The Materials Analyst, Part 111: Understanding root cause—an essential step in resolution

Problem solvers can avoid a time-consuming trip down the wrong path if they keep all five areas of plastic product design in mind.

There has been a lot of attention and training devoted to problem solving over the last 20 years in the world of business. It has become a sort of cottage industry, served in many cases by professional organizations that have little familiarity with the so-called nuts and bolts of the endeavors that they advise. As such, the courses are somewhat generic in content, glossing over the details of specific problems and eschewing case studies in favor of an overly formulaic approach.

This method lends itself to visually impressive aids such as flow charts and fishbone diagrams, but it does not place a high value on industry-specific knowledge and it does little to address the process as it takes place on the ground.

Most importantly, these courses tend to favor a democratic approach to problem solving rather than a firm emphasis on the scientific method. And they do little to prepare the student for the hard-nosed and sometimes hostile environment that can characterize the real world. It is perhaps not surprising, therefore, that in the world of plastic part design and manufacturing, the problem-solving process often has the appearance of an exercise in walking through the woods when lost or, alternatively, is characterized by a hard-headed approach that repeatedly manipulates the same variables in the face of continually negative results.

It has been said repeatedly, but it bears articulating one more time, that the failure of a plastic product to work as expected is caused by lack of attention to detail in one of five areas: part design, mold design, material selection, processing, and validation. Often the culprit is actually a combination of small deviations from the optimal in two or more of these disciplines.

Keeping this framework in front of the team working to solve the problem is critical because it tends to organize the thinking of the group. In fact, it is the guiding principle for creating all those attractive flow charts and fishbone diagrams. It also keeps the focus of the evaluation broad when too often there is a tendency to decide on the answer before any of the data have been gathered and to zero in on one factor to the exclusion of all others.

Focusing on the weldline

An example of this need to keep all areas in mind is the story of a cylindrical pressure vessel molded of ABS. The part enjoyed a successful history of unspecified duration while being molded in a particular grade of ABS. When this grade was discontinued by the material supplier, a new resin was selected based primarily on an approximate similarity in data sheet properties.

Production resumed. It was several weeks before problems were noticed with the product made from the new ABS. The problems were not evident in the as-molded part. It was not until the part was assembled and used that small cracks were noticed that prevented the product from maintaining a proper seal. The cracks were very difficult to see even under magnification; however, they were judged to be related to the location of a weldline.

This created a framework for the problem-solving process. The onset of product problems was related to the change in the grade of ABS being molded. Since the cracks were presumed to occur at a weldline, the natural assumption was that somehow this new grade of ABS lacked the weldline integrity of the original ABS.

This background suggests a couple of approaches. One would be to continue looking for an ABS that has weldline strength equal to or better than the originally specified material. This can be a frustrating endeavor since published data on weldline strength is not particularly abundant. In fact, this approach produced several candidates, none of which provided satisfactory results.

The other approach is to re-optimize the process using a new grade of material in order to increase weldline strength. This is often the time when design of experiments comes into play and selected process variables are manipulated in some combination in order to better understand the influence of these variables on the desired result. This happens when a well-trained staff is working on the problem. In most cases, a process technician is sent out to the press to work his magic, turning dials or pushing buttons in a fairly random manner until some combination of adjustments produces success.

Then there is the question of how the various proposed solutions are to be tested and validated. The failures did not appear until weeks after the parts were molded and they only occurred in assemblies that were in use. The feedback loop between process or material changes and the realization of results in a case like this is highly unsatisfactory. It simply takes too long to determine if the instituted changes have had an effect. And since the failure rate for the product was relatively low, it may take a lot of production to determine whether a real improvement has been realized.

But because this was presumed to be a weldline failure, a test was devised that would generate a high level of stress in an effort to examine the behavior of the weldline. The test was rather manual and therefore could not control an important component of any stress test: the strain rate. In addition, this test had never been conducted on the original material, so there was nothing to which results could be compared. And because the original material was no longer being manufactured, it was not possible to go back and recreate any of these experiments with the original material.

A shift in attention

After much press-side experimentation, parts were sent to the lab. Manufacturers are understandably reluctant to spend resources on laboratory testing. Often the people performing the tests are not familiar with manufacturing processes, so even when the nature of the problem is evident in the data, it may go unrecognized. Even when it is properly identified, the challenge remains of converting the knowledge gained from the tests into a practical solution. In addition, there is a cost associated with testing, whereas playing at the press is free. (We all know that it is not; but the machine, people, and materials are there anyway, so it does not feel like an expense to spend a few days or weeks experimenting.) But in the right hands, and with sufficient background information provided, laboratory tests can provide a much quicker route to a solution.

In this case, parts with cracks were provided along with a sample of a grease that was used to lubricate the inner diameter of the ABS cylinder and ease the sliding process of a piston that contains a rubber disk. The grease was of immediate interest because many lubricants and greases are stress-crack agents for amorphous polymers like ABS, and stress was clearly a component of this application. It was not clear whether or not this grease was a recent addition to the product. The plan was to characterize the composition of the grease and examine one of the fractures to determine how and where it started.

The cracks turned out to be extremely fine, and because they appeared to begin on the inner diameter, visual evidence for them on the outer diameter was scant at best. But a pressure test did confirm that the cylinder was not holding pressure and there must be a path to atmosphere somewhere.

Under magnification, not one but two cracks were found that ran most of the length of the cylinder. And here was the first revelation. Neither of these two cracks had anything to do with a weldline. The weldline could not have been farther away from the cracks. If the weldline was at six o'clock, the cracks were at approximately three and nine. A closer look revealed that the crack was actually a series of shorter, disconnected cracks rather than a single continuous fissure.

Once the fracture surface was opened and studied under higher magnification, the reason for the discontinuous crack pattern became apparent. There were multiple crack origins all along the cylinder that moved progressively from the inner to the outer diameter until one of them finally broke through to the outside diameter. The appearance of the fracture showed clearly that the cause of the failures was environmental stress cracking (ESC), a phenomenon that requires the simultaneous presence of stress and an accelerating chemical agent. If either of these factors is removed or in some cases reduced, the failures stop.

The stress was a fact of life for the application. So the attention was turned toward the chemical aspect-in this case, the grease. Unfortunately, an infrared spectrum of the grease showed it to be a substance that has no tendency to stress crack ABS. However, the same analytical technique applied to the fracture surface turned up a residue that was very much a stress-crack agent. It was a phthalate ester that is frequently used as a plasticizer in flexible PVC and in some thermoset rubber compounds, including the rubber used in the piston assembly.

These findings changed the focus of the evaluation rather abruptly. Suddenly, instead of examining process conditions or more grades of ABS, the root cause became the design of the entire product as it relates to the use of incompatible materials. It is conceivable that experimentation on process conditions and materials could have continued for a long time because the root cause of the problem was unknown. With the new information, the more appropriate solution now involved changing the polymer type altogether to one that is not susceptible to stress cracking by phthalate esters, or changing the rubber component to a compound that does not employ such a plasticizer.

It never fails . . .

Of course, a perfect scientific theory accounts for all observations. So in the interest of objectivity, it may be asked, if the plasticizer in the rubber stress cracks ABS, why did the original ABS not exhibit failure? There are several possible reasons for this, some of which are more ominous than others if you are the manufacturer of the final product.

First, it is possible that the material in the rubber seal changed at approximately the same time that the grade of ABS did. In fact, even a change within the same rubber family to a lower-durometer material might be sufficient to introduce a higher concentration of plasticizer into the system and increase the likelihood of an interaction. In this instance no such change was reported.

Another possible reason for the performance change was an increase in the stress field due to a change in the tolerance stackup between the housing and the piston. This was generally considered not to have occurred, but no data were ever produced showing conclusively that it was or was not a factor.

A third consideration comes from the material. Although no other grade of ABS was found that could match the performance of the original material, it should be remembered that ABS, while a very common material, is not a simple one. ABS is referred to as a terpolymer, a polymer composed of three different monomers. But it is more complicated than that. From a morphology standpoint, ABS is actually a matrix of styrene-acrylonitrile (SAN) with the butadiene rubber grafted onto the SAN backbone. High-magnification microscopy will actually allow for an identification of the rubber particles as a discrete phase. The amount of rubber can vary from grade to grade, as can the size distribution of the rubber particles.

In addition, SAN is itself a copolymer. The percentage of acrylonitrile can vary within a group of SAN polymers. Higher levels of acrylonitrile improve mechanical properties and chemical resistance and also raise the glass transition temperature of the polymer. All of these factors can improve resistance to ESC. These details of composition are proprietary to ABS manufacturers and therefore are not easily determined by processors or end users. This particular grade of ABS may simply have had the right combination of properties to prevent ESC from occurring.

Or, and this is the ominous part, ESC is occurring on parts made from the original material, but because of a slightly superior property or composition profile, the process is taking place more slowly and has not resulted in a complete breach of the part wall . . . yet.

When clients report that failures never occur in a certain material, it is an interesting observation. Never is a long time. It is known that ESC can be a slow mechanism. Even in the case of the material that was failing, the time frame was weeks after the parts were assembled. We usually view a product as never failing when it meets or exceeds the expected lifetime of the product. But when materials are brought into contact with one another that are inherently incompatible, the time frame for failure becomes a sliding scale with a slope that can be difficult to determine without feedback from the customer.

The history of failure analysis is replete with examples of products that were presumed never to fail-until they did. The best practice is to remove from the application all of the possibilities for failure to occur. That is the real purpose of FMEA, one of those other techniques they teach in problem-solving classes.