Ultrasonic Welding of Plastics: Energy Directors, Viscoelastic Heating, Shear Joints and the Acoustic Stack

How ultrasonic welding works: viscoelastic heating, energy director geometry, shear joints for semi-crystalline resins, horn design, process control, testing and India costs.

·

Ultrasonic Welding of Plastics: Energy Directors, Viscoelastic Heating, Shear Joints and the Acoustic Stack

Take a 60 × 40 mm polycarbonate enclosure with a 2 mm wall. The joint runs 200 mm around the perimeter. Mold a 0.5 mm base, 0.25 mm high triangular ridge on one half, press the two halves together under a few hundred newtons, and drive the top half at 20 kHz through 30 µm of peak-to-peak displacement. The entire weld takes about 0.3 seconds. No screws, no adhesive, no solvent, no cure oven.

The ridge holds 12.5 mm³ of plastic — 15 mg. Heating it from 25 °C to 250 °C takes 4.05 J. Heating a 0.1 mm layer of the mating land underneath takes another 12.96 J. The machine, meanwhile, is delivering on the order of 240 J of mechanical work at the horn face over that 0.3 s. Roughly 7% of the delivered energy does the melting. The rest heats the bulk of the part, the horn, the fixture, and the air.

That imbalance is the defining feature of ultrasonic welding. It is a poor melting process and an excellent local heating process. The energy is deposited precisely where the two surfaces touch, in a layer thin enough that the bulk never has time to heat. Get the joint geometry right and the process is faster than any competing assembly method. Get it wrong and the machine will happily deliver 240 J into a cold weld, a cracked boss, or a flash ring that has to be trimmed by hand.

This guide works the numbers behind that outcome: the viscoelastic heating rate, the thermal penetration depth that sets the weld window, the energy director geometry that decides where the strain concentrates, the shear joint that semi-crystalline resins require, the acoustic stack that delivers the amplitude, and the economics that decide whether the process belongs in a product at all.

1. One cycle, and what each part of the stack does

An ultrasonic welder is a resonant mechanical system, not a heater. The generator converts 50/60 Hz mains into 20 kHz (or 15, 30, 40 kHz) electrical drive. A piezoelectric converter turns that into longitudinal displacement — typically 20 µm peak-to-peak at 20 kHz, 31 µm at 15 kHz, 13.5 µm at 30 kHz, 9 µm at 40 kHz. A booster multiplies or attenuates that displacement by a fixed gain, usually between 0.4 and 2.5. The horn — the sonotrode — is a half-wave resonator machined to the part's shape; it multiplies the amplitude again, typically by 1 to 3, and delivers it to the workpiece.

The cycle has five phases:

  1. Load. The two halves are placed in a fixture. The joint must self-locate or be located by pins; the energy director does not align anything.
  2. Horn descent and trigger. The horn contacts the top part and the machine registers a trigger force, typically 50–300 N. The trigger point matters: if it fires late, the horn has already compressed the part; if it fires early, the weld starts before contact is stable.
  3. Weld. Ultrasound is applied for a set time, energy, or collapse distance. The weld phase is typically 0.2–1.0 s, and for small parts often 0.2–0.4 s.
  4. Hold. The vibration stops but the clamp force stays on while the melt solidifies. Hold is typically 0.1–0.5 s. Higher hold force improves strength and hermeticity by squeezing out voids and bringing the surfaces into intimate contact.
  5. Retract and unload. The horn lifts, the part is removed. Total cycle time is usually 3–10 s including handling.

The frequency choice is a trade between amplitude and delicacy:

Frequency · Converter amplitude · Typical part size · Best for · Notes

15 kHz · 31 µm · Large (>150 mm) · Thick walls, semi-crystalline, high energy · Audible noise; large horns; highest power

20 kHz · 20 µm · Medium (20–150 mm) · General structural welding · The industry default

30 kHz · 13.5 µm · Small (5–50 mm) · Delicate parts, thin walls · Less noise, lower stress

40 kHz · 9 µm · Micro (<20 mm) · Medical, electronics, films · Small horn dimensions, tight tolerances

flowchart TB
  G["Generator<br/>20 kHz electrical drive"] --> C["Converter<br/>PZT stack, 20 µm"]
  C --> B["Booster<br/>gain 1.5"]
  B --> H["Horn / sonotrode<br/>gain 2.0, half-wave resonator"]
  H --> P["Top part<br/>energy director at joint"]
  P --> J["Joint interface<br/>melt layer 0.1–0.3 mm"]
  J --> F["Bottom part + fixture<br/>rigid anvil, no compliance"]

The stack is tuned to resonance. Every element — converter, booster, horn — is a half-wave resonator at the same frequency, and the joints between them are at displacement nodes, where stress is highest and motion is lowest. A horn that is 2 mm too long, or a stud that is not torqued to spec, shifts the resonance, and the generator either faults on overload or drives the stack at a frequency where the amplitude at the face is a fraction of its design value.

The key division of responsibility: the machine controls force, time, and amplitude; the part design controls where the strain — and therefore the heat — goes. No machine setting can rescue a joint that puts the energy director in the wrong place, or a wall that flexes instead of transmitting.

2. The heating is viscoelastic, not frictional

The popular explanation is that ultrasonic welding "creates friction between the two surfaces." That is only true for the first few milliseconds, before full contact. Once the surfaces are pressed together and moving together, the dominant mechanism is viscoelastic dissipation: the polymer is cyclically strained, and its loss modulus converts part of that mechanical work into heat.

For a sinusoidal strain of amplitude \varepsilon_0 at angular frequency \omega = 2\pi f, the volumetric heat generation rate is

where E'' is the loss modulus at the driving frequency and temperature. This is the Tolunay relation used in ultrasonic welding models; the same form appears in ultrasonic consolidation and plasticizing models. The important structure is:

Worked example — polycarbonate at 20 kHz, E'' = 100 MPa, \varepsilon_0 = 2\%:

The volumetric heat capacity of PC is \rho c_p = 1200 \times 1200 = 1.44\times10^6 J/m³·K, so the initial heating rate is:

Heating from 25 °C to 250 °C — a 225 K rise — takes 129 ms at that rate, before heat losses and before the land starts to melt. That is the right order of magnitude for a 0.2–0.4 s weld.

The strain amplitude is not the horn amplitude divided by the part thickness. It is the local cyclic strain in the energy director, approximately

where A is the peak-to-peak horn amplitude and h is the energy director height. For A = 30 µm and h = 0.5 mm, that is 6%. Local compliance, the taper of the triangle, and the constraint of the mating surface reduce the effective strain; 2–4% is a realistic working range. The equation is a scaling law, not a precision prediction — but it correctly predicts that a shorter energy director experiences higher strain and heats faster, which is why very small directors can flash while a tall one stays cold.

E'' · \varepsilon_0 = 1\% · 2% · 3% · 5%

50 MPa · 314 W/cm³ · 1,257 · 2,827 · 7,854

100 MPa · 628 W/cm³ · 2,513 · 5,655 · 15,708

200 MPa · 1,257 W/cm³ · 5,027 · 11,310 · 31,416

Two practical consequences follow. First, amplitude calibration is not optional. A horn measured at 28 µm instead of 30 µm delivers 13% less heat; a horn at 34 µm delivers 28% more. The difference between a cold weld and a flash ring can be a 15% amplitude drift caused by a cracked converter or a mistuned horn. Second, the same machine settings do not transfer between materials. PC at 2% strain heats at 1,745 K/s; a glass-filled PA66 with a different E'' and a much higher melting point may need three times the energy, and a low-modulus PP may damp the wave before it reaches the joint at all.

3. Thermal penetration sets the weld window

The heat generation is volumetric, but the heat removal is diffusive. The competition between the two sets the thickness of the melt layer and the time available to form a bond. For a pulsed surface heat source, the thermal penetration depth scales as

For polycarbonate, k = 0.20 W/m·K, \rho = 1200 kg/m³, c_p = 1200 J/kg·K:

Weld time · Penetration depth \delta

50 ms · 0.083 mm

100 ms · 0.118 mm

200 ms · 0.167 mm

300 ms · 0.204 mm

500 ms · 0.264 mm

1.0 s · 0.373 mm

The inverse is more useful for process design:

Target melt depth · Required time

0.05 mm · 18 ms

0.10 mm · 72 ms

0.20 mm · 288 ms

0.30 mm · 648 ms

This is why production weld times cluster at 0.2–0.5 s for small and medium parts: that is the time required to melt a 0.15–0.25 mm layer, which is enough to flow across the joint and fuse, but not enough to heat the whole wall. It also explains why longer is not stronger. Doubling the weld time from 0.3 s to 0.6 s increases the melt depth by only 41% (because \delta \propto \sqrt{t}), while doubling the heat input to the surface. The extra energy goes into overheating the interface, degrading the polymer, generating flash, and warping the part.

The cooling side uses the same diffusion time. A 0.2 mm melt layer needs roughly 0.3 s to freeze; a 0.1 mm layer needs about 0.07 s. Hold times of 0.1–0.5 s are matched to that, not chosen arbitrarily. If the part is released while the interface is still molten, the joint is disturbed and the strength drops; if the hold is much longer than needed, cycle time is wasted.

Semi-crystalline resins add a second time scale: crystallization. The melt must not only cool below the melting point, it must re-solidify with enough crystallinity to carry load. The narrow melting range that makes these resins dimensionally stable also makes the weld window narrow. The joint can go from unmelted to fully molten over a 10–20 K band, and the resolidification is equally abrupt. This is the fundamental reason a simple energy director works poorly on PP, PE, PA, POM, and PBT: the melt flows out of the director and freezes before it has wet the whole land.

4. Energy audit: where the joules actually go

The mechanical power delivered by the horn is force times velocity. For sinusoidal displacement of amplitude A (peak-to-peak), the RMS velocity is

and the average coupled power is

Worked example — F = 300 N, A = 30 µm, f = 20 kHz:

Over a 0.3 s weld, that is 240 J of mechanical work at the horn face. The electrical draw is higher — generator efficiency is typically 80–90%, and the stack has its own losses — so the wall-plug energy might be 280–320 J. (The calculation also assumes the force and velocity are in phase; reactive loading reduces the real power, so treat 240 J as an upper bound on coupled work.)

Now compare the enthalpy actually needed to melt the joint. For the 60 × 40 mm box, 2 mm wall, 200 mm perimeter, 0.5 mm base and 0.25 mm high energy director:

The land beneath the director also melts. A conservative estimate is a 0.1 mm layer across the full 2 mm joint width:

Total joint melt enthalpy: 17.0 J. Against 240 J delivered, the melting efficiency is 7.1%.

That number is not a defect of the process; it is the price of locality. A hot-plate welder heats the entire joint region for 10–30 s and conducts heat through the parts. Ultrasonic welding puts the energy into a 0.2 mm layer in 0.3 s and leaves the rest of the part at room temperature. The same 240 J spread over the whole 60 × 40 × 2 mm box (volume 4,800 mm³, mass 5.76 g) would raise its temperature by only 35 K — which is exactly why the bulk survives.

For semi-crystalline resins, add the latent heat of fusion. PP has \Delta H_f \approx 90–100 J/g for a molded part, and the exact value depends on crystallinity. A 15 mg PP energy director needs roughly

about 33% more than the PC director for the same mass, before accounting for PP's higher amplitude requirement and greater acoustic damping. That is the quantitative version of the familiar statement that "semi-crystalline plastics are harder to weld."

Machine sizing follows from the power, not the energy. A 2 kW, 20 kHz machine can deliver 800 W average comfortably at a 40% duty cycle; a 3.3 kW machine is needed for large horns, high force, or short cycle times where the instantaneous demand is high. Buying a 4 kW machine for a 200 J weld wastes capital and often runs at a poor power factor.

5. Energy director geometry

The energy director is a triangular ridge molded into one half of the joint. Its job is to do three things at once: limit the initial contact area, concentrate the cyclic strain, and supply a known volume of melt to fill the interface.

The standard geometry is a right isosceles triangle for amorphous resins (90° included angle, height = half the base) and an equilateral triangle for semi-crystalline resins and polycarbonate (60° included angle, height = 0.866 \times base). The 60° director is taller and sharper for the same base width; it penetrates the mating surface earlier, melts a larger volume, and gives the semi-crystalline melt a better chance to wet the land before it freezes.

The dimensional rules from the joint-design literature:

For a 2 mm wall, a 0.5 mm base gives a 0.25 mm high, 90° director. The cross-section is 0.0625 mm², and the melt volume per 100 mm of joint is 6.25 mm³. If that volume is spread over the 2 mm land, the average bondline is:

A 31 µm bondline is thin — which is what you want. The strength of a welded joint comes from intimate molecular contact across a thin layer, not from a thick glue line. A thick melt layer has a longer cooling time, more shrinkage, and more voids.

Wall · ED base · ED height · Cross-section · Melt volume / 100 mm · PC melt energy / 100 mm

1.5 mm · 0.38 mm · 0.19 mm · 0.036 mm² · 3.6 mm³ · 1.17 J

2.0 mm · 0.50 mm · 0.25 mm · 0.063 mm² · 6.3 mm³ · 2.03 J

3.0 mm · 0.75 mm · 0.38 mm · 0.141 mm² · 14.1 mm³ · 4.56 J

4.0 mm · two × 0.40 mm · two × 0.20 mm · 0.080 mm² · 8.0 mm³ · 2.59 J

The 4 mm case uses two smaller directors rather than one large one. The rule for thick walls is that the sum of the director heights should equal about 10% of the joint width, and each director should be about 60% of the size that would be used alone. Multiple directors spread the melt across the width and avoid a single deep triangle that would take too long to melt and create a wide, weak bondline.

Three design features are as important as the director itself:

A collapse stop. Mold a ledge or step that bottoms out at the target collapse — typically 0.8–1.1 times the director height. Without a stop, the pneumatic press continues to drive after the director melts, and the parts collapse until the horn force is balanced by the fixture. The result is a weld with variable bondline thickness and a part with a variable height. With a stop, the collapse is mechanically limited and the weld becomes a distance-controlled process even on a time-mode machine.

A flash trap. If the melt volume exceeds the joint gap, excess polymer has to go somewhere. A small groove or reservoir adjacent to the joint catches it. Without one, the flash appears on the cosmetic surface, and someone has to trim it — which is exactly the labor cost the process was supposed to eliminate.

Separate alignment. The energy director does not align the parts. Pins, sockets, or a step in the wall do. If the parts can shift laterally during the weld, the director can miss the land, and the joint is weak on one side.

flowchart TD
  A["Thermoplastic joint"] --> B{"Amorphous or semi-crystalline?"}
  B -->|"Amorphous"| C{"Strength / hermetic requirement?"}
  C -->|"Cosmetic, moderate strength"| D["Butt joint + 90° ED"]
  C -->|"Strong, well aligned"| E["Step joint + 90° ED"]
  C -->|"Hermetic, PC/PSU"| F["Tongue-and-groove + 60° ED"]
  B -->|"Semi-crystalline"| G{"Can the walls telescope?"}
  G -->|"Yes"| H["Shear joint<br/>0.2–0.5 mm interference"]
  G -->|"No, thin wall"| I["Scarf joint<br/>30–60°"]
  H --> J["Flash well + fixture support"]
  I --> J
  D --> K["Collapse stop + alignment pins"]
  E --> K
  F --> K

6. Shear joints and the semi-crystalline problem

A semi-crystalline resin does not soften gradually. It holds its modulus until the crystalline lamellae melt, then drops to a low-viscosity liquid over a narrow temperature band, then recrystallizes just as abruptly on cooling. The melt that flows out of an energy director can solidify before it reaches the far side of the land. The resulting weld is limited to the base width of the director — a narrow strip in the middle of the joint, with the rest of the land unbonded.

The shear joint solves this by changing the geometry of the melt. One half has a recess; the other has a matching step with an interference fit. The parts telescope together as they melt. The initial contact is a small vertical band, and as the parts close, the molten walls smear along each other, eliminating voids and exposing fresh melt to the mating surface. The result is a strong, hermetic joint that does not depend on the melt flowing across a wide land.

Dimensions from the joint-design standard:

Maximum part dimension · Interference

< 19 mm · 0.20–0.30 mm

19–38 mm · 0.30–0.40 mm

> 38 mm · 0.40–0.50 mm

The vertical weld depth is typically 1.0–1.5 mm; it controls the joint strength and can be increased for higher loads. A lead-in of 0.4–0.6 mm guides the parts together, and a flash well of 0.13–0.20 mm clearance contains the displaced material. The outer wall must be supported by the fixture and must be thick enough not to bulge under the hoop stress of the interference fit.

The melt-volume arithmetic explains the cycle-time penalty. For a 0.25 mm interference and a 1.2 mm weld depth:

against 0.0625 mm³/mm for the butt-joint director — a 4.8× increase. The literature reports shear joints requiring 3–4× the weld time of an energy-director joint, which is consistent: more material, more melt, more time. The shear joint is not a free upgrade. It is the correct choice when the resin demands it, and it costs cycle time, fixture complexity, and flash control.

The scarf joint is the fallback for thin walls and circular parts where a shear joint will not fit. The two mating faces are cut at 30–60°; the angles must match within 1.5°, and the outer wall at the scarf must be at least 0.76 mm to avoid blow-through. Scarf joints are sensitive to concentricity and molding tolerance, which is why they are less common than shear joints in production.

Joint · Best material · Strength · Hermeticity · Tooling complexity · Typical use

Butt + ED · Amorphous · Moderate · Poor–fair · Low · Consumer housings, cosmetic parts

Step + ED · Amorphous · Good · Fair · Medium · Enclosures with alignment

Tongue-and-groove · Amorphous, PC · Good · Good · Medium · Hermetic small assemblies

Shear · Semi-crystalline · High · Excellent · High · Automotive, medical, fluid handling

Scarf · Semi-crystalline, thin wall · High · Good · High · Circular/oval parts, thin walls

7. Near-field, far-field, and what the wave actually does

The industry divides ultrasonic welding into near-field and far-field by the distance from the horn face to the joint. The usual boundary is 6 mm. Joints within 6 mm are near-field; joints deeper in the part are far-field.

The physics behind the rule is not a simple wavelength argument. In a thin plastic wall, the relevant wave is an extensional (plate) mode, with speed

For polycarbonate, E \approx 2.4 GPa and \rho = 1200 kg/m³ give c_{\text{ext}} = 1{,}414 m/s, so the wavelength at 20 kHz is 70.7 mm. For ABS it is 69.3 mm, for PP 64.4 mm, for PA66 81.1 mm, for PEEK 83.2 mm. A 6 mm path is only 7–9% of a wavelength — geometrically, it is still the near field.

What changes at 6 mm is mode conversion and attenuation. The horn is a piston radiating into a finite plate. Close to the horn, the motion is predominantly axial and the joint sees the horn's displacement almost directly. Further away, the plate bends, thickness modes appear, and the energy spreads laterally. Semi-crystalline resins scatter the wave at spherulite boundaries; low-modulus resins damp it; abrupt section changes reflect it. The result is that the amplitude at the joint can be a fraction of the amplitude at the horn face, and the fraction is not predictable from a simple formula.

Material · c_{\text{ext}} · \lambda at 20 kHz · Transmission · Practical rule

PS · ~1,700 m/s · ~85 mm · Excellent · Far-field possible

ABS · 1,387 m/s · 69 mm · Very good · Far-field up to ~100 mm

PMMA · ~1,600 m/s · ~80 mm · Very good · Far-field possible

PC · 1,414 m/s · 71 mm · Good · Near-field preferred above 6 mm

PA66 · 1,622 m/s · 81 mm · Moderate · Near-field; dry resin

PP · 1,287 m/s · 64 mm · Poor · Near-field only

PE · ~1,000 m/s · ~50 mm · Poor · Near-field only

PEEK · 1,664 m/s · 83 mm · Moderate · Near-field, high amplitude

For thick sections, the bulk longitudinal speed is higher — PS 2,350 m/s, PA6 2,570 m/s, PP 2,550 m/s — because the lateral constraint adds stiffness. The mode that actually carries the energy depends on the wall thickness relative to the wavelength. A 2 mm wall at 20 kHz is about \lambda/35; it behaves as a plate. A 20 mm section is a substantial fraction of a wavelength — about \lambda/3.5 — and can support a thickness resonance. This is why a part that welds well at 2 mm can behave completely differently at 6 mm, even with the same material and joint.

Design rules that follow:

  1. Keep the joint within 6 mm of the horn face unless the material is a rigid amorphous resin with good transmission.
  2. Stiffen the load path. Ribs, bosses, and uniform wall sections transmit amplitude better than large flat panels that flex.
  3. Avoid abrupt thickness changes between the horn contact and the joint. A step reflects the wave and creates a cold spot.
  4. Measure the amplitude at the joint if the part is far-field. A laser vibrometer or a calibrated PVDF sensor on a prototype is cheaper than a production line that welds on one side only.
  5. Do not assume the horn face amplitude is uniform. Large horns can have 10–20% variation across the face; the part welds where the amplitude is highest.

8. The acoustic stack: converter, booster, horn

The stack is a chain of half-wave resonators. Each element is designed so that its length is a half wavelength at the operating frequency:

For 7075-T6 aluminum, c = 5{,}005 m/s. At 20 kHz, the half-wave length is 125.1 mm; at 15 kHz, 166.8 mm; at 30 kHz, 83.4 mm; at 40 kHz, 62.6 mm. For Ti-6Al-4V, c = 5{,}080 m/s, the 20 kHz half-wave is 127.0 mm. For steel, c = 5{,}900 m/s, it is 147.5 mm.

Material · c (m/s) · 15 kHz · 20 kHz · 30 kHz · 40 kHz

Al 7075-T6 · 5,005 · 166.8 mm · 125.1 mm · 83.4 mm · 62.6 mm

Ti-6Al-4V · 5,080 · 169.3 mm · 127.0 mm · 84.7 mm · 63.5 mm

Steel · 5,900 · 196.7 mm · 147.5 mm · 98.3 mm · 73.8 mm

Amplitude is multiplied through the stack:

A 20 µm converter with a 1.5 booster and a 2.0 horn gives 60 µm at the face. That is in the middle of the 40–80 µm window for polycarbonate at 20 kHz. A 1.0 booster and a 3.0 horn also give 60 µm, but the horn is now doing more of the amplification, which means higher internal stress and a shorter fatigue life.

Horn gain depends on the shape:

Horn material · Advantages · Limitations · Typical use

Al 7075-T6 · Cheap, easy to machine, low loss · Poor wear resistance, fatigue cracks at high amplitude · Soft resins, low volume, prototype

Ti-6Al-4V · High fatigue strength, good wear, stable · 3–5× material cost, harder to machine · Glass-filled resins, high amplitude, production

Hardened steel · Wear resistance, hardness · Heavy, lower gain, difficult to tune · Abrasive fills, special inserts

Coated aluminum · Wear surface without titanium cost · Coating can spall; limited life · Moderate-volume abrasive parts

The horn is part-specific tooling. It must be designed with the joint in mind: the face should cover the joint, be flat and parallel to the mating surface, and have enough mass to avoid flexing. Large flat faces need ribs or a composite horn; a thin plate horn will "drum" and produce uneven welds. The face is usually knurled or textured to grip the part, except on cosmetic surfaces where a smooth, polished face and a protective film are used.

Horn design is a modal problem. The horn is modeled with finite elements, tuned to the operating frequency within about ±50 Hz, and checked for stress at the target amplitude. The node — the plane of zero displacement — is where the horn is mounted to the booster; mounting anywhere else damps the resonance and couples vibration into the press. The stud that joins the horn to the booster is torqued to a specified value, usually 20–40 N·m for a 20 kHz stack; a loose stud heats, frets, and eventually fractures.

A horn that cracks usually does so at a fillet or a sharp internal corner, after 10⁵–10⁷ cycles. The fix is a larger fillet, a smoother gain transition, or titanium. A horn that "stops welding" without cracking is usually mistuned — the face amplitude has dropped, and the generator is drawing current without delivering displacement. Measuring the face amplitude with a laser vibrometer is the only way to distinguish the two.

9. Process control: time, energy, distance

The machine has three primary control modes, and they are not equivalent.

Time mode applies ultrasound for a fixed duration. It is open-loop: the energy delivered depends on the part fit, the horn coupling, and the material batch. A tight fit draws more power; a loose fit draws less. The weld strength varies accordingly. Time mode is acceptable for low-volume or non-critical parts, but it is the wrong choice for production.

Energy mode integrates power over time and stops when a target energy is reached. It compensates for some part-to-part variation, but it still cannot distinguish energy that went into the joint from energy lost at the horn/part interface or in the stack. A contaminated horn face can absorb energy without melting the joint.

Distance (collapse) mode stops when the horn has traveled a set distance into the part. It is the closest to a direct measurement of melt volume: if the director height is known and the stop is at the right collapse, the same amount of material has melted. It compensates for molded tolerance, batch variation, and horn wear. The best production machines combine distance control with energy and time limits as a check.

Parameter · Typical range · What it controls · Failure if wrong

Frequency · 15/20/30/40 kHz · Amplitude, part size · Mistuned stack, low amplitude

Amplitude · 20–120 µm · Heating rate (squared) · Cold weld or flash

Weld force · 200–1,000 N (0.5–2 MPa) · Contact, melt film thickness · Poor coupling or melt squeeze-out

Trigger force · 50–300 N · Start point · Late/early weld, part marking

Weld time · 0.2–1.0 s · Energy input · Under/over weld

Hold time · 0.1–0.5 s · Solidification · Weak joint, voids

Downspeed · 10–50 mm/s · Impact, melt flow · Part cracking, flash

Collapse · 0.15–0.35 mm · Melt volume · Cold weld or over-collapse

A worked collapse window: a 0.25 mm director should collapse by 0.20–0.28 mm at a good weld. Less than 0.15 mm means the director did not fully melt — check amplitude, force, and contamination. More than 0.35 mm means the horn has driven through the director into the land, producing flash and a thin bondline — check the stop, the force, and the weld time.

The control mode also determines what you can monitor. Time mode gives you time and peak power. Energy mode gives you joules and peak power. Distance mode gives you collapse, energy, time, and peak power. The last is the only one that supports meaningful statistical process control: plot collapse and peak power on an X-bar/R chart, set limits at ±3σ, and investigate shifts. A Cpk of 1.33 on collapse is a realistic production target; a Cpk below 1.0 means the process is not capable and the joint design or the molding tolerance needs work.

Amplitude profiling is available on modern digital generators: start at high amplitude to initiate melting, then reduce to control viscosity and avoid flash. It is useful on semi-crystalline resins and on parts with delicate features near the joint. It is not a substitute for correct joint geometry.

10. Materials: what welds, what fights back

The single most useful material classification is amorphous versus semi-crystalline.

Amorphous resins — PS, ABS, SAN, PMMA, PC, PPO, PSU, PEI, rigid PVC — have a random molecular arrangement, a broad softening range above T_g, and good acoustic transmission. They weld over a wide window, flow gradually, and form strong bonds with a simple energy director.

Semi-crystalline resins — PE, PP, PA, POM, PBT, PET, PPS, PEEK — have ordered lamellae, a sharp melting point, and higher acoustic damping. They need higher amplitude, higher energy, and a shear joint for consistent strength. The melt resolidifies quickly, so the joint must be designed to keep the melt in contact with both surfaces.

Resin · Structure · T_g / T_m · 20 kHz amplitude · Weldability · Joint

PS · Amorphous · 100 °C · 30–70 µm · Easy · Butt + 90° ED

ABS · Amorphous · 105 °C · 30–70 µm · Easy · Butt/step + 90° ED

PMMA · Amorphous · 105 °C · 40–70 µm · Easy · Butt + 90° ED

PC · Amorphous · 147 °C · 40–80 µm · Moderate · 60° ED or shear

PC/ABS · Amorphous · 110–140 °C · 60–100 µm · Moderate · Step + 90° ED

PPO · Amorphous · 150 °C · 50–90 µm · Moderate · Step + 90° ED

Rigid PVC · Amorphous · 80 °C · 40–80 µm · Moderate · Butt + 90° ED; ventilate

PA6/66 · Semi-crystalline · 220–260 °C · 70–120 µm · Difficult · Shear

PBT · Semi-crystalline · 225 °C · 70–120 µm · Difficult · Shear

POM · Semi-crystalline · 165 °C · 70–120 µm · Difficult · Shear

PP · Semi-crystalline · 165 °C · 90–120 µm · Difficult · Shear/near-field

PE · Semi-crystalline · 130 °C · 90–120 µm · Difficult · Shear/near-field

PPS · Semi-crystalline · 280 °C · 80–120 µm · Difficult · Shear

PEEK · Semi-crystalline · 340 °C · 70–120 µm · Difficult · Shear, high force

The amplitude ranges are starting points from vendor data; the upper end is for far-field, glass-filled, or high-modulus grades. They are not a substitute for a weld trial.

Glass fill changes the problem in three ways. The fibers raise the bulk modulus and can raise absolute joint strength, but the weld line is resin-rich and the joint efficiency relative to the filled bulk usually falls. The fibers are abrasive: an aluminum horn wears at the face and at the edges, and the wear changes the amplitude. A 30–50% GF compound is a titanium-horn application. Fiber orientation near the joint also creates anisotropic melt flow, so the weld strength is direction-dependent.

Moisture is the most common cause of a good-looking weld that fails a leak test. PA, PBT, PET, and PC absorb water; at welding temperature, the water flashes to steam, creating voids, bubbles, and hydrolysis of the polymer. PA66 should be dried to below about 0.1% moisture — typically overnight at 60–80 °C in a desiccant dryer — and welded promptly. A molded part that has been sitting on a shelf in a humid climate is not the same part that was molded.

Mold release is the second most common cause. Silicone release agents migrate to the surface and form a low-surface-energy layer that prevents molecular contact. The weld looks fine and fails at low load. Specify silicone-free molding, clean parts with IPA before welding if there is any doubt, and never use a release spray on the joint surfaces.

Regrind is acceptable up to about 20–30% in many amorphous resins, provided it is dried and has not been overheated. Repeated heat history degrades molecular weight, and the weld strength tracks the bulk strength. DVS 2216-1 treats recyclate as a material variable that must be controlled, not assumed.

Dissimilar materials weld only when they are chemically compatible. ABS/PC, ABS/PMMA, and ABS/SAN are reasonable pairs. PE/PP, PA/POM, and most semi-crystalline pairs are not — the melts do not interdiffuse, even if both melt. For incompatible pairs, use a mechanical interlock (a snap, a swaged lip, a molded dovetail) or an adhesive, not a weld.

PVC deserves a special note. It welds well, but it releases hydrogen chloride at processing temperature. The gas is corrosive to the machine, the horn, and the fixture, and it is a health hazard. Use coated horns, stainless fixtures, and local extraction. POM releases formaldehyde; halogenated flame retardants can release hydrogen halides. Ventilation is not optional.

What strength to expect

Joint efficiency — weld strength divided by bulk strength — is the number that matters for design. The literature and vendor data give a consistent picture:

For design allowables, take the measured short-term joint strength and apply a factor of 1.5–2.0 for static loads and 2.5–3.0 for fatigue or impact. A welded joint is a stress concentrator; it is not a homogeneous section. If the application is safety-critical, qualify the joint with the same test that will be used in production, at the same temperature and humidity, and add a burst or proof test.

11. Strength, testing, and failure modes

Ultrasonic welding has no radiographic inspection that reveals a cold weld. The joint is a thin bondline inside an opaque part; the only reliable quality signal is a destructive test correlated with a process signature.

Destructive tests. Tensile pull to failure on a purpose-built test bar or a production part; single-lap shear; peel for flexible assemblies; burst pressure for fluid housings; leak rate for sealed enclosures. Record the failure mode, not just the load: cohesive failure through the weld is good, interfacial failure at the bondline is suspect, and parent-material failure is excellent — it means the weld is stronger than the part.

Non-destructive tests. Visual inspection for flash, part marking, and collapse; dimensional check of the post-weld height; weight check (flash loss); pressure decay or helium leak for hermeticity; cross-section microscopy on a sampled part (20–50 µm polish, measure bondline and void fraction); DSC to check crystallinity in the heat-affected zone of semi-crystalline parts.

Process signatures. Collapse, peak power, energy, and weld time. These are the production control variables. A shift in peak power at constant collapse usually means a change in material, moisture, or horn coupling. A shift in collapse at constant energy usually means a change in the director geometry or the fixture.

Symptom · Likely cause · Diagnostic · Fix

Cold weld, low pull strength · Amplitude too low, energy too low, contaminated surface, warped joint · Measure horn amplitude; inspect fracture surface · Increase amplitude/energy; clean; flatten land

Flash ring, trimming required · Too much energy, director too large, no flash trap, force too high · Check collapse and energy; measure director · Reduce energy; add flash trap; resize director

Cracks at corners · Sharp corner, thin wall, excessive clamp force, part flex · Inspect crack origin; FEA the corner · Radius ≥0.5 mm; support the wall; reduce force

Horn marks on cosmetic surface · Horn face roughness, high amplitude, insufficient contact area · Inspect horn face; measure amplitude · Polish face; protective film; reduce amplitude

Inconsistent welds · Time mode, molded tolerance, mold release variation · Plot collapse; check part dimensions · Switch to distance mode; control molding; ban release agent

Voids/bubbles in bondline · Moisture, entrapped air, too-fast collapse · Cross-section; dry resin trial · Dry resin; vent grooves; slow downspeed

Weak semi-crystalline weld · Energy director instead of shear joint · Fracture surface shows unbonded land · Redesign to shear joint; increase amplitude

Insert pull-out · Hole too large, insufficient knurl, thin boss · Measure insert/hole interference; torque test · Resize hole; use knurled insert; thicken boss

Horn cracking · Fatigue, mistuning, sharp fillet, wrong material · Crack location; frequency sweep · Larger fillet; retune; titanium horn

Electronics damaged · Ultrasonic energy coupled into components · Functional test after weld · Weld before assembly; isolate; lower amplitude

Odor/fumes · PVC, POM, halogenated FR · Material audit · Extraction; coated horn; substitute resin

The leak test deserves its own note because it is the most common functional requirement. Pressure decay is the production workhorse: pressurize the assembly, isolate it, and measure the pressure rise or fall over 10–60 s. The leak rate follows

A 100 mL enclosure that loses 2 mbar in 30 s has Q = 0.1 \times 2/30 = 6.7\times10^{-3} mbar·L/s, or 6.7\times10^{-4} Pa·m³/s. That is a large leak by hermetic standards — bubble testing resolves about 10^{-3} mbar·L/s, and helium mass spectrometry reaches 10^{-6} mbar·L/s. For IP67/IP68 enclosures, correlate the decay threshold to an immersion test on the same geometry, then hold the threshold in production.

12. Economics: ₹1.40 per part and the 18,800-piece break-even

Ultrasonic welding is capital-intensive and labor-light. The machine is the dominant fixed cost; the per-part cost falls sharply with volume.

Indicative 2026 Indian prices:

Item · Price · Notes

20 kHz, 2,000 W analog welder · ₹1.35–1.80 lakh · Bench-top, pneumatic, time/energy modes

20 kHz, 2,000 W digital welder · ₹1.80–2.30 lakh · LCD, amplitude control, distance mode

15 kHz, 2,600 W welder · ₹1.50–2.00 lakh · Larger parts, higher amplitude

15 kHz, 3,200 W welder · ₹2.50–3.20 lakh · Heavy-duty, large horns

Servo-driven / imported stack · ₹6–15 lakh · Precision, distance control, medical

Aluminum horn (custom) · ₹20,000–45,000 · 7075-T6, 2–4 week lead time

Titanium horn (custom) · ₹60,000–1,50,000 · Ti-6Al-4V, abrasive/glass-filled

Fixture / anvil · ₹15,000–50,000 · Hardened insert, part-specific

Horn re-tuning / repair · ₹8,000–25,000 · After wear or crack

Cost model for a 60 × 40 mm enclosure at 100,000 parts/year:

At 100,000 parts, capital cost is ₹0.75/part. At 50,000 it is ₹1.50; at 25,000 it is ₹3.01.

Labor: a 9.9 s cycle (5 s load, 0.4 s weld, 0.5 s hold, 3 s unload, 1 s transfer) gives 364 parts/hour at 100% and 309/hour at 85% OEE. At ₹200/hour fully loaded, labor is ₹0.65/part. Energy is 800 W × 0.4 s = 320 J = 0.089 Wh, or ₹0.0007/part at ₹8/kWh. Energy is irrelevant.

Annual volume · Capital/part · Labor/part · Total/part

25,000 · ₹3.01 · ₹0.65 · ₹3.66

50,000 · ₹1.50 · ₹0.65 · ₹2.15

100,000 · ₹0.75 · ₹0.65 · ₹1.40

Compare with four M3 screws: ₹0.80 each = ₹3.20, plus 25 s of assembly at ₹200/hour = ₹1.39, for ₹4.59/part — and that excludes the molded bosses, the screwdriver, and the risk of a stripped thread. The ultrasonic saving is ₹3.19/part at 100,000/year.

Break-even:

That is the strategic number. Below about 18,800 units, with no machine in-house, screws or adhesive are cheaper. Between 18,800 and 75,200, the decision depends on whether the machine can be shared across products. Above 75,200, ultrasonic welding wins decisively — and the gap widens with labor cost, because the process removes 25 s of manual assembly per part.

Adhesive bonding is the other alternative. A two-part epoxy or acrylic costs ₹1.50–3.00/part, plus surface prep, dispensing, fixture time, and cure (5 min to 24 h). It is not hermetic, it is temperature-limited, and it introduces VOC and shelf-life issues. For a sealed enclosure at volume, ultrasonic is usually cheaper and faster; for a one-off or a heat-sensitive assembly, adhesive may be the only option.

Process · Capital · Cycle · Best fit · Limitation

Ultrasonic · ₹1.4–2.3 lakh · 3–10 s · Small/medium thermoplastics, volume · Joint must be designed for it

Hot plate · ₹3–8 lakh · 10–30 s · Large, simple joints, semi-crystalline · Slow; heat damage; wide HAZ

Vibration · ₹8–25 lakh · 5–15 s · Large parts, automotive · Capital; particulate flash

Laser · ₹25–60 lakh · 2–10 s · Medical, precise, clear/black · Capex; pigment/absorber dependent

Adhesive · Low · Minutes–hours · Prototypes, dissimilar materials · Cure time; not hermetic

The hidden costs are horn wear and reject rate. An aluminum horn on unfilled resin may last 50,000–200,000 cycles; titanium on glass-filled resin can last 500,000–1,000,000. A worn horn loses amplitude and produces cold welds, so horn wear should be on the PM schedule, not discovered by a leak test. A 2% reject rate at ₹20 part cost is ₹4,00,000/year at 100,000 units — more than the machine. The process control loop (collapse SPC, periodic pull tests, horn amplitude checks) is what protects that margin.

13. DFM rules and RFQ checklist

The rules that most reliably separate a weldable part from an unquotable one:

  1. Design the joint at the part-design stage. Cutting a joint into an existing mold costs 5–10× more than designing it in, and often requires a new tool.
  2. Keep walls 1.5–3 mm at the joint. Thinner walls flex and heat; thicker walls need multiple directors and longer cycles.
  3. Use a 90° director for amorphous resins, 60° for semi-crystalline and PC. Base 20–25% of wall, height 10%.
  4. Keep the apex sharp. A rounded apex is a cold weld waiting to happen.
  5. Add a collapse stop. A molded ledge that bottoms out at 0.8–1.1× the director height turns an open-loop press into a distance-controlled process.
  6. Add a flash trap if the joint is visible. The trap volume should be at least the director melt volume.
  7. Align with pins or a step, not with the director. The director is a heat concentrator, not a locating feature.
  8. Support the joint from below. The fixture must be a rigid anvil under the joint, not a soft nest under the middle of the part.
  9. Keep the joint within 6 mm of the horn face. If it must be deeper, use a rigid amorphous resin and verify amplitude at the joint.
  10. Radius every corner. Minimum 0.5 mm; sharp notches crack under vibration.
  11. Dry hygroscopic resin and ban silicone mold release. These two changes fix most "mystery" weak welds.
  12. Provide a flat, parallel horn contact surface. Flatness 0.05 mm, roughness Ra 1.6; the horn must not rock.
  13. Weld before installing sensitive electronics, or isolate the components with compliant mounts. Ultrasonic energy does not respect a PCB.
  14. Specify the material grade fully: polymer, maker, GF content, FR package, colorant, and moisture at molding. "Nylon" is not a specification.
  15. Qualify with pull and leak tests, then control collapse with SPC. Visual inspection cannot see a cold weld.

An RFQ for an ultrasonically welded assembly should state:

The numbers to take away

For the molding side of the same part — wall sections, draft, gate placement, and the tolerances that decide whether the joint lands flat — see the injection molding guide. The design-for-assembly guide covers the broader decision between welding, snap-fits, and fasteners, the snap-fit guide covers the compliant features that often replace a weld, and the IP-rated enclosure guide covers the leak-testing and ingress requirements that a welded housing must meet. For threaded inserts in the same housings, the heat-set insert guide covers the thermal installation route; ultrasonic inserts use the same boss geometry but a completely different heating mechanism.

All heating rates, penetration depths, energies, and costs above are computed from the stated assumptions. Loss modulus, acoustic speed, and weld strength vary with grade, temperature, moisture, and frequency; vendor joint-design rules are starting points, not guarantees. The numbers here are design-stage estimates and should be replaced with weld trials, tensile and leak tests, and a measured horn amplitude before a joint is released to production.

More FabFlow blog posts