Failure Analysis - Case StudiesDavid Hershberger
Failure Analysis Consultant
Tonight we will get a feel for failure analysis by looking a six examples of failures for which I've determined the likely cause.
In a town not far from here a water main was leaking seriously, and some determination of the reason for its failure was required. The subject was a large diameter cement lined cast pipe section, flanged on one end, with a crack most of its length and a section of wall missing adjacent to one portion of the crack. Since most cracks start from a surface and propagate inward, it was decided to closely examine the area where the crack came out to the flange at the end of the section. A black coating was observed on the surface of the crack in this area, which matched the black coating observed on the cement coated surface of the pipe itself. Use of SEM and EDAX reveled this coating to be a form of iron oxide. Cast into another portion of the pipe was the date 1874 which indicates the casting was 129 years old. Since black scale is a common occurrence of the casting process itself, I concluded that the origin of the failure was a casting defect (crack) in the flange of the pipe when the pipe was originally cast. Because of the flake like structure of the grey iron, with time (129 years)the crack propagated the length of the pipe section. It was not possible (in 1874) to easily detect small cracks in a large casting (no NDT), so there is very little likelihood the defect was known about. Traffic passing 30" over the pipe in 1874 probably never exceeded a few hundred pounds per axle. Now days it could be several tons. This failure was a natural result of age and old technology, and can be expected at more locations along this water main.
The second example is a hydraulic manifold which is part of a test stand to induce vibrations into vehicles for evaluation. Hydraulic fluid was leaking through this aluminum manifold. A threaded hole was used to connect the main hydraulic line to the manifold, and the crack propagated to this connection. Close microscopic examination showed that the crack actually originated at the root of a thread of this hole. Careful examination of the crack surface (after sectioning to examine it) showed fatigue ripples concentric to that point as a center. This made it fairly obvious that the crack initiated at the thread root and propagated by fatigue through the aluminum which is rather poor at resisting fatigue. Changing to a fatigue resistant steel manifold should solve or at least delay the problem significantly.
The third example is a basket (perforated steel) used to muffle sound in a steam application. Temperature was not precisely known, but was originally suspected in the failure of the baskets which simple seemed to crumble apart. Close microscopic examination of the 1/4" thick 1006 perforated plate and broken surfaces showed lots of rust (as would be expected in a live steam application). On crack surfaces which were less rusted however, fatigue ripples could be clearly seen. Close examination of the cross section of a perforation showed that there was a lot of cold work right at the surface of the perforation (probably the result of punching). Microhardness verified that this was a much harder layer, and therefore quite subject to cracks (several were observed). Once initiated these cracks easily propagate by fatigue which lead to the massive failures of the baskets. Stress-relieving after punching would greatly help this problem.
The fourth example involved a bending mandrel made of 4140 steel. A square slot (rounded corners) was present which allowed for linkage to the next mandrel section using a spring. Examination showed that the failures were initiating from points on this slot. Close etched microscopic examination revealed that a thin white layer of untempered martensite was present on the surface of this slot. This makes for a very hard and very brittle surface, ideal for crack initiation. It turned out that this slot was being produced by wire EDM. Since the EDM was effectively re-casting (melting) minute amounts of material and the mass of the mandrel and aqueous environment was producing a very rapid quench it was evident that this was responsible for the failures. Tempering this material after EDM would greatly mitigate this problem.
A fifth example example is a brass fitting in a domestic dishwasher. A white coating could be on the fracture. One might think this is calcium from hard water, but EDAX showed that it was Zinc oxide. With the zinc oxidizing, it was found that the failures were the result of de-zincified copper.
The final example is a stainless steel fitting used in food processing (making ketchup). The stainless steel fits to an o-ring. It was observed that this ring had been "weld repaired" several times (probably to keep the ketchup production moving). Corrosion appeared to be the problem, and EDAX showed corroded surface material to be high in chlorine. Further discussion with the owners revealed that the system is periodically cleaned out using chlorine bleach. This was clearly responsible for the corrosion to this fitting, and a non-chlorinated cleaning procedure should be found.
That is a sampling of what failure analysis is all about. Basically my philosophy is to start the investigation with careful observation, a magnifying glass, and a magnet, and follow hunches based on that evidence with more sophisticated techniques.