Our part is cracking in the field: do we go thicker, change polymer, or move to a glass filled grade?

Start at the fracture surface, not at the datasheet. A part that cracks in the field almost always fails in one place: a sharp internal corner, the base of a rib, a weld line behind a through-hole, or a screw boss. If that is where the cause sits, a different polymer will not fix it and a thicker wall only half fixes it. So establish where and how it breaks first, and only then choose between a geometry change, a thicker wall, a different polymer or a glass filled grade. Those four routes differ sharply in tool work, lead time and the risk of creating a new problem.

Where do you start when parts are already failing at customers?

With the broken parts themselves. Ask for ten to twenty back and lay them out side by side, because one fracture is an anecdote and twenty are a pattern. Then answer four questions: where does it break, what does the fracture surface look like, when in the service life, and do the cases share a production batch.

That fourth question is the one most often skipped and the fastest to answer. If only parts from two consecutive weeks fail, the design is fine and the cause sits in the process or the raw material.

| What you see on the fracture surface | What it probably means | First check | | --- | --- | --- | | Break exactly in an internal corner | Stress concentration from too tight a radius | Measure the radius on the part, not on the drawing | | Straight break across the wall, away from any corner | Weld line where two flow fronts met | Find the gate and the through-holes, run a fill simulation | | Whitened, stretched zone next to the break | Material taken well past yield: overload or impact | Reconstruct the load case, not the material | | Smooth, dull fracture with no visible deformation, after months of use | Fatigue or brittle failure | Count the load cycles, check temperature and UV | | Only cases from one batch | Process or raw material: moisture, regrind, holding pressure | Ask for the process data and the batch number | | Hairline cracks around a boss or an insert | Environmental stress cracking from torque or cleaning agents | Check tightening torque and the cleaning agents in use |

The table in one sentence: if it always breaks in the same place, a different material is the most expensive way to hide a tight radius.

Is it the design, the material or the process?

Three signals separate them. One location across all batches points at the design. Varying locations under normal load point at material chosen too light, the rarest outcome. The same location but only in part of the series points at the process.

The process side is cheap to investigate, so it goes first. Polyamide that has not been dried to specification moulds perfectly well and gives a part that turns out brittle later. Melt temperature too high, or residence time too long, degrades the polymer. Too little holding pressure leaves voids in thick sections, exactly where a rib meets a wall, and a regrind percentage that quietly crept up does the same. All of that is checkable without touching the tool.

On the design side, look hard at weld lines: invisible on the drawing, visible on the fracture surface. Research on glass fibre reinforced PA6 measures a loss of more than 50% of tensile strength at a weld line, while the unfilled material in the same study barely weakens. A broader review puts the usual range at 40 to 80% of the strength of the homogeneous material. If your crack line sits on a weld line, the question is not which material is stronger, but whether you can move the gate so the weld line runs out of the loaded area.

How much does a thicker wall buy you, and what does it cost?

More than you expect, and it costs more than you expect too. In a flat section loaded in bending, the second moment of area goes with the cube of wall thickness and the bending stress with the square. A wall 20% thicker is therefore about 1.7 times as stiff, and the stress in that section drops by roughly 30%. That is arithmetic from the geometry, not a material claim.

In the tool this is often the cheapest change, because a thicker wall means removing steel from the cavity or core, usually in the existing block. The reverse does not work: a thinner wall, or a rib that has to go, means adding steel, so welding and remachining, or a new insert.

The bill arrives in cycle time. Cooling time scales with the square of the thickest wall, and research on injection moulding puts cooling at 60 to 80% of the total cycle. A worked example on invented assumptions: a 30 second cycle with 20 seconds of cooling, and a wall going from 2.0 to 2.4 mm. Cooling becomes about 29 seconds and the cycle about 39 seconds, nearly 30% longer, while that wall takes 20% more resin. Over a programme of years, that is a structural cost.

Which is why local stiffening usually beats thickening the whole wall. Common design guidelines put rib thickness at 40 to 60% of the nominal wall the rib sits on, and rib height at no more than about three times that same nominal wall. Note what those ratios are measured against: the wall, not the rib. A rib too thick for its wall cools slower than its surroundings and produces exactly the sink and the void you wanted to avoid. A more generous internal radius is also steel removal, and often the smallest change with the biggest effect on a corner crack.

Can our existing tool run a different polymer?

Mechanically, usually. Dimensionally, almost never without a modification. The obstacle is shrinkage. Published PA66 datasheets give roughly 1.5% shrinkage in flow direction and 1.8% across it for the unfilled grade, and roughly 0.45% in flow and about 1.1% across for a 30% glass fibre grade. On a 100 mm dimension in flow direction that is more than a millimetre of difference in the finished part. The tool was cut for one of the two and will deliver a part that no longer fits with the other.

The second difference is anisotropy. In the unfilled grade, shrinkage along and across flow sit close together; in the glass filled grade they differ by a factor of two to three. That gap warps flat parts and lids that have to close. The processing window changes too: melt and mould temperature, drying, venting and often the gate size. A change within the same family, to a different viscosity grade or an impact modified version, is the least invasive route. A move to another polymer family is in practice a fresh sampling round with a dimensional report, even if the tool was never touched.

When does a glass filled grade help, and when does it make things worse?

Glass fibre helps with stiffness and creep, and works against you on impact and at weld lines. The figures from published PA66 datasheets show both: dry tensile modulus goes from about 3,000 to about 9,000 MPa, while strain at break falls from about 30% to about 2.5%. The part gets stiffer and deforms far less before it breaks.

That makes the choice straightforward. If your part deflects too far, or slowly sags under sustained load, glass fibre is the logical direction. If it breaks from a drop, a knock or a rough assembly step, fibre will probably make it worse, because you are removing the very ability to deform that absorbs the energy.

Three things come attached that are not on the datasheet. Glass fibre is abrasive: gates, vents and unhardened steel wear faster, so the guaranteed shot count of your tool changes. The weld line becomes a genuine weak spot. And regrinding breaks the fibres shorter, so the allowable regrind percentage is lower than for the same polymer without fibre.

Which route do you choose, and what does each cost in time and risk?

Choose the cheapest change that actually touches the cause you established, not the change that gets made most often. That means ruling out the process first, then the geometry at the failure point, then the wall, and only then the material.

| Route | Typical change in the tool | Lead time | Biggest risk | When this is the right route | | --- | --- | --- | --- | --- | | Process adjustment only | None | Shortest | Fixes nothing if the cause is in the design | Failures correlate with a batch or production date | | Larger internal radius or an extra rib | Removing steel in the existing block | Short: one machining round plus a sampling | Sink on the visible face, draft | It always breaks at the same corner or rib base | | Move the gate | New gate position, sometimes an insert | Medium | The weld line moves elsewhere, it does not disappear | The crack sits on a weld line | | Thicker wall across the part | Steel removal on more faces, sometimes across the parting line | Medium | Longer cycle, sink, more weight | Failures spread across a flat section loaded in bending | | Different polymer, same family | Often none, but a new sampling round | Driven by resin delivery | Dimensional drift from different shrinkage | The limit is creep, temperature or chemical exposure | | Glass filled grade | New inserts or a new tool, wear resistant steel | Longest | More brittle, weld lines, warpage | Stiffness or creep is the limit and the load is not impact |

One thing matters more than the choice itself: test before you change steel. A printed or machined version of the modified part shows whether a larger radius or an extra rib touches the problem, at a fraction of a tooling round. It says nothing about the strength of a moulded part, because the material orientation differs, but it does say whether you are looking in the right place.

Then run the arithmetic on the parts still to come, not the ones already out. On invented figures: 2% failures on a remaining run of 50,000 parts is 1,000 returns, each carrying handling, replacement and shipping. That number decides whether a tool modification pays.

Have the modified design reviewed before it goes into steel. A DFM review on the injection moulded design catches the radius, the rib ratios and the gate position, and a manufacturability review across the whole chain also looks at assembly, because a part preloaded on the bench breaks later in the field. What a thicker wall does to the visible surface is covered in engineering decisions that impact looks, and what an extra rib does to ejection in optimising your ejection strategy. If it ends in a second tool, comparing mould quotes over the programme makes that call on cost per good part.

Production Heroes takes products from design to series production, from feasibility and cost price to sourcing and supply chain, and supplies moulds and tooling to manufacturers across Europe, from the Netherlands and its own presence in China.

Sources: the shrinkage, modulus and strain values for PA66 come from the public product datasheets of a major European polyamide producer, for an unfilled grade (2026) and a 30% glass fibre grade (2024), measured to ISO 527 and ISO 294-4; other grades and colour variants differ, so always check against the datasheet for the grade you actually run. The loss of tensile strength at a weld line in glass fibre reinforced PA6 comes from an ICCM23 paper (2023), and the 40 to 80% range from an Altair technical article on weld line positions (2021). The rib ratios of 40 to 60% of the nominal wall and a rib height of up to about three times the nominal wall come from public design guidelines published by moulders, among them Fictiv and RJC Mold (2023 to 2025). The share of cooling in the cycle (60 to 80%) comes from RWTH Aachen research on cooling time in injection moulding; that cooling time scales with the square of wall thickness is described in Polímeros (2015) and in technical material from RJG. The worked examples in the text and in the FAQ use invented assumptions and are not a quotation.