Demolding Deep-Draw Silicone Parts Without Compressed Air Assist Damage

A deep-draw silicone part (boot, bellows, cup, sheath) leaves the core like a suction cup. The usual factory answer is compressed air. The usual defect is a rolled sealing lip, a stretch mark, a contamination event, or a part that inverts and never sits right again.

Air assist is a tool, not a personality. You can demold deep draws without damaging air by cutting the vacuum lock, spreading the strip load, and using air only as a measured pulse through the core. A hose in an operator’s hand is how medical and optical parts pick up oil, rust, and a crease that ISO 3302-1 will not call out but the customer will.

Process split: LSR in steel is the high-volume problem. HCR compression boots are often pulled by hand over a mandrel. RTV glove molds invert on purpose. Do not copy the glove-mold peel onto a 16-cavity LIM cell.

Why the part will not drop

Three locks, often stacked:

• Vacuum lock. Mirror ID, little draft, closed end. Atmospheric pressure holds the cup on the core.
• Friction. 0° walls, texture (VDI or MT grain as micro-barbs), self-bonding grade tack.
• Mechanical undercut. A sealing barb or bellows convolution that must stretch.

Soft high-elongation LSR can come off all three if a human rolls it. Robots and cycle clocks cannot roll. Damage starts when the release energy is dumped into a small lip: air that inflates the closed end and then escapes under the seal in a jet, or a gripper that pulls one ear.

If you ignore the lock type, the defect is “we added more air” until the lip blossoms.

Break the lock in steel, not in the hose

DFM that reduces the need for violent air:

• Draft on the core, even 0.5-1° on a deep draw, is a lot of opening. Zero draft is an elongation bet (see the draft article in this cluster).
• Finish: a controlled matte or very light blast can break vacuum without printing a cosmetic grain. Mirror looks premium and holds like a sucker. Texture deep enough to be VDI 30+ on a 0.6 mm wall is a perforation and a barb.
• Stripper ring or plate that pushes on a designed shoulder, not on the sealing lip. The part peels from the open end. This is the most robot-friendly no-damage method.
• Collapsible or split cores when convolutions are honest undercuts.
• A tiny vent or hole in the product, if allowed, so a perfect cup cannot form. Many boots cannot have it. Ask.

Gate vestige on the closed end becomes a tear start when air balloons that end. Put the gate where stretch is low, or use a valve gate and a vestige limit.

HCR compression deep draws add a second damage mode: the operator who “pops” the part with a screwdriver at the closed end. That nick is a tear start on the next fold or the next pressure test. If the mandrel will not release with taper and a straight pull, fix the mandrel. A pry bar is not a demold spec.

If you still need air, make it a circuit

Core air that does not damage:

• Internal passages that open *behind* the closed end so the part balloons uniformly, then the lip walks off a lead-in, not a square shoulder.
• Timed, filtered, dry air as a process parameter (pressure, duration, start relative to open). Shop air with water and rust is a dirty blast.
• Pressure low enough that the part does not invert. Inversion is a geometry plus pressure problem. A thin bellows will turn inside out at a pressure a thick boot ignores. Find the pressure on the real part.
• No aim at the lip from outside. External air knives fold edges and throw flash onto the next nest.
• Medical / food: air is a contamination control problem (ISO 13485, food-contact discipline next to FDA 21 CFR 177.2600 grades). Filter, dedicated circuit, no operator blow-off as a standard.

Sometimes the honest answer is more cure so the lip has tear strength, or less polish, not more bar. Green rubber plus air is a balloon animal.

HCR mandrels: a slight taper, a good finish, and a pull along the axis. Air between mandrel and part can help; blasting the OD to “pop” it will stretch a thin wall out of spec.

RTV: peel the jacket. Compressed air between jacket and part is how you split a 10A mold. Use compressed air between jacket and *master* only if the jacket can take it.

Cell and inspection

Robots should grip a designed rim after the stripper or core air has broken the lock. Vacuum cups on a deep-draw OD will collapse the draw. Cameras catch inversion; they do not un-invert a medical lip you will ship.

First-article checks for air damage: lip roll, witness rings, localized gloss from stretch, contamination, and dimensional neck-down on the wall (a cousin of core shift, but from strip). Put them in the visual standard. If you only CMM the OD at the mouth, you will miss a creased bellows.

RFQ notes for deep-draw release

Deep-draw geometry, draft on the core, finish intent, whether a stripper shoulder is allowed, undercuts, grade/durometer, robotic versus hand, and air policy: none, core circuit only, or not on this part. If the lip is a seal, say that air marks are a fail. If you ignore the lock, T1 will work with a technician and fail with a robot and a hose.

FAQ

Can I demold a deep-draw LSR boot with no air at all?

Often yes if you have draft, a finish that is not a vacuum lock, and a stripper on a shoulder. Zero-draft mirror cores usually need some assist. Design the steel so air is optional. Do not design a sucker and then ban air in a note.

What air pressure is safe for a sealing lip?

There is no universal bar number. Thin, soft lips invert at pressures thick boots ignore. Instrument a core circuit, raise pressure until the part frees without roll, and lock that as a parameter. An external hose has no safe number because the jet is not controlled.

Why does air assist leave a cloudy ring on clear LSR?

Stretch: the rubber yielded locally where the balloon peeled. Finish, draft, and a slower peel (stripper) reduce it; more pressure makes a sharper ring. Clear medical and optical parts should treat that ring as a defect in the visual standard.

Is operator blow-off acceptable on food-grade or medical deep draws?

As a documented, filtered process, core air can be validated. As a person with a gun, it is a contamination and a geometry lottery. ISO 13485 and food-contact programs should not rely on random shop air at the lip.

Will more draft make robotic deep-draw ejection slower?

Draft usually makes it faster and kinder: less air, less stretch, less miss-pick. The CAD volume change is the cost. A 1° core draft on a deep boot is often cheaper than a cell that fights inversion every twentieth shot.