Twelve reasons an ultrasonic weld fails, and how to tell which one you have

When a weld goes wrong, the temptation is to start turning knobs. Resist it. Ultrasonic welding has maybe a dozen realistic failure modes, and each one leaves a different fingerprint. Read the fingerprint first and you will usually find the cause in ten minutes.

Here is the order we work through on our own bench.

1. Weak weld, looks fine from outside

Parts hold in your hand, then come apart when someone pulls properly.

Usual causes, in order of how often we find them:

• Not enough material in the energy director. The bead melts, the parts bottom out, and there is nothing left to fill the joint. Measure the bead. If it is 0.3 mm high and the parts travel 0.5 mm, the maths does not work.
• Fixture lets the lower part flex. If the base can move, the vibration goes into moving the part rather than melting the joint. Your nest should support the joint line directly from below, not just hold the part loosely.
• Worn horn face. A horn that has done a few hundred thousand cycles on filled resin is no longer the shape it was.
• Amplitude too low for the material. Common when someone moves an ABS setup onto a PP part without changing the booster.

2. Weld strong in places, absent in others

Classic sign of uneven pressure. Look at the flash — if the flash bead is thick on one side of the part and nonexistent on the other, the horn face and the part are not parallel.

Check horn face flatness, check the fixture is not sitting on a chip of plastic, and check the part itself for warp out of the mould. Warped mouldings are the single hardest thing to weld consistently, and no machine setting compensates for them.

3. Marking or burning on the visible surface

Three things do this:

• The horn face is rough, scratched or has plastic stuck to it. Polish it.
• Amplitude is higher than the job needs. Try a lower-gain booster before you try anything else.
• The horn is pressing on a thin, unsupported area, so the plastic flexes and heats where you did not intend.

A sheet of thin PTFE film between horn and part is a useful diagnostic. If the marking disappears, it is a contact problem, not an energy problem.

4. Excessive flash

Too much energy, too much travel, or an energy director carrying more material than the joint needs. Reduce weld time or switch from time mode to energy or depth mode so the machine stops when the joint is full rather than when the clock runs out.

If the part design allows it, move the energy director inward from the edge so the flash forms where nobody sees it.

5. Diaphragming — the part rings and cracks away from the joint

A large flat panel far from the horn face starts vibrating like a drum skin. Stress concentrates somewhere that has nothing to do with the weld, and the part cracks there.

Almost always a far-field problem on a semi-crystalline material. The cure is design: support the panel from below, add a rib, or bring the horn closer to the joint. Sometimes a lower frequency with a bigger horn spreads the load enough to stop it.

6. Internal components shaken to bits

The weld is perfect. Something inside has snapped off, or a PCB solder joint has failed, or a small spring has jumped its seat.

Reduce amplitude, shorten weld time, and consider a higher frequency for that part. This is one of the few genuine reasons to buy a 30 or 35 kHz machine.

7. Plastic sticking to the horn

Common on soft materials and on PE. Options, cheapest first: polish the horn face, drop the amplitude, add a very slight air blast, and if it persists, have the horn hard-coated or carbide-faced.

8. Horn cracks

This one costs real money, so it is worth understanding.

The overwhelming cause is an under-torqued stud between converter, booster and horn. A joint that is not tight does not transmit cleanly — it generates heat at the interface, and that heat fatigues the metal until it lets go. Faces must be clean and flat, and the stud must be torqued to the figure in your machine's manual. Not "tight by hand".

Other contributors: running the stack unloaded for long stretches, running amplitude above what the horn was designed for, and sharp internal corners in the horn geometry itself. A horn that cracks repeatedly in the same place was designed wrong, and we would rather redesign it than keep selling you replacements.

9. Converter or generator running hot

Check duty cycle first. A machine rated for intermittent use will overheat on a continuous line, and no amount of settings will change that — you need a water or air cooled chiller, or a machine specified for continuous duty.

After that, check the stack is properly tuned. An out-of-tune stack dumps energy as heat instead of motion.

10. Overload trips

Work through it in this order:

1. Does the stack tune cleanly with no load? If not, the problem is the stack, not the part.
2. Is the booster right for this job? Too much gain and the generator sees a load it cannot drive.
3. Is the horn too big for the wattage? A 15 kHz horn on an underpowered generator will trip every time it meets a real part.
4. Is the joint simply asking for more energy than the machine has? Filled PP on a long seam needs 3,000 W and up. This is not a fault, it is a specification mismatch.

11. Fails the leak test

An energy director joint is not a sealing joint. If you need hermetic, you need a shear joint or tongue-and-groove with correct interference — see the joint design guide. Trying to make a butt joint hermetic by welding harder produces flash and a leak, in that order.

12. It worked yesterday and does not today

Nothing on the machine changed, so something else did. In our experience, in this order:

• New batch of mouldings, different regrind ratio or a different resin lot.
• Parts have been sitting and have picked up moisture — nylon and PET especially, and much faster in monsoon.
• Somebody cleaned the horn with the wrong thing, or did not clean it at all.
• The stud has worked loose. Check the torque.
• Air pressure has drifted. Weld pressure depends on it.
• Mains voltage is sagging. On a 220 V single-phase supply with other machines starting up on the same line, this is more common than people think.

The two-minute check we run first

Before anything else: take the part out, run the stack unloaded, and listen. A clean stack sounds clean. Then put a fingernail's worth of pressure on the horn face and listen again. If the note changes wildly or the generator complains with no load, stop looking at the part — the problem is in the stack.

If the stack is healthy, the fault is in the part, the fixture, or the settings, in roughly that order of likelihood.

Stuck on one? Send us a photo of the failed weld and a photo of the joint design. We have seen most of them, and we would rather tell you it is a fixture problem than sell you a machine you do not need.

Questions we get asked