Before You Sign: 12 Checks on a New Ultrasonic Welder
Most welders get accepted in the buyer’s plant, on the morning the line is supposed to start. By then every problem is a schedule argument. The twelve checks below are the ones we would run before the machine leaves the maker’s floor, ordered so the cheapest problems surface first.
Why the check belongs at the maker’s plant
A machine that fails on your floor costs a week of production and a negotiation about who pays for it. The same machine failing at the maker’s plant costs a re-machining job and a delay on one order. That difference is the whole argument for sending parts ahead and travelling to the acceptance test.
There is a second reason, and it costs nothing to fix today. A welder is a resonant system, and most of what you will need next year exists only on the first day: the amplitude at the tip, the frequency and impedance of a fresh stack, the look of an unworn face. If nobody writes those numbers down before shipment, every later comparison is guesswork. We wrote about how horns wear out separately; this page is about catching the baseline while the machine is still clean.
The twelve checks
| Check | How to test | What counts as a pass |
|---|---|---|
| 1. Amplitude at the tip | Laser vibrometer, or a displacement gauge against a known reference | Number recorded, with the instrument and the method named |
| 2. Amplitude against the datasheet | Same reading, compared with the figure quoted in the offer | Inside the tolerance the maker states in writing |
| 3. Frequency and impedance | Impedance analyser across the full stack | Readings on file for every horn shipped |
| 4. Weld on your own parts | Trial welds on production material, not the maker’s sample coupons | Passes your acceptance criterion; section shows no voids |
| 5. Pull-off force, first article | Five pieces from the first run, pulled to failure | All five above your limit; the spread reported, not the average |
| 6. Tip condition | Magnification, plus a photograph next to a drawing or a new-part reference | Flat face, no ring-shaped wear, no aluminium pickup |
| 7. Clamping pressure | Gauge at the cylinder, watched through twenty cycles | Setpoint reached and held every cycle, no sag mid-weld |
| 8. Continuous run | Eight hours at production rate, on your parts | No over-current trips, no frequency drift, no rising weld time |
| 9. Alarms and interlocks | Trigger each one on purpose: guard, air loss, over-travel, emergency stop | Each stops the cycle and needs a deliberate reset |
| 10. Changeover | Full horn change, timed by your own operator | Inside the time the maker promised, with no re-tuning steps missing |
| 11. Baseline record | Amplitude, frequency, impedance, tip photo, parameter set, serial numbers | Handed over as a file you can open, not a printout in the tool drawer |
| 12. Accessories and spares | Foot pedal, power cable, air line, a set of hex keys, an adjustable wrench; horn and anvil if ordered | Every item on the packing list is in the crate, and each item works when it is plugged in |
Checks 1 to 3: the acoustic baseline
These three are the ones buyers skip, because they need an instrument most plants do not own. Ask the maker to bring theirs, or to run the measurement in front of you. A number you watched being taken is worth more than a certificate.
Two details decide whether the record is usable later. First, write down where the measurement was taken: at the tip, on the horn, or at the booster, because the three do not read the same. Second, ask for the figure the maker is willing to put in the contract. “Around 45 microns” is not a specification; a datasheet number with a tolerance is.
Checks 4 to 6: the weld and the tool
Trial welds on the maker’s coupons prove the machine works. Trial welds on your material prove it works on your job, which is the only question you are paying to answer. Bring production parts, including the batch you are least sure about: a coil with a slightly different oxide, a terminal from a second supplier, a harness with one more strand.
Five pulls from the first run is enough to catch a machine that is out of adjustment. It is not enough to establish a process. What you want on the report is the spread and the lowest value, because the lowest value is what will stop your line. The quality and selection guide covers how to set the limit itself.
Photograph the tip before the first weld. A photograph taken on day one is the only honest reference you will ever have for wear.
Checks 7 to 9: the machine around the weld
Pressure is where cheap machines show themselves. Watch the gauge through a full cycle, not just at the start: a cylinder that reaches setpoint and then bleeds will weld the first part of the shift and miss the last.
The eight-hour run is the check most likely to be shortened to an hour because the hall is closing. Keep it. Thermal drift, a generator that trips on a long shift, and a stack that changes frequency once it is warm all appear in hour three, not in minute ten.
Trigger the interlocks yourself. A guard switch that has been bypassed at the factory is a finding, not a detail, and it is much easier to argue about before the invoice is paid.
Checks 10 to 12: changeover, records, and what is in the crate
Changeover time is a production number, so let your own operator do it while someone times it. If the horn change needs a tuning step that only the maker’s engineer knows, that belongs in the training, not in a phone call next month.
The baseline record is the single most useful file you will get from the whole purchase. Ask for it as a spreadsheet or a PDF that lists each horn by serial number, and keep a copy where maintenance can find it. Without it, “the weld changed” has no reference point.
The accessory list is short and worth counting out loud: foot pedal, power cable, air line, a set of hex keys, an adjustable wrench. Then plug the pedal in and step on it, because a pedal that arrives with a bent cable is the kind of thing that stops a line on day one and takes a week to replace.
If your contract asks for certificates, check them at the same time: ISO 9001, the CE declaration, and the scope line on each. A certificate covering “trade” or “sales” is not a certificate covering manufacturing, and a quality department that receives one finds out at the worst possible moment.
When a check fails
Accept a fix, not a promise that it will settle in. Three responses are reasonable: the maker adjusts and you re-run the same check, the maker replaces the part and you re-run it, or the check moves to the contract as a condition before final payment.
What is not reasonable is signing the sheet with the failing item left blank. A blank line in an acceptance report reads, later, as “the buyer did not check”.
FAQ
How long should an acceptance test take?
One full day for the checks above, plus the eight-hour run, which can overlap with the trial welds. If the machine needs a second day to reach a stable reading, that is information about the machine, not about the schedule.
Do I need a laser vibrometer?
Not necessarily, but you need some way to read displacement at the tip on day one. A displacement gauge against a known reference works, as does having the maker measure it in front of you and recording the instrument used. What does not work is accepting a number with no method attached.
Can the acceptance test be done at my plant instead?
It can, and it usually is, which is why problems turn into schedule arguments. Doing it at the maker’s plant moves the cost of a failed check to the party that can fix it fastest, and it is the only moment when a second machine is still on the floor for comparison.
What if my parts are not ready in time?
Send the closest thing you have: the same alloy, the same thickness, the same surface condition. A test on a similar part still catches a machine that is out of adjustment; what it cannot tell you is whether your acceptance limit is right, and that has to be settled on production material before the line starts.
If you are writing the acceptance sheet for a machine you have not ordered yet, send the part drawing and the joint you need and we will run the trial welds first. The amplitude window data on 70 mm² Cu–Al splices shows what the measurements above look like when they are written down properly.