A window cleaning robot works through four systems: a vacuum pump that clamps it to glass, edge sensors that stop it at frames, a misting head that softens dirt, and a squeegee-vacuum pair that collects the dirty water. When a robot streaks or stalls, one of those four has failed, not all of them.
The mechanics are simple enough to explain to a facilities team in ten minutes. The failure modes take longer, and they are the part vendors skip.
How does the vacuum actually hold the machine on?
A ring seal sits against the glass and an internal pump pulls the cavity to somewhere around 0.3-0.6 bar below ambient, depending on the model. That pressure differential across the pad area is what generates hold. On a Lingdu Lingkong K3-class unit, the seal ring is roughly 300 mm across, which at those pressures gives tens of kilograms of clamp force.
The number that matters is not the peak hold. It is the leak rate. Any path for air to get back under the seal reduces the differential. Sealant ridges, weld spatter, deep scratches and the sawtooth edge of an old gasket all leak. So does dried cement dust, which is why the manual tells you to wipe the glass edge first.
There is a second design worth knowing. Some robots use a vacuum plus a micro-pump on the wiper to recover water, which is the same architecture a wet vacuum uses. Others just squeegee and let a trailing absorbent strip collect residue. The first leaves less streaking on hot glass, where water evaporates before you finish the pane.
How does the robot know where the glass ends?
Two common methods. Optical sensors look for the contrast between glass and frame or sky, which works well on framed windows and badly on frameless glass roofs where the edge is just an end. Pneumatic sensing watches for a change in vacuum resistance when the seal crosses the edge, which is slower but more reliable on frameless panelling.
On a skylight, this is where most trouble starts. A glass roof usually has mullions, gutters, drainage channels and parapet upstands. The robot needs to bridge some, avoid others, and not treat a dark gutter as the edge of the world. Machines built for this, like the Lingyun Y3 and Lingfeng S1, use obstacle handling as a first-class feature rather than a fallback. Our Lingyun Y3 review documents how that behaves on stepped glazing.
What does the water system do besides make things wet?
- Softening: the mist dwell time is what lifts mineral film and bird residue. Too little water and the pad scrubs dirt into the glass.
- Lubrication: a damp pad slides with less drive-motor load, which extends battery runtime on cordless units.
- Recovery: dirty water must leave the glass, not dry on it. Hot glass dries water in under a minute, which is exactly how a clean pane becomes a streaked one.
The recovery step is where cheap machines get exposed. A 2026-era pad that only wipes will look fine on an overcast autumn morning and terrible on a sunny July afternoon. If your glass roof faces west, test the machine in the afternoon before buying.
Why do robots still miss the corners?
Because the seal ring is round and the pane is square. Corner geometry leaves an arc of 40-80 mm the pad cannot reach under full suction. Manufacturers work around it with a corner-boost mode that briefly reduces suction near an edge, or with an offset programme that lets the pad overhang slightly.
| Accessory | Typical cleaning width | Time per 10 m2 | Best use |
|---|---|---|---|
| Standard pad | 240-280 mm | 14-18 min | Flush glazing, few turns |
| Wide pad | 320-360 mm | 10-13 min | Large unbroken skylight panels |
| Corner kit | N/A | +3-5 min | Mansard edges, triangular panes |
| Hand pole head | 400-500 mm | 8-12 min | Frames and gutters between panels |
Anyone who claims a robot cleans a skylight with no manual corner work has not stood on the roof. Budget five minutes of hand work per 10 m2 and you will not be disappointed.
What breaks first in real use?
Pads, then seals, then pumps. A pad on a commercial site lasts 40-80 hours of running; on a dusty industrial roof, half that. Seal rings hold up better, but they must be inspected for nicks. Pump failures are rare and usually caused by running dry or by hard water scale in the misting line.
The most common surprise is not mechanical at all. It is water logistics. A tank that holds enough for 60 m2 means the crew hauls a refill up a ladder mid-job. Facilities managers who want one machine per building should count refills in the schedule, not just runtimes.
Who should care about the internals?
Homeowners, no. Buy on pad width, recovery method and consumable price. Facility teams, yes, because you need to explain to a procurement committee why machine A holds better than machine B on their specific glazing. Once you know that vacuum hold is a leak-rate problem, that explanation gets much shorter.
If you want the field comparison rather than the theory, the consolidated skylight robot reviews cover what each machine does well on pitched glass, gutters and parapet edges.
Key Takeaways
- Four systems do all the work: vacuum hold, edge sensing, water softening and squeegee recovery.
- Hold is limited by leak rate. Dust and sealant ridges matter more than peak suction spec.
- Frameless glass roofs need pneumatic edge sensing; optical sensors struggle with gutters and sky.
- Water recovery, not water delivery, separates good machines on hot west-facing glass.
- Budget five minutes of hand work per 10 m2 for corners and frames.