Mould Shrinkage

The part is always smaller than the steel it came out of. Polymers are far more thermally expansive than tool steel, semi-crystalline ones give up extra volume as they crystallise, and no amount of packing pressure fully makes up for it. Mould shrinkage is that difference expressed as a fraction, and this page is about getting the fraction right — because a tool cut to the wrong shrinkage is expensive in a way very few other mistakes are.

The denominator is the mould dimension. That is the convention, and it matters because the other one exists. Referenced to the mould, $S = (L_{mould} - L_{part})/L_{mould}$, and the tool is cut as $L_{mould} = L_{part}/(1 - S)$. Referenced to the part it would be $(L_{mould} - L_{part})/L_{part}$, and a shrinkage quoted that way substituted into the cutting expression puts the cavity wrong by roughly $S^{2}$. On one 100 mm dimension at 2 % that is 0.04 mm — small enough to be invisible on a single feature and quite large enough to matter across a stack of them, or on a bore that has to take a bearing. The related old error is cutting $L_{part}\times(1+S)$ instead of dividing: it is the same magnitude, the same direction, and it is still in circulation. Ask which reference any quoted shrinkage figure uses, and say which yours does.

How shrinkage is measured is an authority question, and this site names the authorities without reproducing them. ISO 294-4 and ASTM D955 define the specimen, the moulding conditions, the conditioning atmosphere and — critically — when the measurement is taken. That last one is not a formality. Shrinkage keeps happening after the part leaves the machine: the 24-hour figure is not the 48-hour figure, and in a semi-crystalline resin post-mould shrinkage continues for days or weeks as crystallisation finishes, faster if the part is annealed or ever sees service heat. If you are quoting a shrinkage number to a customer, quote which standard produced it. If you are receiving one, ask.

Now the mistake that scraps tools: shrinkage is directional, and the direction that shrinks most depends on what is in the resin. In an unfilled semi-crystalline material, the flow direction usually shrinks more than the transverse direction, because the molecules were oriented along the flow and relax back. Put glass fibre in the same resin and the behaviour reverses. The fibres line up with the flow, they barely shrink at all themselves, and they hold that axis nearly rigid — so a fibre-filled part shrinks far less along the flow than across it, often by a factor of two or three. Applying one datasheet number to every dimension on a fibre-filled part does not give you a slightly wrong part; it gives you a warped one, because the differential shrinkage between axes is what warp is. And which axis is which is decided by where the melt enters and how it fills, which makes the gate position a dimensional decision as much as a filling one.

The process moves the number as much as the material does. Higher packing pressure and a longer hold time both reduce shrinkage, by forcing more material into the cavity before the gate freezes — which is why gate freeze-off time is worth measuring rather than guessing. A hotter mould increases shrinkage on a semi-crystalline resin, because slower cooling lets more crystallinity develop, but it also gives a more dimensionally stable part afterward for the same reason. A thicker wall increases shrinkage, because the core cools slowly and shrinks against a skin that has already solidified — the same mechanism that produces sink marks and voids. Two shops running the same grade to the same drawing will not get the same shrinkage. That is why a first-off is measured rather than assumed, and why the oldest rule in the tool room still holds: cut it steel-safe, sample it, and open the cavity to suit. Steel can always be taken away, and only welded back at a price.