylus records one line in one direction and needs a stable seat and careful alignment. GelSight captures a full 3D area in one press, gives Ra in any direction plus Sa, and works on curved surfaces and in bores.

What does ISO 21920 change for roughness inspection?

ISO 21920 makes the maximum rule the default, so one out-of-tolerance value fails the part unless the drawing calls up the 16% rule. It replaces ISO 4287, ISO 4288 and ISO 1302 and renames Rz as maximum height. Measuring an area makes sure the worst region is found.

What should we prepare before a GelSight demonstration?

Prepare the roughness callout from your drawing, the material and finishing process, and photographs of the area. A sample part or offcut with the same finish lets AIET Group show you live Ra and Sa results on your own surface during the demonstration.