Get a Free Quote

Our representative will contact you soon.
Email
Mobile/WhatsApp
Name
Company Name
Message
0/1000

Blog

 >  Blog

News

A Launch Checklist For A Vision Station

Time : 2026-11-18

A vision station is usually judged on its accuracy, and accuracy is one of the last things to go wrong. What goes wrong first is that the station was installed in a place where it cannot see, calibrated against a part that was not representative, and accepted on a sample that did not include the defects it was bought to find.

These are the checks that belong before each of those stages closes.

Before The Vision Station Arrives

These are structural, and none of them can be fixed cheaply after the station is on the floor.

  • Confirm the floor and the structure will not transmit vibration into the vision station — a station on a mezzanine next to a press will move more than its tolerance
  • Confirm the light environment: a vision station near a window or a roller door sees its illumination change with the weather
  • Confirm the part arrives at the vision station in the same orientation every time, and that the station has a defined position to check it from
  • Confirm there is space to service the vision station from behind as well as in front, because half the adjustments are at the rear
  • Confirm who will own the vision station after handover, by name, before it is installed
  • Confirm the vision station's working distance is fixed and repeatable to within 0.5 mm, because scale error follows distance directly
  • Confirm the vision station images a feature of 0.1 mm with enough contrast to threshold reliably, not just enough pixels to resolve it
  • Confirm the vision station's images are stored with the part identity, so a disputed rejection can be reviewed a month later

Before The First Part Is Inspected

These checks concern the vision station's own state rather than the parts.

  • Run a reference target through the vision station and record the result as a baseline, not just a pass
  • Check that the vision station's illumination is stable over an hour of continuous running, not just at switch-on
  • Confirm the vision station's field of view covers the part with margin, so a part presented slightly off-centre is still fully imaged
  • Confirm the station rejects a part it cannot measure, rather than passing it by default

At First Article

This is where a vision station is accepted or quietly found wanting, and the sample matters more than the average.

  • Include known-bad parts in the sample, with defects at and just below the acceptance limit, and confirm the vision station finds them
  • Include parts from a different casting or batch, because a vision station tuned on one batch may be tuned to that batch
  • Confirm the vision station's measured values against a reference instrument on at least ten features, and record the differences
  • Confirm the vision station's reject path works end to end: part flagged, part removed, record created
  • Run the same sample twice and compare, because a vision station that gives two answers has a stability problem rather than an accuracy problem

During The First Week

The first week is when the vision station meets production, and it is the week that determines whether it will still be trusted in a year.

  • Log every rejection with its image, and review the log daily rather than weekly
  • Count how many rejections the operators override, because that number is the vision station's real acceptance
  • Check the vision station's reference target again at the end of the week and compare with the baseline from day one
  • Confirm the variant identification is being read from the part rather than selected by the operator
  • Confirm somebody has already added a new defect class to the vision station, because if nobody has, nobody can

The vision station that passes these checks is not a better instrument than one that does not. It is the same instrument installed where it can see, calibrated against parts that include the bad ones, and accepted on evidence rather than on demonstration.

If You Only Keep Three

Keep the known-bad sample, because it is the only thing that proves the vision station still works after a change. Keep the reference target baseline, because it is the only thing that shows drift before it reaches the customer. And keep the override count, because a vision station whose rejections are routinely overridden is not in service, whatever the report says.

Get a Free Quote

Our representative will contact you soon.
Email
Mobile/WhatsApp
Name
Company Name
Message
0/1000