A FAT and SAT protocol for busbar machinery you can actually use

Buying capital equipment from a supplier on the other side of the world comes down to one question: what happens if it does not do what the quotation says? That is decided before the deposit is paid, by whether the acceptance criteria are written down, measurable and attached to the purchase order.

What follows is a protocol you can lift, edit and attach as a schedule. It covers punching, shearing and bending machinery for copper busbar, and it is deliberately specific about sample sizes and limits, because a criterion without a sample size is not a criterion.

Before the FAT: agree the conditions

Nine tenths of acceptance disputes are about conditions, not results. Settle these in writing before the machine is built.

The test piece. Section, material, temper and supplier. Pick a section near the machine's rated capacity, not a flattering one: a machine that punches 8 mm bar beautifully tells you nothing about 16 mm. The bending force calculator will tell you whether the section you chose actually loads the machine. And supply the bar yourself. If the supplier provides it, they will provide good bar.

The programme. One of your own production parts, chosen to include your worst hole pitch, your longest bar, your tightest bend and a tool change.

The measuring equipment. Named, calibrated, certificates available on the day, agreed by both parties. Whose instrument wins in a dispute has to be settled before the dispute.

The environment. Temperature, and the supply voltage the machine will see during test.

Measurement capability, which is where most protocols fail

Two problems will invalidate your results if you ignore them.

Resolution. The working rule is that instrument resolution should be no worse than a tenth of the tolerance you are policing. For a ±0.1° bend tolerance the total band is 0.2°, so you want 0.02° of resolution. Most shop-floor digital protractors resolve 0.05°, which is a quarter of the band. Either bring better instrumentation, or write the acceptance limit against what you can actually measure and say so in the contract. Do not write a limit you cannot verify.

Temperature. Copper expands about 17 µm per metre per kelvin. Over a 6 m bar, one kelvin is 0.10 mm and five kelvin is half a millimetre, which is comparable to the whole length tolerance. Reference temperature is 20 °C (ISO 1). Soak the bar and the measuring instrument together for at least four hours, record the temperature on the test sheet, and correct if you are outside 20 ± 2 °C. Measuring 6 m of copper to ±0.1 mm is harder than making it, and any protocol that does not acknowledge that is decorative.

The FAT tests

1. Hole position over the full bar length

Test piece: 6,000 mm bar of the agreed section. Programme: holes on a stated pitch from the datum end, e.g. Ø11 mm at 500 mm intervals, twelve holes. Method: measure datum face to each hole centre on a CMM, or with an optically centred travelling microscope against a calibrated scale. Bar soaked, temperature recorded. Sample: three bars, all holes on each. Acceptance:

• Cumulative position error within the contracted figure over full length. A punch-shear line specified at ±0.20 mm/m gives ±1.2 mm at 6 m; a processing centre specified at ±0.10 mm/m over 6 m gives ±0.6 mm.
• Adjacent hole-to-hole pitch within ±0.10 mm.
• Hole diameter checked with go/no-go plug gauges at nominal +0.1 / −0.0 mm.
• Burr height on the exit face not exceeding an agreed figure, commonly 0.1 mm on copper, measured at three holes per bar.

Record every value, not a pass/fail. The distribution tells you more than the extremes, and you will want the baseline at SAT.

2. Bend angle repeatability

Test piece: 30 consecutive parts, identical programme, no operator intervention, no manual correction between parts. Method: warm the machine with ten parts first and discard them. Measure each part across a stated span, not at the leg tip, with a calibrated angle instrument. Acceptance:

• All 30 parts within the contracted band, e.g. ±0.1° for machines specified at that figure.
• Sample standard deviation s ≤ 0.025° for a ±0.1° band. That is the value that gives Cp ≥ 1.33 on a 0.2° band, and it is the number that tells you whether the process will still be inside tolerance next Tuesday.
• Range recorded separately.

Then run one extra test that matters more in production than the other thirty: after the machine has stood idle for 30 minutes, bend one part and measure it. The first part after a break is the one that scraps material, and hydraulic machines behave differently cold. Springback compensation gets demonstrated here rather than claimed on a datasheet.

3. Shear length tolerance

Sample: 30 consecutive cuts at a nominal length in your normal range, plus five at the shortest programmed length and five at the longest. Acceptance:

• Length within the contracted mm/m over the measured length.
• Cut face squareness within an agreed figure, e.g. 0.5° measured with a square and feeler.
• Rollover and draw-in depth at the cut edge stated and measured, since it affects joint contact area.
• The five short parts also verify the minimum piece the feed can hold. A machine published with a 55 mm minimum oddment should produce a 55 mm piece, and you should watch what happens to the stub below that.

Measure the 30 nominal parts by stacking them against a common reference. The spread matters more than the absolute, and a stack shows it immediately.

4. Cycle time at stated conditions

Cycle time is the criterion most often written unenforceably. "200 hits per minute" is a punch rate at minimum pitch with no repositioning. It is not throughput, and nobody has ever shipped a hit. Write the condition set into the contract: section, part programme, material, manual or automatic loading, and whether the clock includes bar load and part unload.

Method: run a defined 30-part batch. Wall clock from first bar clamped to last part in the bin, divided by 30. Acceptance: not more than the contracted seconds per part plus 5%.

Record hits per minute separately if you want it, and label it in the report as a tool rate rather than a throughput.

5. Noise

Method: emission sound pressure at the operator position per ISO 11202, during continuous punching of the agreed section at the agreed rate. If the workstation level exceeds 80 dB(A), take sound power per ISO 3744. Acceptance: a stated number, for example not exceeding 85 dB(A) at the operator position while punching the contracted section.

Machinery Directive 2006/42/EC Annex I 1.7.4.2 requires the manufacturer to declare A-weighted emission sound pressure at workstations where it exceeds 70 dB(A), the peak C-weighted value above 130 dB, and A-weighted sound power where the workstation level exceeds 80 dB(A). Directive 2003/10/EC then sets your obligations as an employer: lower exposure action value 80 dB(A) LEX,8h, upper 85 dB(A), exposure limit 87 dB(A). The supplier's declaration does not discharge your duty, and a figure measured on 6 mm bar will not represent your shop punching 16 mm.

6. Safety circuit function

Ask for the paperwork first: the risk assessment, the list of safety functions with the required performance level per ISO 13849-1 (or SIL per IEC 62061), and the validation record per ISO 13849-2. ISO 13850 sets PL r c or SIL 1 as the floor for the emergency stop function itself, with higher levels where the risk assessment demands them.

Then test rather than inspect. Every safety function, every device, in every operating mode:

• Emergency stop from each device, in each mode. The stop must override all other functions, and releasing the device must not restart the machine.
• Guard interlock: open each guard mid-cycle and confirm the stop. Then defeat one channel of a dual-channel interlock and confirm the machine faults rather than continuing.
• Light curtain or two-hand control: verify function, and verify any muting behaviour does what the documentation says.
• Restart after power restoration: the machine must not restart by itself.

Record the measured stopping time and stopping distance for each guarded hazard. That measurement is what fixes the minimum safeguard distance under ISO 13855, and a calculated value from the supplier is not a substitute for a measured one on the machine you are buying.

For machines destined for the EU, note that Regulation (EU) 2023/1230 applies from 20 January 2027 and replaces Directive 2006/42/EC. Ask which the technical file is built to, and get the answer in writing if delivery straddles the date.

7. Alarm handling and diagnostics

Provoke faults deliberately. At least these: pull a bar mid-cycle, load an undersize bar, jam the scrap chute, drop air pressure below threshold, open a guard during motion, obstruct an axis to trip a servo, interrupt the program transfer, and power down mid-cycle then restart.

Acceptance:

• Every alarm has a code, a plain-language cause and a documented recovery action in the manual.
• The HMI language set includes the language your operators actually read.
• Message text names the sensor or IO address. "System error 07" is not a diagnostic.
• For each fault, record whether the operator can resume without scrapping the bar, and how much material each recovery costs.

That last figure is worth more than most of the accuracy data, because it is the one you will pay every week.

Writing criteria the contract can enforce

Each acceptance criterion needs eight elements. Missing any one of them is how criteria become opinions.

Beyond that:

Attach the protocol to the purchase order before the deposit. Not after. A protocol produced during commissioning is a negotiation, not a specification.

Delete unmeasurable language. "High quality finish", "industry-standard accuracy", "no excessive noise" and "commercially acceptable" cannot be enforced. Replace each with a number and a method, or delete the clause and stop pretending.

Define failure and retest up front. One retest permitted after correction, on a fresh sample of the same size, within a stated cure period. Say what happens if the retest fails.

Tie payment to sign-off. A structure such as 30% deposit, 60% on FAT sign-off, 10% on SAT sign-off works because the final tranche is large enough to matter and small enough that the supplier still ships. Retention that is too small buys nothing.

Name standard editions. ISO 13849-1

State that the warranty period begins at SAT sign-off, not at shipment. Otherwise three months of sea freight and customs come out of your cover.

SAT: what changes on site

The SAT is not a repeat of the FAT with a different letterhead. It tests the things the supplier's factory could not.

Power quality. Measure at the machine's own terminals under load, not at the distribution board. IEC 60204-1 expects a steady-state supply voltage of 0.9 to 1.1 of nominal and frequency of 0.99 to 1.01 of nominal continuously (0.98 to 1.02 short-term). Check phase rotation, PE continuity and bonding, and confirm the available fault current at the machine against its rating. Record the voltage dip when the largest motor starts; a hydraulic pump starting on a weak supply is a classic commissioning failure that nobody predicted because nobody measured.

Foundation and levelling. Level the bed to the manufacturer's figure with a precision level and record the readings. A 3,600 kg machine over roughly 33 m² is not heavy per square metre, but it is long, and a long machine follows the floor. Re-check after 30 days and again at three months; concrete moves, and the length accuracy you accepted moves with it.

Ambient conditions. Record the shop's summer and winter extremes and check the hydraulic oil grade against them. A machine commissioned at 30 °C behaves differently at 8 °C in January, and the first-part-after-idle result from the FAT is the one that will drift. In humid or coastal sites, confirm cabinet IP rating and whether anti-condensation heaters are fitted and wired.

Repeat the accuracy tests. Hole position, bend repeatability and shear length, on your material, at your temperature, run by your operators. The delta between FAT and SAT is the transport and installation contribution, and it is the number to argue about while the commissioning engineer is still in the building.

Operator competency sign-off. Named operators, each producing a test part, each signed off on: normal operation, tool change, program load and transfer, alarm recovery, daily maintenance, emergency stop and isolation for maintenance. Keep the record; it is also your evidence of training if anyone is injured later.

Spares handover. An agreed critical-spares list with part numbers, quantities and stated lead times. Wear parts, punch and die sets, shear blades, seals and filters, with a stated life in strokes or hours. And a named commercial source for every item that is not OEM-specific, because bearings, valves and seals should be buyable locally at 3 am.

Documentation checklist. Sign this off item by item, at handover, before the final payment:

• EU Declaration of Conformity (or Declaration of Incorporation), CE marking, technical file reference
• As-built electrical schematics with IO list; hydraulic and pneumatic schematics
• PLC and CNC program backup, parameter backup, and the licence terms for both
• Safety function validation record and the risk assessment
• Calibration certificates for every instrument used to demonstrate acceptance
• Operating and maintenance manuals in the operators' language
• Spare parts catalogue and lubrication schedule
• Machine serial number and as-delivered configuration record

Why the accuracy numbers matter downstream

There is a reason to be strict about hole position and bend repeatability beyond pride: your own product verification depends on them. Under IEC 61439 the assembly manufacturer verifies by test, by calculation, or by comparison with a verified reference design, and temperature rise and short-circuit performance both depend on busbar geometry and joint quality. The standard permits 105 K on bare copper busbars and 70 K at terminals for connection of external conductors, verified against a specific bar, hole pattern and joint.

A machine that drifts on hole position changes joint contact area. A machine that drifts on bend angle changes clearances. Acceptance limits on a punching or bending machine should trace back to the tolerances your verification assumed, not to what looks impressive on a datasheet.

Machines whose published figures you will be testing against, for reference: the EMAC-BP-60 at ±0.20 mm/m hole pitch, the EMAC-BB-H12 at ±0.1° bend accuracy, and the IMAC-CENTER 80 at ±0.05 mm positioning and ±0.10 mm/m length accuracy over 6 m. Those are the numbers to write into the schedule, with the method and the sample size beside them.