Running a Useful Factory Acceptance Test

Bring your own parts, including the bad ones

A machine demonstrated on ideal samples tells you nothing about production. Bring parts from actual production, including the variant at the edge of tolerance and the one that only runs twice a year.

Bring known bad parts too. A test that only proves the machine accepts good parts has not tested the checking at all.

Test the failure cases deliberately

Remove a part mid cycle. Block a sensor. Open a guard while running. Stop the air. Each of these should produce the behaviour described in the sequence of operation, and you should watch it happen.

The recovery matters as much as the detection: what happens to the part that was in process, and what does the operator have to do to restart?

Measure the rate over time, not over one cycle

A single fast cycle is not a rate. Run the machine continuously for a meaningful period and measure what it produced, including any stoppages.

If changeover is part of your operation, time a real changeover during the test rather than accepting an estimate.

Check the documentation while you are there

Electrical drawings should match what is in the panel. Wire numbers on the drawing should be on the wires.

The sequence of operation should match what the machine actually does. Walk it step by step.

Spare parts lists, maintenance schedules and any calibration certificates should be present rather than promised.

Bring the people who will live with it

A maintenance technician spots things in an hour on the builder's floor that take weeks to discover on your own. So does an experienced operator.

Anything they raise is cheaper to address before the machine is crated than after it is installed.

Write down what is outstanding

A short, specific list of open items agreed by both sides at the end of the test is worth more than a general sense that it went well. Vague acceptance becomes an argument at installation.