
The problem
Symptom. Two injection molded housing halves leave the ultrasonic welder looking fine, then separate in a drop test, squeak under load, or crack around the boss a week later. Turning up the weld energy makes the flash worse but does not make the joint stronger.
What it costs. A production line running at 600 assemblies per hour can produce several thousand suspect parts before anyone notices. Scrap is the cheap part; the expensive part is a customer field return and a tool modification that needs another four weeks.
This guide gives you a repeatable diagnosis order, the joint dimensions that actually control strength, and an acceptance test you can put in the purchase order before the first article is made.
- 1. What a failed ultrasonic weld actually looks like
- 2. Trace the failure back to the joint, not the welder
- 3. The energy director is a specification, not a detail
- 4. Material and moulding choices that quietly break welds
- 5. Set an acceptance test before the first article
- 6. What to change first when the line is already running
What a failed ultrasonic weld actually looks like
An ultrasonic joint fails in a small number of recognisable ways, and each one points at a different stage of the process. A weld that never reached melt temperature leaves a dull, grainy interface with almost no material flow. A weld that was over-driven squeezes the energy director flat, pushes molten material out as flash, and leaves the two halves sitting on a hard stop with nothing left to bond them. A joint that was contaminated by mould release or moisture can look perfect and still peel apart by hand.
Before touching the welder, record which of these you are seeing. The distinction matters because the corrective actions are almost opposites: an under-welded joint needs more energy or a bigger director, while an over-welded one needs less energy and a mechanical stop.
| Observed symptom | Most likely cause | First check on the line |
|---|---|---|
| Joint peels apart, grainy unmelted interface | Energy director too small, or horn not contacting | Measure director height on 5 moulded samples |
| Heavy flash, halves bottomed out | Over-weld; no collapse stop | Check weld-by-distance setting and part height after weld |
| White stress marks around the boss | Far-field transmission into a rigid, unfilled grade | Measure distance from horn contact face to joint |
| Boss cracked circumferentially | Hoop stress from interference fit or oversized director | Check boss OD and mating hole clearance |
| Good-looking weld, low peel strength | Release agent, moisture, or regrind contamination | Weld a solvent-wiped control sample |
| Strength varies across the batch | Part-to-part flatness or shrink variation | Measure joint plane flatness across 20 parts |
Table 1 — Reading the failure before changing the weld schedule
Need a fixed price for this moulded part?
Send the 2D drawing, resin grade and annual volume. Our engineers return a DFM review, a material recommendation and a written quotation you can compare line by line.
Trace the failure back to the joint, not the welder
Ultrasonic welding delivers mechanical vibration to a joint through one half of the part. Everything between the horn contact face and the joint line either transmits that energy or absorbs it. Adjusting amplitude and weld time only changes how much energy arrives; it cannot repair a joint that was never designed to concentrate it.
A six-step diagnostic sequence
- Confirm the resin and the grade on both halves.
Two halves moulded from different grades, or one half with a different regrind percentage, will not fuse reliably. Check the lot labels before anything else. - Wipe a control sample and weld it.
If a solvent-wiped part suddenly welds, the problem is surface contamination, not the weld schedule. Mould release agent and silicone-based handling aids are the usual culprits. - Measure the energy director on unmoulded and moulded parts.
Use a profile projector or a calibrated microscope. A director specified at 0.30 mm that moulds at 0.16 mm has already lost half its melt volume. - Check joint-plane flatness across the batch.
A 0.15 mm bow means the horn loads one side of the joint first, so energy is spent flattening the part instead of melting the director. - Verify horn contact and support fixture.
The lower half must be supported directly under the joint. A compliant nest absorbs energy and produces a weak weld with a good-looking surface. - Only then adjust weld time and amplitude.
Run a short matrix and pull-test every point. Record collapse distance as well as time; collapse is the measurement that transfers between machines.
The chart below is a typical result from that last step. It shows why a small energy director cannot be rescued by longer weld time: the joint reaches a ceiling well below the acceptance line, while an oversized schedule on a good joint eventually degrades strength again.

The energy director is a specification, not a detail
On a near-field joint the triangular director concentrates strain into a very small volume, which is where melting starts. Its height sets how much molten material is available, and its width sets how quickly that material spreads. Both need to be dimensioned on the part drawing, with a tolerance, or the mould shop will make an aesthetic choice and the welding shop will inherit the consequence.
| Joint parameter | Typical target | What goes wrong outside it |
|---|---|---|
| Director height (semi-crystalline) | 0.30 – 0.50 mm | Below 0.25 mm the melt volume is insufficient; strength plateaus early |
| Director height (amorphous) | 0.20 – 0.35 mm | Over 0.40 mm drives flash and long weld times |
| Included angle | 60° – 90° | A blunt director spreads energy and delays melt initiation |
| Joint plane flatness | ≤ 0.10 mm over the joint | Bow loads one side first and produces uneven collapse |
| Target collapse | 60 – 80 % of director height | Full collapse means the halves have bottomed on a hard stop |
| Horn-to-joint distance | ≤ 6 mm for near-field | Longer paths need a rigid, low-loss grade and a full-wave horn |
Table 2 — Joint dimensions to put on the part drawing
For glass-filled grades the numbers tighten. Glass fibre does not melt, so the effective bonding area is smaller and a director that works in unfilled nylon will produce a weak joint in a 30 % glass-filled version of the same part. If the assembly must carry load, specify the filled grade on the drawing and qualify the weld on that grade, not on a cheaper unfilled trial.
Material and moulding choices that quietly break welds
Four decisions made during material selection and moulding have a larger effect on weld strength than the welder settings. Moisture in hygroscopic resins turns into steam at the joint and leaves a foamed, porous interface. Regrind changes melt viscosity and, above roughly 25 % in many engineering grades, reduces the repeatability of collapse. Internal and external mould release agents migrate to the surface and act as a bond breaker. And a colourant package with a high loading of certain pigments can change how the resin absorbs ultrasonic energy.
| Upstream choice | Effect at the joint | Practical control |
|---|---|---|
| Resin dried to spec, then held > 4 h in an open hopper | Foamed interface, low peel strength | Dry and weld within a defined window; use a hopper dryer |
| 25 – 40 % regrind | Variable collapse, batch-to-batch strength spread | Cap regrind at a qualified percentage and record the lot |
| Silicone-based mould release | Bond-line contamination invisible to the eye | Ban silicone release on welding parts; use a wipe test control |
| 30 % glass-filled grade substituted for unfilled | Loss of effective bond area | Requalify the weld schedule on the actual production grade |
| High cavity count with unbalanced filling | One cavity welds, another does not | Track weld results by cavity number, not just by batch |
Table 3 — Upstream decisions that determine weld performance
Set an acceptance test before the first article
Without a defined test, first-article approval becomes an argument about appearance. Agree a measurable criterion while the tool is still being built: a minimum pull-off or burst force, a drop or torque test that reflects the assembly's real load, and a sample plan that covers every cavity and both ends of the moulding window. Then weld a control set at the lower and upper process limits.
A workable specification is usually three lines: minimum joint strength as a percentage of parent material strength, maximum permitted flash, and the number of parts to be tested per cavity. Everything else can be negotiated later, but those three numbers decide whether the assembly is shippable.
Not sure whether your joint is weldable?
Send us the joint detail before the tool is cut. Our engineers will review the energy director, the grade and the horn access, and return a marked-up recommendation with any changes that would reduce tooling risk.
What to send us
- 2D drawing of the joint cross-section
- Resin grade and regrind percentage
- Required pull-off or burst force
- Expected annual volume and cavity count
What to change first when the line is already running
When the line is producing parts, the order of fixes is driven by how quickly each one can be validated. Process settings are free but have a low ceiling; tooling changes are slow but permanent. Work down the list, validating strength after each step rather than changing five things at once.
| Order | Change | Typical lead time | Strength gain in field cases |
|---|---|---|---|
| 1 | Dry the resin, wipe a control batch, verify horn seating | Same shift | 0 – 25 % |
| 2 | Re-tune collapse distance and hold time | 1 day | 10 – 30 % |
| 3 | Add a hard collapse stop to the fixture | 2 – 3 days | 15 – 35 % |
| 4 | Modify horn face or nest support | 1 week | 10 – 40 % |
| 5 | Enlarge or add the energy director in the tool | 2 – 4 weeks | 30 – 80 % |
Table 4 — Corrective actions ranked by validation speed
Before you release an ultrasonically welded housing
- Energy director height, angle and tolerance are dimensioned on the part drawing
- Joint-plane flatness is specified and checked across every cavity
- The production resin grade, colourant and regrind cap are written into the specification
- Mould release agents are prohibited on joint surfaces, with a wipe-test control
- A measurable minimum joint strength and maximum flash are agreed with the supplier
- First-article samples cover both ends of the moulding process window
Print this and attach it to the RFQ. Most weld failures are already decided at the drawing stage.
Can ultrasonically welded parts be reworked after a failed weld?
Rarely. Once the director has been consumed the joint has no reserve melt volume, so a second pass usually produces flash rather than strength. Plan the process so that a suspect assembly is scrapped or downgraded rather than re-welded.
How do I know if my joint is near-field or far-field?
Measure the distance from the horn contact face to the joint line. Under about 6 mm is near-field. Beyond that the part itself has to transmit the vibration, which requires a rigid, low-loss resin, generous radii to avoid stress concentration, and a horn tuned to the part geometry.
Why does one cavity weld well and another fail?
Unbalanced filling changes shrinkage and joint flatness cavity by cavity. Record the cavity number on every test part; a consistent cavity pattern almost always points to the mould rather than the welder.
Does a bigger energy director always make a stronger weld?
No. Beyond roughly 0.5 mm in semi-crystalline resins the director needs more energy to melt, produces more flash, and can leave voids as the excess material is displaced. Strength peaks in a band, which is why the weld schedule should be qualified with a pull test.
What is a sensible minimum strength criterion?
For structural housings, 70 – 80 % of the parent material's tensile strength at the joint, measured on a pull or burst test that reflects the real load path. For cosmetic covers a lower figure with a defined drop test is often more meaningful.
Can we weld parts made from two different plastics?
Only if they are chemically compatible and have similar melt temperatures. ABS to ABS, PC to PC and PMMA to PC blends are workable with qualification; dissimilar semicrystalline pairs generally are not. Ask for a weld trial on the exact two grades before committing the tool.
Get a weldable part, not just a quote
Send us the assembly and we will review joint design, material and tooling together, then return a DFM report and a fixed quotation with a defined first-article acceptance test.
To get an accurate quote, send:
- 3D file or 2D assembly drawing
- Resin grade and any regrind limit
- Required joint strength or load case
- Annual volume and target unit price
Request a DFM review and quote
Prefer to talk it through first?
Correio eletrónico [email protected] or message us on WhatsApp +86 177 1752 9949. Send a drawing, a sample or a specification and an engineer will reply.
Why buyers work with Uplastech
ISO 9001, ISO 14001 and ISO 45001 certified. Founded in 2016, with injection moulding lines in Shanghai and an extrusion plant in Jiangsu. Moulding machines from 90T to 1100T and a maximum shot weight of 5100 g. Samples and a written quotation are available on request.