MATERIALS: Don’t Guess Your Way Through Root Cause Analysis

Like most phrases that become buzzwords in the business world, root cause analysis gets talked about a lot more than it gets done. While there are a lot of reasons for this, they come down to two critical factors: lack of commitment from upper management that manifests on the ground as a lack of resources, and lack of training in the key tools and techniques that help get to root causes. It is possible to fill management books on the first factor. I am more interested here in the second factor and how it plays out in the problem-solving process.

When performance problems arise with a product, there is a tendency for those affected to view the problem through the lens of their training and background. Processors will typically focus on the raw material. Material suppliers have traditionally provided a significant amount of technical support to both processors and end users to assist with design, property assessments, and processing issues. Much of this assistance is proactive, and some of the processing assistance is provided, at least in part, to manage the perceptions of molders that raw-material variation is a common source of processing difficulties.

Manufacturers of the raw material will be particularly adept at identifying problems with part design, tooling, assembly processes, and processing techniques. The end user, who is ultimately responsible for coping with the fallout from the problems with either the raw material or the molded component, often lacks the ability to make a determination regarding root cause but takes the position of demanding that the supply chain correct the problem by whatever means are available. 

If within that supply chain there are secondary-operations providers such as painters or platers, they will have their own input on how to solve the problem. However, much of this weighing-in is guesswork and posturing in an effort to deflect responsibility for the problem onto someone else. It is rare that anyone, at least in the early stages of the problem-solving process, brings anything like hard data to the discussion.

It is at this point in the discussion that you will hear statements like, “The molding machine is computer controlled, so the problem cannot be with the molding process.” Or, “We sell this material to a lot of our customers and no one else is seeing this problem.” Or, “The polymer never fails; we understand these materials and know their capabilities. If the material is failing it is automatically a case of expecting too much from the compound.”

Because the molder is frequently the small operation caught between a large material supplier and an even larger end user, the burden of proof often lands on the molder’s doorstep and the problem-solving activities focus on the molding process. This leads to a lot of experimentation at the machine that can range from trials where random adjustments are tried to more structured and scientific approaches such as designs of experiment (DOEs) that are intended to find the so-called “red X.”

While all of this is well intended, it often occurs without any real knowledge of what the most likely causes of the problem might be. This results in a great deal of activity without any guidance on what direction the problem-solving process should take. Repeated trials, followed by some period of testing, consume a lot of time. Problems often continue for months or even years without resolution because the participants assume that they know the source of the problem and often persist in investigating possible causes without any real knowledge that they are going down the correct path. 

If the problem continues for a long period, arriving at a solution is made more difficult by the lack of continuity that results when people move into different activities and job responsibilities, leaving new people to continue the process without the benefit of thorough documentation or a complete history of what has gone on before they arrived on the scene.

Recently I worked on a problem of cracking in a molded part. The defect occurred in a relatively small percentage of the parts and took quite a bit of time to appear after the part was produced. Attempts to reach a resolution had absorbed a substantial amount of time and talent. The mold had been moved to a different supplier at one point without eliminating the problem. A lot of time had been spent developing tests that attempted to replicate the failure under controlled conditions. These had not been successful. Finite-element analysis (FEA) had been performed at some point; but the results were not useful because a linear analysis had been run. This led to a conclusion that the maximum stress on the part was over eight times the yield strength of the polymer.

If this were true the cracks would not have taken weeks or months to manifest—they would have occurred instantly. At no point in the process had anyone conducted any analytical tests on the raw material or the molded parts. These tests, when properly prescribed, conducted, and interpreted, can provide a wealth of information about the composition of the material, the origin of the crack, and the mechanism by which it is occurring; and can answer essential questions regarding whether or not the polymer was degraded during the molding process or was made to the correct specifications.

An evaluation of the molding process by a third-party expert identified a number of concerns that had escaped the attention of the people who had been investigating the problem. This was partly because no one on the problem-solving team recognized the need for a greater level of control over the process than was already being provided. This was compounded by the type of tunnel vision that often sets in when the same people have been working on a problem for an extended period of time. An outsider will often ask the questions that have not been considered by “the team.” 

In this case, the processing consultant identified several potential processing concerns that could contribute to the cracking. These included a poor machine response, a large and uncontrolled variation in cycle time due to semi-automatic operation, a large temperature gradient in the nozzle, and the possibility that the material, a polyester of undetermined composition, was not being dried completely. While these were all important considerations, there was no hard data linking these problems to the cracking.

This is where analytical testing comes in. There are more than 50 different test techniques that can be used on polymers to answer questions about root causes of failures. Of these, typically five or six are used frequently. In this case, melt-viscosity tests ruled out degradation. Thermal analysis and infrared spectroscopy answered important questions about the composition of the polymer and showed that the material was undergoing a change in structure near the gate caused by the geometry of the flow path and potentially aggravated by a process condition.

And microscopy showed that the crack was starting from the same area near the gate where the structural change had been detected. Once all the tests were performed, the cause of the problem was clear and the needed corrections to the mold and the process, as well as a plan for controlling the key variables of the process, followed logically.

If this is such a successful approach, why does it take so long to get around to it? Part of the problem is that many people with manufacturing backgrounds are not aware of the potential for analytical testing to solve problems. But there is another barrier, which involves the ability of the analytical people to deliver a useful result. This will be the topic for Part 2 of this series.