Common Mechanical Prototype Failures

A failure is useful only when you can say what it is evidence of. A crack in a printed polymer is weak evidence about a moulded boss in another resin. The same crack, in the production resin, at the production torque, is evidence about that joint. This article sorts common mechanical prototype failures by what they can and cannot tell you. Every figure is hypothetical teaching maths, not a BrahmWorks price, yield, timeline or test result. Turning a failure into a disposition before a repeat build is part of How to Take a Hardware Prototype to Production.

Printed and machined prototypes are often altered until they work: a filed edge, a selected screw, a dab of slack. The unit then passes, and the alteration is forgotten. The next unit, unfiled, is called a regression. It is the design you actually have.

Write the adjustment or delete it. If a latch closes only after a knife takes off a corner, that corner is still in the CAD, and a tool will repeat the CAD. A printed fit can be tight from a ridge or loose because a thin wall sagged. Neither reading transfers to a moulded wall with draft and shrink. Measure the failed feature and compare it with the drawing before you move the nominal out of irritation.

A cantilever snap is a spring. On a tough printed resin it may survive a demonstration and then creep, whiten, or break in a stiffer production grade. The demo did not measure strain. It measured one person's thumb. If you cannot estimate the strain, or at least bound it, you do not know whether the snap is a joint or a consumable.

Bosses fail when they must locate, clamp and resist a cable side load at once, on a short engagement. The screw strips, the boss splits, or the head crushes the plastic. Extra length without a wall tie only makes a taller lever. Thread-forming screws need a hole, a wall thickness and a torque window that match the resin. "Tighten until firm" is how one good unit and one cracked unit share a revision.

A fastener is not a structural member you forgot to design. If the product holds its shape only because screws preload two flexible housings, the housings are moving. Give the load a path through walls, ribs and interlocks. Let the screws clamp that path.

Hinges and sliding latches fail by wear, not by the first operation. Ten slow cycles in the office say almost nothing about daily use, dust, or a year of parked load. A printed pin can look smooth and still be abrasive. Call a short demonstration a demonstration. If life matters, cycle the duty you claim on the material you will ship, and record the count as that test only.

Seals fail for geometric reasons a soft prototype gasket conceals. A warped lid, a parting line across the bead, an ejector pin on the land, or screws that clamp the corners and leave the middle free. Hand pressure during a splash is not the screw pattern. Put flatness and clamp pitch in the failure note, or the next gasket will be blamed for a lid that was never flat.

Layered printing is weak between layers. A boss that splits on a print plane may be orientation, process, or a real notch. You cannot tell from the crack alone. Reprint the same file in a new orientation, or machine one unit, before you thicken the CAD. If the crack moves with the layers, the thicker model would have been fiction.

Heat, sunlight and chemicals are the other absences. A latch that passes in an air-conditioned room has not seen a hot cabin or a disinfectant wipe. A sticky elastomer or a clip that no longer returns can be a material failure with innocent geometry. An exposure you did not run is an open item, not a pass.

The teaching product is a handheld meter. Five doors, printed in a tough polymer, latched and unlatched in the lab by the person who designed the latch. The door was then tooled in a stiffer production resin family. Early moulded doors either needed two thumbs or cracked at the root on the way off. Nothing in that story is a measured BrahmWorks trial. It is a pattern.

The prints had answered whether a person could open that shape in that polymer, slowly, a few times. They had not answered strain in the stiffer grade, creep after a week latched shut, or draft moving the hook. One tight print had also been eased with a blade, and the model was not updated. The tool repeated the unfiled hook. The unanswered questions had been in the prototype the whole time.

Hypothetical beam arithmetic for teaching. Not a material test, a safety case, or a BrahmWorks result.

Treat the snap arm as a cantilever. Assume length L = 12 mm from root to the point of the hook, depth h = 1.6 mm in the bending direction, and a deflection y = 1.2 mm to pass the catch. For a simple rectangular cantilever, strain at the root is approximately

strain = (3/2) × h × y / L²

= 1.5 × 1.6 × 1.2 / 144

= 2.88 / 144

= 0.02, or 2%.

The formula ignores the fillet, the rate and the temperature. A fillet lowers the peak. A sharp root raises it. Two percent may be comfortable for one grade and absurd for another. The point of writing it down is to stop arguing from the demo. If you will not look up an allowable strain for the grade you mean to mould, the 2% is a question, not a pass. Changing y from 1.2 mm to 0.8 mm, everything else assumed equal, gives 1.5 × 1.6 × 0.8 / 144 = 0.013, or 1.3%. That is a design lever you can see. It is still not a fatigue life.

• The failed feature is measured, not only photographed.
• The material, the process and the revision are on the note.
• Any hand filing or selected part is written, or the unit is discarded as evidence.
• You can say which question this failure answers, and which it does not.
• A printed crack is checked in a second orientation before the CAD grows thicker.
• Snap deflection and root thickness are on the drawing, not only in a render.
• Boss engagement, hole size and torque are a set. "Firm" is not a torque.
• Loads have a path that is not the fasteners alone.
• Cycles, chemicals, heat and humidity you care about are listed as run or not run.
• The next build changes one accused cause, so the result can be read.

Not by itself. Print layers, a brittle prototype resin, or a ridge on the surface can create a crack the moulding will not repeat. The reverse is also true: a tough print can hide a sharp root that a production grade will open. Use the print to find the location and the load case. Confirm it in the production material, or a named stand-in, before you call the joint safe.

No. Thickness can move a hinge that then never opens, create a sink opposite a show face, or raise the force until users pry with a tool and break something else. Ask whether the section was under-designed, over-deflected, or simply the wrong material. A thicker snap at the same deflection can raise strain. Check the arithmetic before you add plastic.

No. Ten cycles show that the feature can operate ten times, in that material, at that speed, by that person. Fatigue, wear and creep need the duty you intend to claim, including time under load with the latch parked shut. A short pass is allowed to be encouraging. It is not allowed to be copied into a reliability statement.

The revision, the material, the process, what was done to the unit by hand, the measurement of the feature, the load or the cycle count, and the question you think it answers. A photo with no scale and the sentence "latch broke" cannot be designed against. A note that says the hook opened 1.2 mm and cracked at a sharp root, in printed resin, on cycle 6, can.

Bring the broken unit, the revision it was built from, and the question you believe it answered. BrahmWorks can help separate a prototype-process artefact from a joint the production material will still have to survive.

Review the next stage with BrahmWorks