How Should Silicone Shrinkage and Tolerance Be Defined Before Tooling?

Select dimensions from function for silicone shrinkage and tolerance

In a silicone tooling review, the first dimensional question is not how much the compound shrinks; it is which finished surfaces locate the part when it performs its intended job. A seal lip may be located by a housing shoulder, a lid by a container rim, and a flexible grip by the user's hand rather than by the geometric center of a CAD model. Ask the buyer for the mating components, assembly direction, load, temperature range and failure mode before placing datums. Where a feature does not affect fit, function, safety, appearance or assembly, a restrictive limit may buy no useful risk reduction.

We would mark dimensions by consequence before assigning numerical limits; the first tier contains interfaces that can cause leakage, detachment, interference, poor sealing or unsafe assembly; the second covers features that affect handling, appearance or pack fit; the remainder may follow a general drawing rule. This classification is a purchasing decision as well as an engineering one because tighter limits can require different tooling, more sampling, fixtures, sorting and scrap allowance. A tolerance table is most useful when every important row names the function and the method that will establish acceptance.

A flexible product can change shape when installed, so the drawing should distinguish free-state dimensions from dimensions achieved in the assembly; a free-state opening may need to be smaller than a mating neck, but that difference is a design hypothesis until actual samples are fitted and tested under the intended conditions. Document how much insertion force, retention or compression matters and which part is allowed to move. Without this boundary, one supplier may quote a relaxed unstressed dimension while another assumes an installed seal requirement, making the prices and technical offers incomparable.

Separate drawing targets from measurement rules

A dimension without a measurement rule is only half a specification for silicone; the part may flatten under a caliper, stretch when removed from a fixture or relax after molding and trimming. Before tooling, define whether the part is measured unsupported, on a low-force fixture or installed on a master component; state the orientation, measurement points and instrument contact method. If the feature is a thin sealing lip, an optical image or functional go/no-go fixture may be more repeatable than squeezing it between metal jaws. The selected method must be feasible for incoming, trial and production checks, not only for a specialist laboratory.

The measurement instruction should also state when the reading is taken; freshly molded silicone, post-cured material, parts conditioned overnight and parts stored compressed in a package are not automatically equivalent states. ISO 23529 provides general procedures for rubber test-piece preparation and conditioning, but its public scope specifically does not create a whole-product measurement method for every geometry. We therefore agree a finished-part timing and support rule with the buyer, then use it consistently across trial rounds. If timing must change, the old and new data should be compared under both methods before a trend is interpreted as tool movement.

Write down the resolution and uncertainty appropriate to the decision rather than assuming more decimal places mean better control; a soft wall measured by two operators can vary because of hand pressure, edge selection or part temperature, even when both instruments are calibrated. Run a small method study with the same parts, operators and fixtures before negotiating an aggressive tolerance. Retain photographs of the supported part and the probe position with the drawing revision. When repeatability is inadequate, improve the fixture or choose a functional test; do not tighten the product drawing to compensate for a measurement system that cannot resolve the proposed limit.

State the shrinkage hypothesis

A useful first mold trial is designed to locate variation, not merely to produce attractive samples; label parts by cavity, tool revision, material lot, pigment condition, cure setting and molding sequence; preserve untrimmed examples where flash or release evidence matters. Record the time after molding and any post-cure or conditioning before measuring. The buyer should ask for enough pieces to see whether a feature is centered, drifting or different between cavities, with the quantity linked to the number of active cavities and the risk of the dimension. One selected sample is a demonstration of possibility, not evidence of a stable distribution.

Start measurements with the agreed fixture and operator method, then compare them with a functional test on the named mating component; a dimension may be outside a provisional target while the assembly still works, or inside the target while a lip folds during installation. These findings should trigger a documented review of the target, method and product design rather than an informal decision to ignore the number. Keep raw values, sample identities, photographs and observation notes together so a later tool correction can be compared against the same baseline. Do not average away a cavity-specific defect by reporting only the pooled mean.

When the first trial reveals an offset, investigate the cause before changing steel; ask whether the apparent movement follows cavity position, material batch, pigment, cure condition, measurement timing or a change in the mating part. Repeat selected checks using retained samples and a second measurement route when the method is suspect. If a correction is justified, document the feature, direction, expected effect, affected cavities and other dimensions that might move. A buyer-approved correction request should identify what the next trial must prove; otherwise successive rounds can become a sequence of cosmetic changes without closing the functional risk.

Design a measurable mold trial

Tolerance approval needs evidence from a production-representative state; confirm the intended compound, color route, tool revision, active cavities, trimming and post-cure, then sample across the range of conditions that will be used commercially. The acceptance decision should review both individual critical readings and how the part behaves on real mating components. A preliminary T0 sample may establish tool function, while later trials may be needed for dimension, appearance, cavity balance and repeatability. The labels T0, T1 and T2 are project milestones, not a substitute for a written definition of what each round is authorized to approve.

For each critical characteristic, the release matrix should record the drawing target, measured range, method, sample source, functional result, approved limit and disposition; if the original limit proves unrealistic, change it only with the buyer's design and quality owners after checking the consequence of a wider range. Some dimensions may need a tighter local limit, while noncritical exterior features can use a more practical class. A conditional deviation must have a quantity or time boundary, an owner and a closure action; it must not silently become the standard for all repeat orders. Retained physical samples should support, not replace, the numerical record.

Where an inspection plan uses sampling, define the lot, cavity rotation, defect severity and escalation response explicitly; iSO 2859-1 can offer attribute-sampling context, but dimensional capability and critical fit risks may require targeted checks or more complete inspection. The buyer should know whether first-off, in-process, final or shipment checks are intended to detect a shift and how an out-of-limit reading stops or contains output. A single laboratory report on a material plaque does not validate finished-part geometry. Approval should cover the exact product revision and production route whose results were observed, with unresolved risks visible to procurement before the purchase order is released.

Approve limits from representative evidence

Control change and repeat orders