One event that is guaranteed to ruin the day of any plastics engineer is a phone call from the field telling of a product failure. Initially there is the hope that the damage is the result of a previous encounter with a forklift truck or some other extraordinary occurrence. As more calls come in, it becomes apparent that, unfortunately, that is not the case. Clearly there is a fault in design or manufacturing which may result in a dreaded product recall.

Assuming someone has saved the failed parts and not subjected them to further damage (like cramming them into too small a box), the next step is an examination of the problem. With luck, the fault is obvious. Severely degraded material, a sharp inside corner with a high stress concentration factor and an open weld line (also known as a “knit line”) are examples of faults that are usually visible to the naked eye. However, other deteriorating characteristics may be devilishly difficult to discover. Furthermore, multiple phenomena may be occurring that result in confusing symptoms. For example, a weld line will always be weaker than the surrounding material in a molded part. But, it might not be a problem in a given application until some additional stress is applied.

One source of this type of stress is the extraordinary effect of heat on plastics. Basic physics describes the phenomena whereby objects expand with heat and contract upon cooling. An object not constrained in some way behaves exactly that way. In manufacturing, stress results from the contraction of the polymer as it cools from its processing temperature, particularly as it encounters something that is shrinking at a slower rate. In molding, this is referred to as “post-molding shrinkage”, however the problem is not confined to the molding processes. Following processing, the product is subjected to heating and cooling in use as well as in shipping, storage, decorating and joining. The resulting expansion and contraction causes further stresses.

This type of problem almost never occurs in a single part product such as an ice scraper for a windshield. It is more common, but still rare, with multi-part assemblies that are made of the same material. It becomes a significant issue with multi-part assemblies using different polymers, particularly if some are glass-filled, as plastics have widely different thermal properties. Some polyethylenes have a Coefficient of Linear Thermal Expansion (CLTE) more than sixteen times that of certain glass-filled thermoset polyesters and a post-molding shrinkage rate one hundred times as great as the same polyesters. Multi-part assemblies involving metal components are equally vulnerable to this type of problem as most metals have a CLTE near that of glass-filled polyester. These are the extremes, but they clearly demonstrate the potential for problems.

Effects of Differentials in Post-Molding Shrinkage
In the processing operation, polymers are elevated in temperature until they are soft enough to be formed or flow in a mold and they shrink as they cool. The problem is most pronounced in molding. The molten plastic begins to cool the instant it enters the mold, which is usually liquid cooled. When the mold opens and the cavity side is removed, the part is exposed to the air and the cooling process accelerates. As the part cools, it stiffens. When it has done so enough to withstand the stress of ejection, it is removed from the mold.

The most common type of failure resulting from post-molding shrinkage is a cracked boss around a molded-in threaded metal insert. The amount of expansion exhibited by the metal insert when the molten plastic surrounds it is infinitesimal. The plastic shrinks around the insert as it cools, and the resulting stress can cause the boss to crack, normally at its weld line. Those that fail immediately are readily discarded before they reach the market. Those that do not fail until some additional application or environmental stress is applied can cause a product recall. Presuming the part has been properly molded, the solution lies in a thicker boss wall thickness. This type of problem is not confined to threaded metal inserts, it can occur with any type of insert, metal or plastic. Overmolding, two-color molding or multi-part molding are three names for the same process. A part is molded, then a second or even a third part is molded over or around it. In effect, a plastic insert is molded as part of the process (instead of independently created). In most applications, it is the outer part that is most vulnerable to failure. Thin sections of the part that cools more quickly in the mold behaves very much like molded-in plastic inserts. Consequently, molders try to gate into the thickest section.

Post-molding shrinkages can range from 0.0005 in/in to 0.050 in/in depending on the plastic. The thermoset plastics (those that undergo a chemical change and cannot be remelted) are usually listed with very low post-molding shrinkage rates. However, that is because they are normally used with glass or mineral fillers which tend to reduce shrinkage. The thermoplastics (those that can be remelted) are essentially divided into two groups, amorphous (having random coil molecular chains) and semi-crystalline (with a very regular, lattice molecular structure). In general, the amorphous thermoplastics have much lower post-molding shrinkages than the semi-crystallines. The range for the amorphous group (without fillers) is 0.001 to 0.008 in/in whereas the semi-crystallines range from 0.002 to 0.050 in/in, however most of them are above 0.010 in/in. High shrinkage rates also make it much more difficult for the processor to hold tight tolerances.

When comparing the post-molding shrinkages, the engineer should bear in mind the limitations of the information provided on most data sheets. The usual test for mold shrinkage is ASTM D955, a test for initial shrinkage. That test is performed 48 hours after molding. While it is true that 75 to 95 percent of the shrinkage takes place in the first two hours after molding, shrinkage continues after that. In most cases, nearly all of it has taken place by the end of a week, however some semi-crystallines can take a year to reach equilibrium. Shrinkage rates will also vary with direction of flow from the gate; normally lower in the direction of flow than in the transverse direction for unfilled resins. Molding parameters also affect shrinkage and, within limitations, the molder has some control over shrinkage.

Finally, there is the matter of wall thickness. Shrinkage figures are typically based on sample plaques with wall thicknesses of 0.125 in. With high shrinking polymers in particular, significantly thinner walls may shrink less, and thicker walls more, than the published figure. Thicker sections also require more time to cool. Parts that have wall thicknesses that vary more than 25 percent are subject to external sink marks and high levels of molded-in stress.

Effects of Differences in Coefficient of Linear Thermal Expansion
Differences in the CLTE are of particular concern when joining parts of different materials. If the parts are welded, the welds will be stressed beyond the values inherent in the basic product design. If the joint is a mechanical one, the part with the greater expansion rate will expand more than the other part, thereby creating problems. If the part with the higher CLTE is the outer part, the two pieces can loosen and come apart at elevated temperatures or tighten and produce sufficient force to cause failure at lower temperatures. If the part with the lower CLTE is on the outside, the reverse could occur. When examined at room temperature, the phenomenon is no longer taking place, leaving the observer to wonder why the parts cracked or came apart. The greater the dimension, the more significant is the hazard. Press fits are particularly vulnerable to this type of problem.

There is an enormous range in coefficient of linear thermal expansion through the thirty-four major families of plastics. Most, however, are far higher than their metal counterparts. The metals range from 4.8 x 106 in/in/oF (titanium) to 13.4 x 106 in/in/oF (aluminum), with carbon steel around 8.3 x 106 in/in/oF. The plastics range is from 7.5 x 106 in/in/oF (glass-filled polyester) to 122 x 106 in/in/oF (polyethylene). Therefore, if we take a 10 in long ABS (CLTE = 61 x 106 in/in/oF) part fastened to a stainless steel (CLTE = 6.6 x 106 in/in/oF) frame at room temperature of 72o F and shipped in a freight car that reached 140o F, the plastic part will expand 0.037 in more than metal one [(.000061-.0000066) x 10 x (140-72) = 0.037 in]. If it were only a metal insert of one inch diameter, the differential would still be 0.004 in. It had better not be a press fit or a part with a weak weld line.

Engineers can also use differences in CLTE to their advantage, particularly in the assembly of mechanical joints. One of the secrets of efficient assembly is to heat the outer part so that it fits easily over the inner part. On cooling, the outer part grips onto the inner part.

Again, a word about the information provided on material data sheets. The coefficient of linear does vary with the direction of flow. Furthermore, it can vary through the temperature range. Best to contract the material supplier to obtain more detailed information for critical applications.

Conclusions
Engineers who want to avoid having that dreaded product failure report land on their desk must take into account differentials in post-molding shrinkage and the Coefficient of Linear Thermal Expansion in order to avoid expensive product recalls. Where possible, materials should be selected to minimize the thermal effects. If that is not possible, expansion joints should be employed or adequate strength be designed into the part to account for these differentials.

Jordan Rotheiser is president of Rotheiser Design Inc. and the author of the critically acclaimed, best selling “Joining of Plastics - Handbook for Designers and Engineers”, the product design chapter of the “Modern Plastics Handbook” and the plastics chapter of McGraw Hill’s recently published “Handbook of Materials for Product Design”.