The Materials Analyst, Part 8: Detecting molded-in stress

When a part fails, a good analyst will initially focus attention on simple matters like degradation of the material and appropriate composition. Previous articles in this series have been devoted to techniques used to investigate these issues. But sometimes these simpler techniques reveal no difference between a good part and one that failed. In this case, methods that can focus on the structure of the material become useful. One such method is called thermomechanical analysis--TMA, for those who spend a lot of time in the lab. For engineers, TMA is easier to appreciate than techniques like DSC and TGA (covered in previous articles) because it measures a physical property--the coefficient of expansion.

How TMA Works

Figure 1 shows a TMA device. It consists of a quartz cylinder called a stage, with an opening in the side for inserting samples and a quartz probe that moves vertically. Samples are placed on the quartz stage and the probe is brought to rest on the top side of the sample. For best results the sample should have straight edges on top and bottom--so this is not a technique for fragments or pellets. Carefully machined pieces of a molded part are used. A small force is applied to the probe to ensure good contact with the sample, and then a furnace is placed around the sample so that it can be heated.

As the material expands or contracts, the dimensional change is measured very precisely by the analyzer and can be plotted as a function of temperature. The slope of the line is the coefficient of expansion. Figure 2 shows an example of this for a piece of a polycarbonate tensile bar.

In a perfect world all TMA results would look like the polycarbonate in Figure 2. But molded parts can exhibit some interesting behavior in a TMA that can reveal a lot about the way in which they were molded. One problem that TMA is particularly good at detecting is molded-in stress. In a transparent part, molded-in stress can be observed visually as birefringence--colored patterns that appear when polarized light shines through the part. But this technique is not useful for opaque parts. In these cases, TMA can be used.

Field Failure for ABS
Our customer's problem was a pressure vessel cap. It was a large, thick-walled part molded from 10 percent glass-fiber-reinforced ABS. It was shaped like a hemisphere with a maximum diameter of nearly 12 inches and had a threaded hole in the center. The part was filled through a large sprue that emptied into a generous diaphragm gate. The diaphragm entered the part through the threaded section and was removed to produce the hole. The exterior of the part was then filament-wound with an epoxy glass-fiber system to improve the burst strength to the required 800 psi. Some of the parts were cracking in the threaded area while others were not.

We were given several parts, some of which had been wound while others were in the as-molded state. We started with a check for composition. There was no difference. DSC tests verified that the ABS was free of any contamination, and TGA tests confirmed a glass content of 9.8 to 10.2 percent in both good and bad parts. Melt-flow-rate tests revealed no signs of degradation.

Then we noticed a pattern. None of the as-molded parts we had been given were cracked. A call to the customer confirmed that cracking was only observed after the filament winding process. Anyone who has seen filament winding knows that it involves draping a resin-reinforcement combination over a substrate. Usually the resin is a thermoset material, and is kept soft and pliable with solvents that evaporate after the system has been applied and allowed to dry. An obvious conclusion was that the ABS was being chemically attacked by the solvents. But why just some parts? And why did the cracks only appear in one small area of the part when the solvents were in contact with the entire part?

Differentiating Good and Bad Parts
The answers came from the TMA. Small samples were machined from the gate area of a good and a bad part and placed in the TMA. Heating the material caused the expected expansion as the sample temperature increased.

However, as Figure 3 shows, the bad part expanded more rapidly. And when the samples reached the glass transition temperature, just more than 100C, a dramatic difference was observed. Amorphous resins like ABS become soft above the Tg. If unreinforced they actually collapse under the weight of the probe like the polycarbonate in Figure 2.

In this case, the glass fiber in the material caused the good part to simply level off and slowly compress as the material temperature increased. However, the bad part began to expand even more quickly above Tg. By the time the sample reached 115C, the bad sample had expanded twice as much as the good one and it still wasn't done. Another expansion was under way when the test was stopped at 150C. Figure 4 shows the coefficient of expansion vs. temperature, and provides an even more dramatic comparison.

What would cause such a difference in behavior and how did it relate to our problem? The sudden expansion in the bad part was a manifestation of stress relief. The good part had been molded properly and behaved predictably. However, the bad part had been overpacked. The molecular overcrowding that results can be thought of as a bunch of compressed springs waiting for a chance to relax. Below Tg, the material is too rigid and immobile to permit any stress relaxation.

As the material begins to warm up, though, the increased molecular motion starts to unleash these constrained polymer chains. At the glass transition temperature, the mobility increases dramatically and the polymer literally explodes at the molecular level.

Why Failure Occurred
Because the gate area sees the highest pressure in the mold cavity, the problems associated with molded-in stress due to overpacking are usually seen in this area. Now imagine this overstressed area surrounded by the rest of the part. The solvents in the epoxy resin increase the mobility of the ABS in a manner similar to the increase in temperature. However, the surrounding material in the molded part constrains the highly stressed region and causes a stress buildup that results in the cracks.

This was an excellent example of environmental stress cracking (ESC), a phenomenon where the combined application of stress and a chemical produce a failure where neither the stress or the chemical alone would adversely affect the material. In this case, the stress came from within the part.

Further from the gate, much smaller differences are seen between a good part and a bad part. Figure 5, shows TMA results for samples taken from near the end of flow in a good and bad part. While the bad part still expands more, it responds appropriately once it achieves Tg. Figure 6, shows the effects of annealing both the good and the bad parts. After heating both parts above the Tg and cooling them to room temperature, samples taken from near the gate behave identically and are almost stress free.

Because the part had to be held to very close tolerances, it was not always possible for the molder to limit the packing pressure as a means of controlling molded-in stress. Annealing of the parts prior to filament winding provided a means of preventing part failure without handcuffing the molder to an unrealistically tight processing window.

Environmental stress cracking is the number one cause of product field failures. Often the problems appear after months in the field. But very often, the stress component of the failure doesn't come from the application, it comes from within the part. Molded-in stress can't be readily seen in an opaque resin, but it can be detected by TMA.