How is Failure Analysis Done?
This article is an introduction to failure analysis of the inner structure of metal and plastic parts. It covers issues with that type of failure analysis and provides examples.
Failures, indeed events in general, happen when factors interact in unexpected ways. This is important to understand. Many companies assign “getting a failure analysis” to people without much technical background. The results of such work are often disappointing and confusing (Fig. 2). They often leave everyone wondering why they bothered to spend any money at all on the analysis.
The typical way this happens is that someone sends a part to a lab to find out if “the material met the specification.” Here’s that special mystery. Everyone can see the dimensions, and everyone can see the color. Everyone can see that the subject part is rusted, broken, deformed or worn out. There are a lot of people who are fairly good at understanding the mechanical aspects of what their “gizmo” does (Fig. 3).
What everyone can’t see is the hardness, strength, composition and microstructure. The microstructure is essentially the record of the processing history of the part. If the right material was used, the microstructure is what will shed light on whether it was processed correctly. You can have the right hardness and the wrong microstructure. Right and wrong are useful terms, but they rarely have the absolute objective meaning that people who are not trained in materials engineering think they have. The “cause” of many, perhaps most, of the failures I look at can be seen as a failure to specify the part manufacturing process in enough detail.
The Laboratory Analysis
Going back to why we “send the failed parts to a lab,” let’s forget about the microstructure and processing history for a bit and focus on what is probably included in most specifications that the typical smaller company cannot check on its own. This includes determination of the basic material composition. For metals, this is relatively straightforward if there is a readily accessible flat surface that can be ground on a belt sander to expose a half-inch-diameter circle of bare metal. Then the part can be analyzed with a spark-type spectrometer.
Many parts, however, are too large to fit in the chamber of the instrument, so they will have to be sectioned. Thus, we see the first problem with substituting a certification check for a failure analysis is that most of the tests that are used to determine if something meets a material specification are destructive. Once the assembly is taken apart – once the part is cut – we lose forever some of the information that might have helped confirm or disprove some of the possible failure scenarios.
If the part is too small for a “spark test,” then some portion of it must be dissolved in acid or consumed by burning as part of the analytical method. This is obviously even more destructive.
If the “failed part” is made of a polymer resin, different methods are required to analyze it. In this case, it is the molecule types, not the atomic species that are of interest. Polyethylene, polystyrene and polybutadiene all are made of carbon and hydrogen. Some type of infrared analyzer is commonly used to determine the resin type by looking at the atomic bonds. But if the polymer is filled with glass or mineral, additional tests must be performed in order to determine if the correct amount of glass and mineral fill were included in the resin. Note that glass is used to strengthen most resins, while the mineral is used to reduce the cost and can weaken the material.
So, knowing the amount of filler is not enough. We must know the amount of glass and the amount of mineral. If you want to differentiate between different types of nylon or polyurethane, additional tests are needed. By now, you have probably spent somewhere between $500 and $1,000 if you have to go to an outside lab.
Worse yet, whether we are dealing with a metal or a plastic part, we have not done anything to determine whether the distribution of the different constituents making up the material is uniform or not. That is usually evaluated with techniques of microscopic examination of specially prepared cross sections (Figs. 5-6). This may be done on metallic and polymeric components. But since uniformity of distribution of the composition elements is one of those usually unspecified things, let’s return to the things that are more often specified.
For a metal part, hardness is a commonly specified characteristic. There is no problem here if the parts are large and the engineering print clearly specifies where the test is to be performed and with what scale. This might be the case on a 5-kg forging or casting that specifies a Brinell hardness. Unfortunately, people who put hardness specifications on engineering prints are often blissfully ignorant of how the tests are actually performed.
The common practice for small metal parts is to use a Rockwell B or C scale. However, there is a minimum thickness for all parts, a minimum diameter for curved surfaces and a requirement for either a flat surface parallel to the surface to be tested or a convex curved surface. Technically, it is not valid to test a concave surface with a Rockwell. For convex surfaces, such as solid bar stock, there is a round-work correction to make up for testing some air, but there is no way to correct for extra metal on a concave surface.
The story gets more complicated if the part is case hardened. Microindentation hardness profiles are the proper way to test these parts, and that requires preparation of a cross section. Hopefully by now, the point has been illustrated that “just get a hardness test and composition check” is not so simple! Even if the situation does not demand a full analysis, if it is worth doing any testing at all, it is worthwhile to have someone knowledgeable about the use of the part converse with whomever is going to be supervising or performing the tests. Therefore, the tests can be customized to shed light on why the part did not perform as desired. Without doing this, the results of the testing can actually be misleading.
Elements of Failure Analysis
What is the difference between a series of tests for conformity to the specification and a failure analysis? The essential difference is that a failure analysis includes a visual inspection, document review and a plan to perform a consistency check on all the data.
There are many types of visual inspection. When parts are manufactured incorrectly or field inspections for a service problem are planned, a visual inspection might be arranged. For economic feasibility, the inspectors – human or machine – must be given a finite list of things to look for. Maybe it’s a scratch, a dent or some rust. Maybe the paint is the wrong color or a slot never got punched. Whatever it is, nobody wants to pay someone to look for a scratch, a dent, some rust or the wrong shade of fuchsia if the problem is that the slot was never punched. Humans are simply incapable of looking for an indefinite number of features simultaneously. We basically have to memorize the list of features and look for each one separately. We are incapable of looking for “everything that might be unusual” unless we have plenty of time.
Having that extra time is the first of two major differences between inspection for defects and for failure analysis. If the part has fracture surfaces visible, then part of the visual inspection will be a fractographic analysis of the characteristics of the fracture faces and surrounding areas. This includes an evaluation of the shape of the part and the ways that the loads are transmitted to the part. It is often possible to determine whether the part was loaded in bending, by twisting (torsion) or in shear. Some components are intended to be loaded in tension or compression.
A key piece of information that a fracture analysis provides is a comparison between the apparent orientation of the loads that caused the fracture and the orientation of the loads that were intended for the component. Torsional loading can create stresses that are double those that would happen with tensile or bending loading. Note the caption of Figure 3. Also, unplanned-for loads can alter the stress pattern in such a way that a location that was never intended to sustain a high stress is now subject to one. Note that this was not the case in the part shown in Figure 1. The fractographic portion of the visual examination is critical to a fracture analysis.
The other “half” of the visual inspection for a failure analysis includes something that most people are not familiar with. It requires looking at the part, preferably with as few preconceived ideas of what might have caused the problem as possible, to see what is there in an attempt to find things that would never occur to anyone to put on a list.
One of my favorite parts to illustrate this is the stainless steel tube shown in Figure 7. It was determined that the corrosion that caused the hole was due to the acid excretions of a bacteria colony. The bacteria like the acidic environment. So, as the corrosion proceeds and pits are formed on the corroding part, the bacteria colonize the pits, eventually tunneling between the inside and outside surfaces of the tube wall.
When I reported this to my client, he said, “Hmmm ... yeah ... right. There were pigeons flying around inside the plant. Bird-poop corrosion! OK. Makes sense.”
The analyst was never at the plant and never saw the part in service. This is an excellent example of a “consistency check.” There was actually a reason that there might be a high level of bacteria on the part. On casual inspection with this fact in mind, the shape of the stain indeed looked like it could have been made by a pigeon plop (Fig. 7).
This brings up another issue. When moving the physical evidence, it is important to document the positions of mating parts before disassembly. When looking at any component from a piping system of any kind, knowing whether the portion of the system in question was in a vertical or horizontal run can make interpretation of the data much easier. Learning to extend this principle of evidence preservation to all the other possible types of failures in all the possible types of situations in which components fail is a lifelong task and one of the biggest challenges of failure-analysis work. Recommending a simpler inspection for the remaining screws in service mated to that shown in Figure 2 would have been possible if we knew the orientation of the larger and smaller cracks in service.
The results of the visual inspection and time that has passed allowing the different bits of evidence available from this initial review of the component are what allow intelligent selection of test locations for composition and hardness (and microstructure if desired) so that the data obtained is most likely to provide the necessary information while retaining the greatest chance of getting more quality data if needed at a later time. IH