Engineering Answers / Prototyping & Validation
A prototype test is only worth running if a failure would change the tool. Most test plans are written to confirm a decision that has already been made, which is why they cost a month and change nothing.
Before steel, test the four things a tool cannot fix later: material performance under heat and compression, tear resistance at the thinnest section, sealing force at the actual geometry, and chemical resistance to whatever the part touches. Run them on parts from the route you will productionise, at the hardness and wall thickness you intend to ship. Production tolerance, flash and cycle time cannot be answered before a steel tool exists, so do not pretend a prototype answered them.
A steel tool fixes geometry. It does not fix a compound that takes a permanent set at 120 °C, a wall that tears at the sealing lip, or a material that swells in the cleaning solvent the customer uses.
That is the filter for a prototype test plan: if the answer could still be changed after the tool exists, the test can wait. If not, it belongs before steel.
| Question | Test | Route it can run on | Decision it informs |
|---|---|---|---|
| Does the seal recover after heat and compression? | Compression set, ASTM D395 Method B | Plaques from the production compound | Compound grade and cure system |
| Will the thin section tear on assembly? | Tear strength, ASTM D624, plus a physical assembly trial | Parts from the intended route | Wall thickness and base radius |
| Does it seal at the actual geometry? | Leak or sealing-force test on the real part | Printed mold or bridge tool | Lip geometry, interference |
| Does it survive the media it touches? | Immersion, volume swell and hardness change | Plaques | Compound family |
| Does it survive ageing? | Heat ageing, 70 h at service temperature plus margin | Plaques and parts | Compound grade, antioxidant package |
If the part is a seal or a gasket, compression set is the property that decides whether it works in month six, and it is the one buyers most often skip because hardness and tensile data look reassuring.
A compound can be exactly 50 Shore A, have good tensile strength and elongation, and still take a 35% permanent set after 22 hours at 175 °C. That means a seal designed with 20% interference loses most of its contact stress within weeks at temperature, and the leak appears long after the customer has accepted the first shipment.
The pass criterion should come from the application, not from a catalogue: for most static seals we work to under 20–25% set at the service condition, and tighter where the seal is expected to be reused or where the temperature cycles.
Most prototype test plans are written to confirm a decision. Three design choices make the difference.
Test at least three hardnesses. Not to pick a favourite, but to find the edge. If 30, 50 and 70 Shore A all seal, the design is robust and you can choose on feel. If only 50 works, the interference is marginal and the tool needs more of it.
Test two wall thicknesses at the thin section. The tear limit is a function of thickness, and the difference between 0.6 mm and 0.8 mm is often the difference between a 99% yield and a 92% yield on the stripping step.
Test above the service temperature. A part rated for 120 °C that is tested only at 120 °C has no margin. Run 150 °C and see what breaks first — that tells you what the failure mode in the field will be.
An assembly trial, not a bench test. A sealing lip that measures perfectly and tears when a technician pushes it onto a connector by hand is a design failure that no tensile test predicts. Ten minutes with the actual mating part and a stopwatch finds it before the tool does.
Three questions cannot be resolved before a production tool exists, and pretending otherwise is how a validation plan gets written twice.
If your customer's quality agreement demands evidence on those three, plan the validation on the production tool and treat prototyping as design verification, not process validation. Mixing the two is the most common reason a launch slips by a month: the report exists, and it does not satisfy the clause it was written for.
A test report that cannot be compared to the next one is a cost, not an asset. Insist on:
That last point is what makes a second test run cheap. If the first report has values, the second one is a comparison. If it only says "pass", the second one is a repeat.
Compression set, if the part is a seal. It measures how much of the seal's thickness is permanently lost after being held compressed at temperature, and it predicts the leak that appears months into service rather than at first article. Hardness and tensile strength both look fine on a compound that has poor compression set, which is why a part can pass every incoming inspection and still fail in the field.
For material properties, yes, provided the compound and the cure are the same as production. For geometry-dependent behaviour, only partly — a printed cavity is less rigid, so the part's dimensions differ slightly and a sealing force measured on it is indicative rather than final. What you cannot transfer at all is anything that depends on the production process: flash, cycle time, tool wear, and tolerance capability.
Enough to see variation, which in practice means at least five per condition and three conditions. Testing one part at nominal temperature confirms almost nothing — the interesting result is always the one at the temperature or the compression you did not plan for. A run of five at three hardnesses and two wall thicknesses costs a fraction of a tool change and settles the design.