Silicone Quality Control: From Compound Certificate to Final Inspection

What a compound certificate actually tells you

A silicone compound arrives with a certificate of analysis listing hardness, tensile strength, elongation at break, tear strength, compression set and specific gravity. Those are the properties of a cured slab, tested to a standard, from a batch of compound that may be several tonnes. The certificate is a statement about the batch, not about your part.

The gap between the two is the whole problem. A slab cured in a laboratory press at an optimum temperature for an optimum time is the best case the material can produce. Your part is cured in a production mould, at a temperature measured somewhere other than the cavity, for a time set by the cycle rather than by the chemistry. The certificate sets the ceiling, and the process decides how far below it you land.

The parameters worth reading first are hardness and specific gravity. Hardness drifts with filler loading and with cure state, and a batch at the top of the tolerance plus a press running slightly hot gives parts at the top of the tolerance on both counts. Specific gravity catches filler variation, which shows up as a change in stiffness and in cost per part at the same time.

Compression set deserves a closer look than it usually gets. It is the property most sensitive to cure state and post-cure, and it is the one most often quoted in a customer specification for a gasket or a seal. A certificate value of 20 percent at 22 hours and 175 C is achievable in a lab slab and unachievable in a part that has not been post-cured, and no amount of final inspection will reveal the difference.

Incoming inspection worth doing

Retesting every property on every batch is a waste of money. The certificate is produced by a supplier who tests to the same standard, and repeating the full suite doubles the cost of incoming material to confirm what you already know. The useful incoming checks are the ones that detect the failure modes a certificate cannot cover: batch to batch variation, contamination and ageing.

Hardness is cheap to measure and the fastest indicator of variation. A durometer reading on a cured button from each batch, compared against the previous batch and against the certificate, catches a filler change before it reaches the press. The measurement takes a minute and the alternative is a week of parts that are all slightly too stiff.

Specific gravity by water displacement is the second cheap check and it catches filler substitution, which is the most common form of supplier drift because it is invisible in every other property. A compound that has moved from 1.15 to 1.19 has had three percent more filler added, and it will mould differently, feel different and cost less.

The third check is a cure or scorch test on a sample from each batch. A moving die rheometer curve, or a simple press cure of a button at production temperature, shows whether the batch cures at the rate the previous one did. Compound that has aged on the shelf cures more slowly, and the first sign in production is a part that is under-cured at the edges.

Table 1 — An incoming plan that costs minutes, not days

Controlling the process, not just the part

The chart on the right puts compound variation at 30 percent of silicone defects and curing at 24. Together they are more than half, and both are process variables rather than part variables. That is the argument for controlling the press rather than the output: by the time a part can be measured, the decision that made it has already been taken.

Mould temperature is the first control point and it is almost always measured in the wrong place. A probe in the platen reads the platen. The cavity surface runs cooler, and it runs cooler at the edges than at the centre. A contact thermometer on the cavity surface at the start of a shift, and again after an hour, tells you whether the tool is at the temperature the cure time assumes.

Cure time is the second. Silicone cure is a reaction that runs to completion and then stops, so a part that is cured for slightly longer than necessary is not harmed, while a part cured for slightly less is permanently deficient. Setting the cure time with a 20 percent margin over the rheometer curve, rather than at the minimum that releases cleanly, is the cheapest quality decision in the plant.

The third control point is the press operator's release check. A part that sticks, tears at the gate or shows a dull patch is telling you something about cure state, and the information is only useful if it is recorded. A shift log with three columns, cavity temperature, cure time and release condition, will explain most of the variation that a final inspection report can only describe.

What final inspection can and cannot catch

Final inspection catches what is visible and measurable: dimensions, flash, tears, discolouration, missing features and contamination on the surface. It cannot catch under-cure, wrong compound, wrong post-cure or a compression set that will fail in service. Those are latent defects, and no sampling plan finds them because the part looks correct.

The chart on the left shows why this matters commercially. A defect found at incoming compound costs one unit of effort. The same defect found at final inspection costs ten, and at the customer it costs forty five. The whole argument for process control is that it moves detection upstream, where the correction is cheap and the information is still useful.

Dimensional inspection of silicone is a special case because the material is compliant. A caliper closes on a soft part and deforms it, so the reading depends on how hard the operator squeezes. The standard fix is a controlled-force gauge, or a go and no-go gauge that measures the part in the same way every time. Either is better than a caliper reading with no defined force.

Visual inspection needs a written standard or it is not inspection. A boundary sample for each defect type, kept at the station and replaced when it degrades, converts a subjective judgement into a comparison. Without one, two inspectors will pass different parts and the customer will see the difference before you do.

Writing a control plan that fits the part

A control plan is a list of the characteristics that matter, the method used to check each one, the frequency and the reaction when a check fails. The mistake is copying a plan from a similar part. The characteristics that matter are a function of what the part does, and a gasket and a keypad need different plans even when they are the same compound in the same press.

Start from the failure modes rather than from the drawing. A seal fails by compression set, by tearing at assembly or by taking a permanent distortion. A keypad fails by a dead key, by a worn legend or by a tear at the web. Each failure mode points at a characteristic and a control point, and the resulting plan is short and defensible.

Frequency should follow risk and cost. A dimension that is set by the tool and does not move can be checked at setup and at the end of the run. A dimension that depends on cure shrinkage and compound variation needs a check every few hours. Putting both on the same frequency wastes time on the stable one and misses the drifting one.

Table 2 — Five stages, five records, one traceable lot

When a batch has to be quarantined

A quarantine decision is a commercial decision dressed as a technical one. The technical question is whether the parts meet the specification. The commercial question is what it costs to be wrong in each direction, and the two answers are often different. A batch of medical components and a batch of promotional keyrings do not warrant the same risk appetite.

The trigger for quarantine should be written down before it happens. A failed incoming check, a cavity temperature excursion, a cure timer that was reset mid-run, a compression set result outside the limit: each of these has a defined response, and the response should not depend on who is on shift. Without a written trigger, batches get shipped because nobody wants to make the call.

Once quarantined, the batch needs a disposition path that is faster than the customer's complaint. Retest a sample against the full specification, including compression set if the part is a seal, and record the result. If the batch passes on retest, release it with the retest attached. If it fails, scrap it or offer it with a concession, and do that before the shipment date rather than after.

The last step is the one that gets skipped: closing the loop back to the process. A quarantined batch is evidence that a control point was missing or was not working, and the control plan should be amended before the next run. A plant that quarantines the same failure mode twice has a control plan that is not being read.