Scroll

The robot executes.
Vision decides.

Robotic cells where artificial intelligence inspects every part, even inside the channels the eye cannot reach.

Send us your partsOur solutions

Advanced Robotics, Brescia

We design and build robotic cells for precision manufacturing. The mechanical side is ours, backed by sixty years of experience; AI vision decides whether the part is good, and keeps the record.

30systems built since 2020
85–99%accuracy, by defect type
100%of parts inspected, no sampling
8 mmthe narrowest bore we inspect

Solutions

Two standard vision cells, five applications

Quality Robocell

AI surface inspection, 6 s per part

Bore Inspection Robocell

The borescope inside every channel, from 8 mm

Vision-guided cells

Foresight or AI, depending on how much the part varies

Machine tending

Up to 4 hours unattended

Deburring

Cast iron and light alloys, path adapted to the casting

Custom cells

24 bottles per pick, at 7 bar

Palletizing

Mixed SKUs and interleaving, without stopping the line

The starting point

Where manual inspection falls short

Quality depends on who checks.

Two inspectors, two verdicts. The same inspector at the end of a shift, a third one. When your customer asks for objective results and full traceability, manual visual inspection stops being defensible.

And finding people for that station gets harder every year.

You cannot see inside the part.

Internal channels, blind holes, bores a few millimetres wide. Manual borescope inspection is punishing, demands constant concentration and is still only a sample.

The defect that gets through is found by your customer.

What we inspect

Real defects, on real customer parts

Real analyses on customer samples: on the left the part, on the right what the model saw.

Brass

Brass fitting and its anomaly map

Fittings with missing parts. A good part stays blue; where the model finds something it has never seen, the map lights up.Anomaly detection

Aluminium

Flaking on an aluminium casting

Die casting: porosity and flaking, measured pixel by pixel.Segmentation

Cracks

Part with a longitudinal crack, on the right the crack segmented in red

Fine cracks along the part, followed along their whole length.Segmentation

Plating

Plated component with an uncoated area highlighted

Missing plating, even on a spot a few millimetres wide.Segmentation

Porosity

Surface porosity, on the right the pores segmented in red

Surface porosity, found pore by pore.Segmentation

Food

Food products in a tray, on the right the two defects segmented

Products in a tray: the defect is found even among dozens of similar pieces.Segmentation

Steel

Threaded hole with porosity at the bottom

Threaded holes with porosity at the bottom, where light barely reaches.Anomaly detection

Bore Inspection Robocell

Inside the channel

The robot inserts the borescope into every bore and the model marks the defects: ridges in red, grit in blue, lumps in green. Hover over an image to see the original frame.

Large ridge on the wall, segmented by the model
Ridge
large, on the wall
Thin ridge, side view, segmented by the model
Ridge
thin, side view
Ridge, front view of the bore, segmented by the model
Ridge
front view of the bore
Grit residue in the channel, segmented by the model
Grit
residue in the channel
Large protruding lump, segmented by the model
Lump
large, protruding
Small lump on the wall, segmented by the model
Lump
small, on the wall

How we work

From your defective samples to production

No system of ours starts from generic images. It starts from your real non-conforming parts, because those are what the model has to get right.

Your samples

We analyse your real non-conforming parts and agree what has to be detected.

Optics and robot

Cameras, lighting and gripping designed around your part.

Training

Around 50 samples to train the model. Anomaly detection, segmentation or presence checking, often in the same cell.

Validation

The model is validated on an independent dataset before it reaches the line.

In production

Every part stays on record with its result and its image, and adds to the dataset.

About us

Thirty years of robotics, sixty of mechanics

Advanced Robotics was founded in 2020 by Danilo Verzeletti, with thirty years of robotic applications, and the Buffoli brothers, with over sixty years of precision engineering. We are part of the Buffoli Industries group.

We don't sell priority support packages: we build cells that shouldn't need one. In six years we have delivered thirty, almost all to returning customers or by word of mouth.

Robots we integrate: ABB · FANUC · Yaskawa · Estun · KUKA · cobots

Our site in Via Ferrini, Brescia, from above