Horns don’t fail all at once. They wear, and the first thing that changes is your weld quality, not the horn. By the time you can see a crack, the line has probably been running out of spec for weeks. Below: what goes, in the order the warning signs appear, and how to measure each one instead of guessing.
What wears out
The horn is the part that touches the workpiece. Some call it a sonotrode. It’s a resonant component, designed to vibrate at one frequency with one amplitude, and every property that matters comes from that geometry. Four things degrade it, and people mix them up.
Fatigue cracking. The horn flexes millions of times, and cracks start where stress concentrates: the base of the amplitude step, slot roots, the thread. This is the one you can’t fix.
Aluminium pickup. Weld aluminium and a thin layer transfers onto the tip. Left alone it builds up, changes the contact condition, and can shift the resonant behaviour. It’s also the most preventable item here.
Face wear and cupping. The tip loses flatness and goes concave. Contact area grows, pressure per unit area drops, and welds get less consistent with the settings untouched.
Loosening at the joint. The horn-to-booster interface is a threaded joint sitting in vibration. Wrong torque, dirt, or repeated assembly lets it fret. You’ll feel that as heat at the joint and amplitude that won’t stay stable.
Five signs, earliest first
Ordered by when they show up rather than by severity. The early ones are warnings. By the fifth, you’re buying a new horn.
One: amplitude drops at the same setting. The generator says the same percentage, but the tip is moving less. Nothing else on the line changed, yet you need more time or more pressure to pass the same weld. It’s the earliest reliable indicator, and the one most plants never measure.
Two: frequency drifts and impedance climbs. A worn or cracked horn stops sitting exactly on the stack’s design frequency. The generator works harder, runs hotter, and may trip on over-current during a long shift. Compare impedance and resonant frequency against the readings from a new horn. Then you’re looking at data rather than a feeling.
Three: pull-off force trends down. Sample testing still passes, but the trend line slopes. If your records say 5,800 N in month one and 5,200 N in month eight with identical parameters, the tool changed, not the operator.
Four: the tip looks different. Bright polished patches, dark oxide, a visible aluminium layer, or a ring-shaped wear pattern instead of full-face contact. Photograph the tip when it’s new. Human memory is useless at this scale.
Five: a visible crack. Usually at a slot root or the base of the step. At that point the horn is scrap. A cracked resonant component can let go and take the transducer above it with it. Stop using it.
What to measure
All five can be turned into a measurement, which is what a maintenance schedule needs.
| Check | How | When |
|---|---|---|
| Amplitude at the tip | Laser vibrometer, or a displacement gauge against a known reference | On install, then at fixed intervals |
| Frequency and impedance | Impedance analyser across the stack, compared to the new-horn baseline | Every horn change, every scheduled stop |
| Weld strength trend | Pull-off sampling, plotted over time rather than against one limit | Per shift or per batch, per your control plan |
| Tip condition | Visual with magnification, plus a photo against the new-part reference | Daily at start-up |
| Cracks | Dye penetrant on suspect areas, or replace on schedule before they start | At each deep clean |
Two of these are worth setting up even on a small line: a new-horn baseline record (amplitude, frequency, impedance, and a photo of the tip) and a pull-off trend chart.
With those two you’ll see signs one through four weeks before sign five appears, and you can plan a horn change instead of reacting to one.
What kills a horn early
The same horn run two ways lasts very differently. The variables that matter:
Running above design amplitude. Fatigue life falls as amplitude goes up. Cranking the generator to “get the weld through” buys a short-term fix and a much shorter tool life. Per our amplitude window data, going higher buys 5% more pull-off force for 46% more resistance.
Too much clamping force. You need pressure. Over-pressing loads the horn in a direction it wasn’t designed for and speeds up face wear.
Dirty or oxidised aluminium. Contamination accelerates pickup, pickup changes contact, and stress concentrates somewhere else.
Dry runs. Firing the stack in air with nothing to absorb the energy is one of the fastest ways to overheat and fatigue a horn.
Heat. Long duty cycles with no cooling let the stack drift in frequency and put extra stress on the horn.
Material choice. Aluminium, titanium and steel horns have different fatigue behaviour and different resistance to pickup. The right one depends on the workpiece, not on the price list.
Resurface or replace?
A worn face doesn’t mean a dead horn. But resurfacing only works within limits, because removing material changes the resonant geometry.
Resurface when it’s face wear or aluminium pickup with no cracking, and the amount to come off is small enough that the horn can be re-tuned back to frequency. It has to be re-measured after machining. A resurfaced horn that wasn’t re-tuned runs off-frequency, and that’s worse than the wear you were fixing.
Replace when there’s any crack, when the face has cupped past the usable limit, when the thread is damaged, or when it can’t be brought back to design frequency any more. And don’t keep a cracked horn as a spare. It will fail on the day you need it.
One more: if the horn-to-booster joint keeps coming loose, find out why. Re-torquing it every week doesn’t fix fretting damage, it compounds it.
Keep a spare
Horn replacement looks like a maintenance problem. It’s usually a lead time problem. A line that stops because the only horn is worn out loses a lot more than the price of a spare.
Keep at least one identical spare per machine, already characterised, with its baseline recorded. Then a swap doesn’t mean re-qualifying the process on the spot with production waiting.
Mark each horn with its own identity and record it against the machine. Otherwise your trend data says “the line”, and you can never tell which tool was responsible.
And for tooling made to your part geometry, expect a manufacturing lead time. Plan the spare before the old horn dies, not after.
We machine horns in-house, so the wear mechanisms above come from our own tool room: here’s how the plant is set up. If you need a replacement or a spare set for a line that’s already running, send the part drawing and the existing horn’s frequency.
FAQ
How often should an ultrasonic horn be replaced?
There’s no calendar answer. Life depends on amplitude, clamping force, workpiece material and duty cycle, so two identical horns on different lines can last very different amounts of time. Replace on measurement rather than on a schedule: track amplitude, resonant frequency and pull-off force against a new-horn baseline, and change it when those trends leave your process window.
Can a worn horn be re-machined instead of replaced?
Yes, if the damage is face wear or aluminium pickup and there’s no cracking. The face gets refinished and the horn re-tuned back to design frequency, then re-measured. A resurfaced horn that isn’t re-tuned runs off-frequency, which is worse than the wear it was meant to fix. Any crack means replacement, not repair.
How do I know if the amplitude has dropped?
Measure it. The generator’s percentage display shows the setpoint, not the actual tip displacement, so a worn horn can read the same on screen while moving less. Use a laser vibrometer or a displacement gauge at the tip and compare against a reading taken when the horn was new. Amplitude down at unchanged settings is the earliest sign of wear.
Does horn material matter?
Yes. Aluminium, titanium and steel differ in fatigue behaviour, acoustic properties and resistance to pickup. The right choice depends on the workpiece, the amplitude you need and the duty cycle, not on which one costs more.