Blog
Moulding defects: narrowing down the cause without stopping production
No material is defect-proof. A defect is the output of a system — material, its condition, machine, tool, process, geometry — not a property of the pellet. The same grade on two machines produces different scrap rates.
When scrap appears, the practical question is not "which grade is better" but a different one: which node of the system more probably holds the cause, and how to narrow that area without stopping production for a week.
"Narrow" is the operative word. A single trial does not establish a cause.
The nodes where the cause sits
Material — class, viscosity, filler content, stabilisation, additive package, lot-to-lot consistency.
Condition of the material — residual moisture, regrind fraction, storage. For polyamides and polyesters this is a separate node: the same pellet at different moisture levels behaves as two different materials.
Machine — screw and non-return valve condition, dosing accuracy, actual versus set temperatures, melt residence time, ability to hold the set injection speed.
Tool — venting, cooling circuit balance, surface condition, gating system, wear.
Process — temperatures, speed profile, switchover point, holding pressure and time, back pressure, cycle.
Geometry — wall thickness and transitions, gate positions, ribs, inserts, weld line locations.
Disputes arise because each party sees its own node: the supplier sees material, the process engineer sees settings, the toolmaker sees the mould. Some failures belong to no single node at all but to an interaction: a grade with a narrower processing window runs stably on one machine and not on another.
Cross-trials: what they actually deliver
A cross-trial does not prove a cause in one run. It changes the priority of hypotheses.
If the defect repeatably follows the lot under identical preparation and identical actual process parameters — investigate the lot.
If it follows the machine or the tool — investigate that node.
If the failure appears only in a particular combination — the cause is an interaction between material and process, and neither side alone is "at fault".
"Repeatably" is not a formality here. A single lot swap establishes nothing: something else usually changes with the lot — moisture after a different drying cycle, actual mould temperature, non-return valve behaviour. The working scheme is A–B–A or A–B–B–A, with actual parameters recorded rather than set values.
The outcome of a cross-trial is stated carefully: the hypothesis became more probable, not proven.
Diagnostic table
The order in the second column reflects checking practice, not a rigid hierarchy.
| Defect | Competing hypotheses | How to narrow it down |
|---|---|---|
| Silver streaks | Moisture; volatiles from additives or regrind; thermal or shear damage; entrapped air; reinforcement showing through | Shape, location and direction of the streaks separate the mechanisms. Start with residual moisture measured before the machine, then residence time and actual melt temperature |
| Sink marks | Insufficient holding pressure level; premature gate sealing; no melt cushion; leaking non-return valve; pressure losses in nozzle and runner; massive section | Step up holding time while monitoring part weight or cavity pressure: a plateau indicates gate sealing. Check cushion, valve and pressure losses separately |
| Warpage | Differential shrinkage along and across flow; fibre orientation; unbalanced cooling; uneven wall thickness; part temperature at ejection; later moisture uptake | Measure actual surface temperatures and circuit balance; check part temperature at ejection; analyse flow direction. A deliberate temperature difference between mould halves is tested as a separate hypothesis — for glass-filled grades it rarely removes the cause |
| Weak weld line | Front temperature at the meeting point; venting in the weld zone; speed profile and switchover point; holding pressure; gate position; wall thickness; fibre orientation | Raise melt and mould temperature; check venting specifically in the weld zone; gate relocation as a design hypothesis |
| Local burns (diesel effect) | Trapped gas at the end of fill, in a rib or where flow fronts meet; blocked or missing venting; high speed at the end of fill | Clean the vents; reduce speed over the final section |
| Burnt streaks (thermal degradation) | Excessive actual melt temperature; long residence time; high shear in narrow sections; dead zones; excessive screw speed or back pressure | Measure actual melt temperature; shorten residence time; inspect dead zones in barrel and hot runner |
| Jetting | Melt jetting into free cavity volume; excessive initial injection speed or a poor profile; gate type, cross-section and position; abrupt section change | Reduce speed over the initial fill section; change flow direction, gate type or position so that the melt meets a wall immediately |
| Black specks | Degradation in barrel or hot runner dead zones; contaminated feedstock; screw and barrel wear | Purge the barrel with purging compound; strip the hot runner; examine particles under a microscope |
| Brittle parts | Polyamide hydrolysis from processing wet pellets; excessive regrind; overheating; weld line in the loaded area | Viscosity number before and after processing; actual regrind fraction; testing after conditioning |
| Short shot | Grade viscosity against flow length; actual temperatures; speed profile; venting; gate cross-section | A series of short shots with stepwise dose increase |
| Dimensions drift over weeks | Moisture uptake in hygroscopic polymers; post-crystallisation; residual stress relaxation; creep under assembly load; thermal ageing; storage conditions | Conformity is assessed in the condition stated on the drawing. If no condition is stated, that is a specification issue, not a moulding issue |
Typical diagnostic errors
Silver streaking reduced to one mechanism. Silver streaking is a description of appearance, not an established cause. Moisture streaks, shear damage and reinforcement showing through look similar and are treated differently. They are separated by shape, location and accompanying signs — and the work starts with moisture, because it is the cheapest check.
Warpage treated by extending the cycle. Longer cooling reduces the symptom and leaves the cause untouched if it lies in shrinkage anisotropy. For a glass-filled grade, fibre orientation stays as it was.
Weld lines attacked with holding pressure alone. Pressure helps in part. In the weld zone fibre orientation often becomes unfavourable relative to the main load direction, so reinforcement efficiency drops — and pressure cannot fully compensate for that.
Brittleness blamed on "a bad lot". For polyamide the first hypothesis is different: hydrolysis from processing wet pellets. The chain is cut in the barrel, and a brittle part comes out of material that looks perfectly normal.
Burns treated by venting without checking temperature. If these are burnt streaks from degradation, cleaning the vents changes nothing while the cause keeps working.
One change per run, or a designed experiment
These are two different instruments and should not be mixed.
To test one clear hypothesis, change one factor per run. The method is slow, but the result is unambiguous. It does not reveal interactions.
If an interaction between several factors is suspected, use a designed experiment in which the changes follow a matrix rather than intuition. A typical factor set: melt temperature, mould temperature, speed profile, holding pressure.
Before planning the experiment you need a stable machine and dryer, measurable actual temperatures and speeds, a defined response, replication and safe factor ranges. Without those, the matrix produces elegant numbers with no meaning.
When the cause really is the material
These are three different situations, and confusing them is costly, because the actions differ.
A non-conforming lot. Contamination, viscosity deviation, wrong filler level, excess volatiles, poor dispersion, foreign pellets, unstable colour, additive package deviation, regrind of unagreed origin. Action: incoming inspection and a claim against the specification.
Material capability mismatched to the process. Processing window narrower than the actual process spread, drying requirements beyond the site's capability, viscosity unsuited to the geometry, stable filling unachievable on the available machine. Action: a different grade or a process change — whichever is cheaper.
An error in the original selection or design. Insufficient heat resistance, unsuitable stabilisation, excessive anisotropy, insufficient hydrolysis resistance, absence of the required impact or fatigue behaviour. Action: revisit material choice together with the geometry of the assembly.
Changing the grade is justified when one of these three has been confirmed. Otherwise it may shift the sensitivity of the system without removing the root cause.
Boundaries of responsibility
The supplier's responsibility is defined by the agreed specification, lot documentation, the quality agreement and the product change notification procedure. The extent of those obligations is not universal: not every supplier issues a CoA for every lot or notifies every change of production site.
The processor is responsible for observing the approved storage, drying and process conditions, and for the condition of machine and tool.
Unregulated boundaries are fixed in writing before series production. What most often remains undocumented is the definition of what counts as a change — and until it exists, each side is certain that nothing changed.
What to do when scrap appears
- Record the defect: photograph, location on the part, percentage, the shift it started on
- Check what changed over the past week: lot, operator, masterbatch, regrind fraction, tool repair
- Measure residual moisture before the machine
- Record actual process parameters, not set values
- Run a cross-trial on an A–B–A scheme with repeats
- Change one factor per run and record the result; move to a designed experiment if an interaction is suspected
- Before concluding anything about the material, have the lot input data, not only observations of the part
The role of Material Wizard
We help gather the input data, compare lot values against the specification, rank the hypotheses and build a verification plan that narrows the area of the cause.
Where a non-conforming lot or a capability mismatch is confirmed, we propose a grade with a different processing window and a plan for verifying it. Where the data point to the process or the tool, we say so directly, even when a material change would have been the more profitable answer.
Pinning down the node remotely is not always possible: often the correct output is a ranked list of hypotheses and a plan that separates them.
Sources and limits of applicability
The "symptom — probable causes" relations above draw on troubleshooting guides from polyamide and polyester producers and on established thermoplastics processing practice. Specific settings, tolerances and threshold values depend on grade and geometry and are taken from the material documentation.
This text covers injection moulding. For extrusion, part of the causes and the order of checks differ.
The article is not a failure analysis protocol. A list of probable causes does not become a diagnosis without measurement: in safety-relevant cases the conclusion follows from verification results, not from a table.