Designing Undercuts for Robotic Ejection in High-Volume LSR Molding Lines

A technician can roll a 20 Shore A boot off a barb. A robot cannot “just peel it.”

High-volume LSR (liquid silicone rubber, platinum, hot steel, usually a valve-gated cold runner) lives or dies on a repeatable exit. Undercuts that were free in a 2-cavity hand cell become cycle-time and scrap generators the day you add a sprue picker, a side-entry gripper, and a 16-cavity cold deck. The rubber’s elongation is real. It is not a license to leave a geometry that only a human wrist can solve. Design the undercut as an ejection specification: direction, stretch, gripper, and what happens on the shot that is 3% stickier than the mean.

HCR compression with an operator pulling parts is a different rule book. RTV glove molds that invert are a third. Do not mix them on the same drawing.

Name the undercut by how it must leave

Not all undercuts are equal in a cell:

• Local lip / sealing barb that must stretch over a core diameter. The part stays on the core; something has to balloon it off (air, stripper, or a gripper that pulls along the axis).
• Side hole or window that needs a lifter, a side core, or a collapsing action because the rubber cannot stretch that far without tearing a thin wall.
• Internal groove on a bore that a collapsible core or a split core was born for.
• Cosmetic wrap that a human peels and a vacuum cup will fold into a crease.

Write the strip vector on the drawing. Robots move in lines and rotations you programmed. If the honest demold is a rolling peel around an axis that is not the press open, you need a tool action or a gripper that can do that roll every shot, including the one that flashed.

If you ignore the vector, the defect is a robot that marks the part, a core that never sheds, or an operator standing next to a “lights-out” cell.

What LSR will actually stretch

Elongation and tear (think ASTM D412 and ASTM D624 on the *grade*) set the budget. Soft 10-30 Shore A with high tear can come off surprising barbs. 60-80 Shore A with a knife-edge lip cannot. Fillers, self-bonding grades (tackier, more likely to stay on steel), and a mirror cavity (vacuum lock) all spend that budget before the undercut does.

Useful shop checks, not folklore constants:

• How much hoop stretch is required (change in circumference), and is that a small fraction of the grade’s elongation at break?
• Is the lip thick enough to take the stretch without a tear that then grows (cut-growth thinking, ASTM D813 family)?
• Is there a lead-in so the lip walks off, or a square shoulder that must snap?
• Will a textured core act as a thousand extra undercuts?

Zero draft plus an undercut is two problems. Give free walls draft so the action is not fighting friction and a barb at once.

Self-bonding LSR on a metal or plastic insert is another trap: the robot may be pulling against a chemical bond you wanted. Mask the core, or accept that the part stays on the insert and the insert is the handle.

Tool actions a robot can live with

Prefer mechanisms that present a known face at a known time:

• Air assist through the core. Works on boots and bellows if the part can balloon without inverting into a wrinkle. Air that blasts a medical lip across a dirty floor is not a process. Aim the circuit, time it, and do not use it as a substitute for draft.
• Stripper plate or stripper ring that pushes the part off a core along the open. Robots love this. Thin lips can roll under a stripper if the speed is wrong.
• Lifters and side cores for holes that should not stretch. More steel, more maintenance, more flash risk at those shut-offs. LSR will leak through about 0.005 mm; sliding shut-offs on a hot tool are a flash discipline problem.
• Collapsible cores for internal grooves. Expensive, justified when stretch would tear.
• Rotate / unscrew for threads. LSR threads can sometimes strip; a robot that unscrews is more honest than hoping.

Valve-gated cold runners help because there is no sprue for the robot to grab as a handle. You must then grip the *part*. Design a non-cosmetic pad, a thick rim, or a runner nub you deliberately kept. Vacuum cups on 10 Shore A will collapse the feature you just molded.

Ejector pins in LSR are easy to flash around and easy to print. If you need pins, keep them off sealing lands and accept a witness, or use a sleeve.

Cell timing is part of DFM

Cure is not binary. An undercut that strips clean at 25 s may tear at 18 s because the lip is still green, or stick at 40 s because it shrunk onto the core. The robot program cannot “feel” that the way a technician does unless you instrument it (vacuum sensor on the cup, gripper stroke, camera).

Build the part so the nominal shot ejects, then check:

• First shots of the shift (colder tool, stickier).
• After a weekend (different scorch in the deck, different tack).
• Self-bonding grades and clear grades (often tackier).
• Flash: a 0.1 mm flash flake on a barb turns a stretch-off into a tear-off. Flashless lands are an automation spec, not only a cosmetic spec.

ISO 3302-1 tolerances on a lip OD are meaningless if the lip is nicked every twentieth shot. Put demold yield in the first-article plan, not only CMM.

RFQ notes for robotic undercut cells

Process: LSR injection, robotic demold. Send:

• Every undercut with strip direction and whether stretch is allowed.
• Durometer, grade family (self-bonding or not), and tear/elongation if you have them.
• Gripper concept: vacuum, ID grip, OD grip, sprue grip. Where marks are forbidden.
• Tool actions you will pay for (stripper, air, lifter, collapsible).
• Cycle target and whether the cell is truly unmanned.

If a face cannot take a gripper mark, say so in millimetres of allowed witness, not in “cosmetic.” A drawing that says “undercut OK, silicone stretches” is a hand-tool drawing. Rewrite it before you buy the robot.

FAQ

Can a robot peel a deep LSR undercut the way an operator does?

Usually not: operators roll and invert, while robots pull along programmed axes. If the honest exit is a peel around a barb, add air, a stripper, a lifter, or change the geometry. Counting on a six-axis arm to “figure it out” produces marks, tears, and a person standing next to the cell.

Does a softer durometer always make robotic ejection easier?

It makes stretch easier and gripping harder. 10-20 Shore A comes off a barb and then collapses in a vacuum cup. 60 Shore A grips well and tears at a sharp lip. Pick durometer for the product, then design the gripper and the undercut to that rubber, not the other way around.

Where should the robot grip if there is no sprue on a valve-gated part?

On a designed pad, rim, or thick wall you declared as a handling surface. Do not invent a grip on a sealing lip or a class-A face in the robot program. If every surface is cosmetic, you needed a stripper plate that leaves the part on a known nest.

Will air-blast ejection damage a medical sealing lip?

It can fold, invert, or contaminate the lip, and it can blow flash onto the next nest. Aimed core air with a timed pulse is a tool. A shop air wand is not. For ISO 13485 cells, treat air as a validated process step: filter, aim, and inspect the lip, or use a mechanical stripper instead.

Do self-bonding LSR grades stick to cores more in automated cells?

They can. The same adhesion promoters that bond to PC or metal make steel tackier if the core is not isolated. Mask or plate the core, keep it out of the bond chemistry, and expect first-article stickiness to be worse than a standard grade. Automation sees that as a missed pick, not as a “sticky shot.”