Sponsored By

The Materials Analyst, Part 16: Following the path of a failed part

January 4, 1999

10 Min Read
Plastics Today logo in a gray background | Plastics Today

This series of articles is designed to help molders understand how a few analytical tools can help diagnose a part failure problem. Michael Sepe is our analyst and author. He is the technical director at Dickten & Masch Mfg., a molder of thermoset and thermoplastic materials in Nashotah, WI. He has provided analytical services to material suppliers, molders, and end users for the last 10 years. He can be reached at (414) 369-5555, Ext. 572.

Periodically, I have the opportunity to give seminars on the rules of proper material selection. With somewhere between 24,000 and 28,000 different grades available and with globalization bringing many unfamiliar trade names into our environment, picking the material that provides the best balance of cost and performance is not easy. This type of decision making is best done with some fundamental rules that can eliminate a lot of choices and narrow the field quickly.

One of these rules deals with how amorphous and semi-crystalline materials handle fatigue. In general, semi-crystalline materials are more durable in situations where fatigue is a factor in the application. While an amorphous material may offer other advantages such as impact resistance or short-term strength, these advantages can disappear over time with repeated stress cycles.

After sharing this piece of information with a group one day, I was approached at the break by a gentleman who excitedly related a story that proved my point. He had just switched an application from a 10 percent glass-reinforced polycarbonate to a 30 percent glass-reinforced nylon 6/6 and had solved a fatigue failure that had given his customer considerable trouble. While it is always gratifying to get support, the more we talked the more I became convinced the original material selection wasn't the problem in this particular case. Lot-to-lot inconsistency or a processing upset seemed to have been the real cause of the failures.

The application had been in the field for almost two years before trouble appeared. And the parts that were failing were from recent production. All of the old product was still functioning well. As we talked, I sensed he was also becoming unsure of the real cause of the problem. The nylon had solved the problem, but was it a necessary adjustment? I suggested if he wanted us to review good and bad parts molded in polycarbonate, he could send them to us, and we could determine if something had gone wrong.

Figure 1. DSC comparison of good and bad polycarbonate parts.

Running the Tests
Shortly after I returned to the office, a box containing good and bad parts arrived. The good parts looked great while the bad parts were full of blisters and splay. This seemed like an open-and-shut case of poor drying. However, we have learned over the years it is still wise to verify composition before running viscosity tests. Contamination or even a variation in glass content can cause problems with viscosity.

This was especially important because this material came from a small compounding house that had no direct control over the feedstock. We ran DSC and TGA tests first to check the material. Figure 1 compares the DSC results for a good and a bad part. The good part gives a typical result for polycarbonate, a clean glass transition at 148C with no hint of any other activity. The bad part produced a rather different result. First, the glass transition for the polycarbonate was down at 132C. A shift this extreme is a sure sign of severe degradation. This was followed by a series of melting events that looked like a mixture of polypropylene, nylon, and polyester. Performing melt viscosity tests would certainly have been a waste of time.

Figure 2. TGA comparison of good and bad polycarbonate parts.

Figure 2 compares the TGA results. Here again, problems abound. The good part produces a typical TGA scan. The material decomposes near 500C, produces the expected amount of carbon, and leaves a residue of 9.4 percent: the glass fiber. The bad part begins to degrade almost 100C earlier and produces less carbon. Both features are signs of contamination from less stable polymers. In addition, the glass content was too low at only 5.5 percent.

Obviously, the material that produced the bad parts was contaminated with a variety of impurities. But the source of the contamination was not clear. Fortunately, our client had the retains of a good and bad lot. Our DSC analysis of the pellets showed both the good and the bad material contained a small amount of polyethylene. This is a fairly common occurrence in compounded materials where a universal carrier is used in the concentrate that colors the material.

The pellets from which the bad parts were made also showed evidence of polypropylene but no nylon or polyester. Nevertheless, the glass transition temperature for the good material was 146C while the bad material had a Tg of 139C, a clear sign that the bad virgin material was also degraded. And while the TGA on the bad material gave the proper glass content of 9.9 percent, there was evidence of poor stability in the way the material decomposed. Even if all the contaminants had not come from the material supplier, there was enough wrong here to be very concerned.

Finding Fault
At this point, the relationship between our client and the material supplier was becoming adversarial. They were blaming each other for the contamination. As for the inconsistencies in the raw material, the compounder defended its position by sending in a copy of its data sheet and challenging anyone to find a property that wasn't in compliance with any of the specs.

Figure 3. Comparative stress-strain curve for good and bad polycarbonate.

The problem was that the data sheet only had four properties on it: tensile strength, flexural modulus, notched Izod, and heat deflection temperature. We took some of the retained material from both lots and produced test specimens for the various properties. In addition, we compared both lots of material to a lot of virgin 10 percent glass-reinforced polycarbonate from one of the major suppliers. We focused our standard tests on the tensile and impact properties.

One of the problems with data sheets is they often fail to capture the critical performance aspects of a material. Figure 3 shows a comparative stress-strain curve for the good and bad batch of polycarbonate. The tensile strength quoted on the data sheet is the peak of the stress-strain curve. If we report just the peak, the material that was failing has the same tensile strength as the good material. In addition, the modulus, which is the slope of the stress-strain curve in the linear region, is much higher for the bad material. Yet an examination of the total curve shows there are clearly several problems with the failed material.

The good batch of material produces a rounded stress-strain curve as it passes through the yield point and shows some extension past the yield point before failing. This leads us to expect some impact resistance. The bad product fails suddenly at less than two percent elongation. Flex tests showed the same disparity in performance. All very interesting, said the material supplier, but with an equivalent tensile strength and a higher modulus, where was the problem? The property sheet didn't certify an elongation to failure, so there was no responsibility. Clearly, we had to examine the impact properties.

The problem with the notched Izod test is that, once glass fiber is added to a compound, the results have as much to do with the glass as with the polymer. Fiber length and orientation are a factor, and the rigidity provided by the glass tends to reduce ductility and make everything look the same. Even unfilled polycarbonate, with a notched Izod of 15 to 18 ft-lb falls to 2 ft-lb with the addition of as little as 10 percent glass. Once the numbers are that low, it is difficult to document degradation. Falling dart tests are far more useful in tracking the declining performance of a material in question.

Figure 4. Schematic of impact test fixture.

We started with a variation of the Gardner method where the striking device is instrumented with a transducer and the force of the test is designed to cause each part to fail. Figure 4 shows a schematic of the instrument that was used. Figure 5 gives a comparison of the good and the bad lot of material. The solid curves that start at zero, rise to a maximum, and return to zero represent the load on the sample. The dotted lines are the total energy required to cause the sample to fail.

Here again, a picture is worth a thousand words. The good product shows a rounded load curve that spans almost eight milliseconds, an eternity for a high-velocity impact test. The sample deflected almost a full inch before failing, and the total energy to break was almost 19 ft-lb. This compared with 25 ft-lb for a virgin compound from one of the big boys.

The bad product shows a jagged load curve, a much lower load to failure, and a total energy of only 2.5 ft-lb. The total deflection on the bad samples was less than 0.2 inch, about 20 percent of that obtained for the good batch, and the time to failure was less than 1.4 milliseconds. The results indicating the differences in overall performance were clear to everyone. Unfortunately, there was no Gardner or falling dart specification on the data sheet. Only the Izod test was recognized by the material supplier.

Getting Results
At this point, the only way to prove the obvious difference in impact performance was to perform the Izod test. The data sheet called for a value of 2 ft-lb/in. This is fairly typical of the values quoted by the major suppliers as well. The good batch came in at 2.1. The bad batch fell just below 1 ft-lb/in. An 85 percent reduction in real toughness appeared as a 50 percent reduction in the notched Izod test. The material supplier was finally willing to concede that the polycarbonate in this batch was not up to par.

The viscosity tests never were performed; after all, the supplier did not quote a melt flow specification on its data sheet. If they had been run, they no doubt would have revealed the same thing we have seen repeatedly during this series of articles. The molecular weight of the brittle material was not up to the task posed by this product. The contamination certainly didn't help.

Figure 5. Falling dart instrumented impact test results for good and bad polycarbonate parts.

Given the stresses associated with this application, it was difficult to believe no melt flow limits had been placed on the material at the beginning of the product development cycle. But the small compounder was charging less for the material than the majors were, and, unfortunately, as is often the case, this is what drove the initial buying decision.

Ultimately, the molder had solved the problem by changing material families altogether and going with the nylon. Certainly, the nylon he chose can be expected to do the job. However, the costs incurred by retooling for different shrinkage and flow characteristics of the new material, in addition to the cost of dealing with the failed parts in the field, probably consumed the original savings in raw material cost.

Without a good program to screen the lot-to-lot consistency of the nylon, something that was not in place to begin with, there is no guarantee the problem won't resurface again at a later date.

Sign up for PlasticsToday newsletter

You May Also Like