Tag Archives: trouble-shooting

Fix the Problem XII: When Chemicals Attack

Please see the end for a “good-faith” disclaimer.

Resistance to chemical attack and galvanic corrosion are big factors in many designs, especially at elevated temperatures.   Even at ambient temperatures corrosion can be an issue: the lowly chain-link fence has its steel protected by a zinc coating which, over time, corrodes instead of the steel – but given enough time, that protection disappears and then the steel itself starts to rust.

In this essay I will outline three examples I have seen where chemicals attacked solid materials, and for each highlight some lessons for people to consider – whether in a design or failure-analysis mode.

Plastics and Acids

At one point in my career I was asked to do some thermal stress analysis because of a series of returned field failures. This was a plastic container made of a relatively new material chemistry whose operation cycle involved being immersed in hot water to heat the contents, and then dunked in cold water to shock-cool them. Over time, cracks would develop in the corners of this injection-molded container. In some instances, catastrophic failure would occur resulting in ruptures of the vessel during the cold-water dunk. To the best of my knowledge nobody was injured in any such event.

The plastic material was believed to be immune to the mildly-acidic contents, so the prevailing theory was that the thermal shock was creating the cracks. This conclusion was logical considering that these failures occurred during the cold dunk operation. My job was to verify this presumed failure mode.

But my first analyses showed something very different than the observed failures. Both partially-failed returns, as well as analysis of sectioned pieces, showed the cracks initiating on the inside of the vessel’s corners. In contrast, every analysis I did showed that any thermal stress cracks should initiate on the outside, where the outer layers – exposed to cold first from the cold dunk – should contract first resulting in crack-opening tensile stresses on the outside preferentially.

Eventually, more detailed and at-temperature chemical compatibility testing showed that the acids inside were attacking the insides of the corners of the vessels.

Result: The material was pulled from this application.


  • Chemical compatibility charts and theoretical assumptions are good guides, but verification testing needs to be done on the actual application with real-world conditions.
  • Molded-in stresses at the corners likely contributed to the problem; corners in injection-molded pieces tend to have higher stresses than the plastic material of the part in general.
  • The thermal shock cycles, while not the root cause, doubtless did contribute. Rapid thermal shocks should be avoided where possible, or included in any factor-of-safety calculations if unavoidable.


Stainless Steel Stress Corrosion

A heavily-welded assembly was found, in the field, to have a pin-hole leak at a weld joint resulting in its hot and pressurized contents leaking out (an obvious safety issue; as above, to the best of my knowledge nobody was harmed). This piece was removed from service and sent for cleaning prior to repair. The cleaning process involved, among other things, the assembly being repeatedly soaked and washed in an agent recommended by the cleaning facility for this type of application and assembly material. A spot-weld repair was then made over the pin-hole leak.

Not long after being returned to service the assembly, which had multiple independent channels only one of which was to be “live” at any time by design and operating procedure, was found to have channel cross-talk – leakage of the working fluid from one channel to another. It was again pulled from service; I was put in charge of the failure analysis.

Even a visual analysis of the piece showed an obvious crack on an internal surface. Pulling prints of this assembly, plus comparable assemblies, showed several differences – only one of which is specifically relevant. To wit: the design (unique to this one variation) resulted in a thin-wall condition in certain locations. A worse-case tolerance stack up would exacerbate the situation significantly.

Sending the assembly out for metallurgical examination, including with it the process history of the cleaning and spot-weld repair including the cleaning agents used, resulted in several key findings.

  1. Sectioning of the cracked location and subsequent microscope examination showed clear evidence of stress-corrosion cracking, with cracks fully penetrating from one channel to another. The pictures of the cracked area were “stereotypical” of this failure mode.
  2. The primary cleaning agent, though recommended for the base material, was discovered in a literature search to create stress-corrosion cracking in this material at high temperatures and applied stresses. The suspicion – believed without explicit proof – was that trace amounts of the cleaning agent remained in the relatively-rough surface finish of the machined surfaces.
  3. Welded structures in general, especially ones with significant machining and welding such as this one, tend to have high residual stresses which are known to play a large factor in stress-corrosion cracking.
  4. In operation, tightening of threaded fasteners close to the thin-wall area introduced further stresses in that region from the torques required.
  5. The design, with its thin-wall condition even at nominal design dimensions, was at the root of the problem.

Result: All assemblies of this design were pulled from service and scrapped. Additionally, this particular design concept was ruled out from consideration in the future.


  • Check designs for thin-wall areas.
  • High temperatures exacerbate any stress-corrosion situation.
  • Stresses, whether residual from processing (e.g., machining or welding) or from other operations (e.g., threaded fastener tightening) can also introduce stresses which increase chemical-cracking susceptibility.
  • When using any agent to clean, not only vet it with both the base material supplier and chemical agent manufacturer’s recommendations, but check with a literature search to see if similar situations have been reported as causing failures.
  • In parallel with the above, go above-and-beyond when it comes to rinsing, in particular if the surfaces being cleaned have a rough surface finish. (And if there is a thin-wall area and it is visible and/or directly accessible, spot-rinse the area to beat heck.)
  • This investigation relied heavily on the expertise of a skilled consultant; in this case, a metallurgist. When specialized expertise is needed, don’t hesitate to ask an expert once you realize you do not have the requisite knowledge.


Galvanic Action Surprise

In one application where a cermet component was used in an oil-based environment, the assembly where these small cermet components were used was initially had a carbon steel base part. For machining convenience and better machining-process yield, a switch was made to a stainless steel base.

Not long after this change a new failure pattern of the cermet pieces was observed in parts returned from the field. Initially thought to be a defect in the cermet piece, both the cermet piece supplier and an independent metallurgist provided links and other information pointing to galvanic corrosion as the likely suspect after being informed of the base material change.

This seemed strange, as galvanic corrosion requires water; this was immersed in oil. However, chemical analysis of the oil showed not only measurable amounts of water, but of materials (salts mostly) in the oil that – when exposed to the entrained water and then dissolved in it – would form a conductive path which is essential for galvanic corrosion to occur.

The switch from carbon steel to stainless moved the situation from one where the carbon steel had been the material degrading across a large surface (and therefore going unnoticed) to one where the much-smaller cermet was preferentially attacked – as stainless steel has a much higher resistance to galvanic corrosion than carbon steel.

Result: The application was switched back to a carbon steel, though a different one than the original for better and easier machining. No further failures of this type were observed.


  • Even oil environments can have enough water and trace elements to create a pathway for galvanic action to occur.
  • If changing materials for machining considerations, consider a material of a similar type (e.g., in this case merely a different carbon steel), rather than a major change of material type.
  • If necessary to change to a material more resistant to galvanic action, keep an eye on materials in the system that have not changed as they may now be attacked by galvanic corrosion when they were not before.
  • Most suppliers want the application to succeed; tap them as knowledge experts before making such a change. If they’ve been in the business for a while, odds are they’ve seen similar situations before.
  • Again, don’t be afraid to consult with outside subject-matter experts on your dime to be sure there is not a conflict of interest (i.e., the supplier had their own metallurgist; the one we hired was a second opinion without that potential influence of their working for the supplier).


Disclaimer: I was very opaque about details like when and where these occurred, and the specific applications – and deliberately so. I am attempting to tread the fine line separating writing with just enough detail to be educational, and not disclosing anything proprietary (or damaging, though note: there were no dangers from the third example, and in the first two – to the best of my knowledge – nobody was harmed and in all instances situations have been remedied and are no longer a danger going on multiple years now). No company names are mentioned, nor are process parameters, design details, or specific material callouts given except in an “arm waving” mode to set the stage. Should there be a desire to change anything on the part of the companies involved, please contact me directly and make some suggestions as to rewording to remove any information that might currently be creating a material-damage situation – a situation which I have labored in good faith to avoid with my deliberate vagueness.


© 2015, David Hunt, PE

Fix the Problem IX: Three High-Scrap Issues on a Product Family

In Fix the Problem VIII I introduced a simple concept, requoted here:

You can only claim to have truly solved the problem when you understand it well enough to turn it on and then back off again.

I want to discuss three production problems, all on the same product family – but with three different root causes.


While working on a cost-reduction team at Ford Motor Company that focused on material and labor savings on existing products, we would occasionally be called upon to act as extra eyes on specific problems in the plant. I was asked to look at three different radiator lines and their “high fin” scrap rate.

Let me define “high fin”. In automotive radiators there are thin, flat tubes that carry the radiator coolant from one header tank to the other; heat gets conducted through the tube walls and into the corrugations made from thin aluminum sheet. This sheet – really an aluminum ribbon – is transformed into the corrugated fins by use of a star-shaped wheel that acts as a mandrel to form the up-and-down corrugations. Functionally, these fins wick heat from the tubes of hot coolant and provide a large surface area for moving air to remove heat from them, thus providing a heat sink.

These fins, in turn, are placed between the flat tubes; the headers that formed part of the coolant reservoir and manifold (which evens the flow distribution between tubes) are placed, the assembly compressed, and steel clamps put on to maintain the tension until the radiator is brazed. As a part of the brazing operation, there is a wash with soapy water that sprayed down onto the radiator face to remove grease and other contamination prior to the braze heat being applied.

In each of these products, fins were being knocked loose from between the tubes and then brazed in that out-of-position position. This was not repairable, resulting in a piece of scrap and a brazing cycle wasted.

First Steps

In all three cases I got the last three months’ worth of scrap data to look for trends, and had scrap items put aside at the end of the line for me to look at, with the time and date of the exit from the braze oven written on them. I also had an intern temporarily assigned to me, and had them – on one shift per day (ideally this should have been for all three shifts, but it wasn’t practical) – watch the radiators going into the braze oven to see if we could determine where in the process stream from raw material to end product the defects were occurring. (Note: no high fins were seen going into the braze oven; this point, first made in Fix the Problem II: Medical Blister-Pack Punctures, is important – knowing where a problem starts to be seen can be a critical clue in identifying the root cause.)

Radiator One

This particular instance is shown on a portfolio page, here.

The most important thing was that over the last three months the fallout rate was getting worse. Something was clearly going in the wrong direction, so I started here. The next thing was that I put the radiators in a row, by time and date. I also graphed the data; my hope was to see if there were any shift dependencies or pattern in the locations of the high fins. They appeared to be random in both time and location.

Time for basics. What holds the fins in place until they’re brazed?

Looking at the prints there was an interference between the nominal fin height and the gap formed by the tubes. In other words, after the fixture compressed everything, the fins were compressed like springs, pushing against the tubes with friction holding them in place (thus the need for the braze bars to hold the compression). Clearly, something was not holding them in place, and not doing so in an apparent random fashion.

I polled some colleagues about the problem to get outside perspectives. What could be causing this? We did an Ishikawa fishbone diagram, went through a few Five Why exercises, and the dominant factor implied something was awry dimensionally (something could have changed in the material, too, but this seemed the best first-guess approach especially as there were similar products with no issue); so what dimensions would be critical for this function? I had a pretty good idea, of course, but I wanted other minds’ input to make sure I didn’t miss anything. The consensus was that three dimensions were the critical ones:

  1. The center-to-center distance on the aluminum headers into which the tubes were inserted prior to brazing.  This was an inspected dimension.
  2. The outside-surface-to-outside-surface of the flat tubes.  Again, this was an inspected dimension.
  3. The top-to-bottom distance of the corrugated fins.  Once more, this was an inspected dimension.

But what did the data say about these? For the header dimensions and the flat tubes, these were spectacularly consistent, with high process capabilities as indicated by their CpKs. But the fins were another matter.

We had a laser system that would measure the peak-to-peak dimensions of the corrugations. The way it worked was that a fin would be put in, and the data would be aggregated into an average and recorded. I wanted the raw data, and measured several fins and then graphed them: height as a function of position along the length of the fin (which comprised, perhaps, 80-100 peak-to-peak dimensions). And then I stared at those graphs.

Sometimes, all that’s necessary to solve a problem is “sit and stare” and let intuition and your subconscious work. In a EUREKA! moment I said “There’s a cycle in there.” So I took the data and – manually – did a best-fit sine wave, finding that the height of the fins was going up and down with a cycle every 26 corrugations.

Clearly the flat ribbon coming into the fin-making machine didn’t have a pattern in corrugations; there no were corrugations at that point. Similar to my example, here, I “walked the line” to find where the corrugations were formed by the fin machine. Then I went to the maintenance guys, those people who knew the equipment, showed them my sine wave, and asked what might be causing the 26 cycle. Immediately they replied “the star wheel”. The star wheel was what formed the corrugations, having 26 teeth.

Fortunately, the maintenance foreman was a sharp cookie; he had the machine shut down and taken apart. What they discovered was that the bearing on which this star wheel rotated was deteriorating, and said deterioration was introducing an eccentric motion as the wheel spun while it formed the corrugations. The bearing was replaced, everything aligned, and we started up again.

I redid the measurements – no cycle at all. CpKs were much better. Most importantly, high fin scrap fell to zero. To prevent this from happening again, the fin measurement procedure had an alert callout added for when the CpK started to drift above a given limit; standard SPC practice to alert the technician that the process was going out of control. The machine maintenance plan was also modified to include periodic alignment and bearing checks.


  1. Averages have their uses, but lots of information can get washed out by relying on them without examining the raw data.
  2. Using the five-why process, in this case to drill down to dimensional variation (a basic engineering fundamental), was instrumental in identifying the root cause.
  3. Intuition again played a key role, as did leveraging the different perspectives and experience sets of other people.
  4. Identification of where the problem was happening – somewhere inside the brazing process – was also a key datum.

Radiator Two

Unlike the first radiator, there was a definite and specific location to the high fin: it always happened right next to the side-rail Z-notch that was in the top and bottom rail forming the outside of the radiator, and it did it across all shifts; nor was there any change over the months indicating this was a chronic problem. Let me discuss the Z-notch.

The side rails were aluminum C-channel rails whose long center flat was brazed to the fins. In each rail was a stamped cutout, called the Z-notch. It created enough structure for handling and processing, but at the end of the line, the remaining material would be cut so that as the radiator heated and cooled there was a place in the rail allowing it to expand and contract freely – like heat expansion joints in a bridge. The high fins in this case were always, always, always at this location.

Since this problem was not observed across other lines, the question to answer was “What’s different?”

This ended up being very simple; I went to the prints – usually my first stop in any problem-solving exercise. Was the material different? No. What about other parameters? The length, one of the things that defines the stiffness of a beam, was “in the pack” of other side rail lengths. Nothing sticking out there. And the cross-sectional dimensions of the C-channel were identical to the others as this raw material was used across virtually every line. But as it turned out, the Z-notch was a different design than the others.

Looking at the geometry, the stiffness of the remaining material at this joint – the moment of inertia that quantified the stiffness of that rail at that point – was less than half what it was in any other design. In essence, that lack of stiffness formed a hinge point, allowing the forces from the compressed fins pushing outward to deflect it more, thus lessening the retention force on the fins – so that the force of the cleaning spray could knock it out.

I checked downstream with the operation to see if changing the Z-notch would impact them; no. I then approached our Design group, pointed out the differences, and asked if we could do a trial run with a standard Z-notch design. A week’s worth of test pieces were made, the appropriate trial forms signed, and a run was made. High fins disappeared with no impacts downstream – as checked for beforehand. Based on my findings and the successful trial, an ECO was written.


  1. In looking at a problem occurring on one product of a multiple-product family, the very first question while working through the possibilities should be “What’s different?”
  2. As with the dimensional analyses above, engineering fundamentals – in this case, an application of my Strength of Materials course – were essential to understanding the issue.

Radiator Three:

As with radiator two, the high fins would always occur right next to the side rails. However, there was one key datum that stuck out: they only happened on one shift out of three. One compression-assembly machine, but – obviously – three different operators. Therefore, in Ishikawa parlance, the likelihood was that it was a man (of man, machine, material, method, measurement, mother nature) that was the first thing to be investigated.

Under the guise of “just poking around” I stayed late on the second shift to watch the operator out of the corner of my eye. He, like the others, worked diligently… but there was one key difference in how he worked. (“What’s different?”) All the other operators would place the radiators on the conveyor belt, while he dropped them. But although he was dropping them, it was only from a few inches.

Getting permission from the foreman on the first shift, I did a few tests to drop the same radiator onto the belt. And nothing. This was interesting. I did some drop tests with other radiators, and nothing there as well. Clearly the drop was a difference, but not the determining difference.

So I talked with the operators; asking “What’s different about this radiator?” I was thinking that there might be some subtle difference, but all the operators I spoke with, on two shifts, had no comments. A review of the prints didn’t show anything obvious. And a key question, which had to be handled very diplomatically, was “What was different about this operator?”

As it turned out, he was a new hire, having been brought on within the past year; all the other operators were veterans with years of experience. Why did this make a difference?

As I’d mentioned in the introduction, steel clamps were put on to hold the tubes and fins compressed until the brazing operation could fuse them together into a solid piece. In talking with the operators, one casually mentioned that the clamps, called braze bars, had to be installed through small positioning notches in the C-section side rail, and had to go through both sides. If they didn’t go through, placing them on the conveyor could create a force that could distort/damage one of the C-section sides “especially if the radiator were dropped” since they were always put on a conveyor with the braze bars underneath the radiator. AHA!

I spoke with the second shift foreman, who had a discussion with the operator. He was not aware that the braze bars had to go through both sides of the side rail notch. Most of the time he got it right, but sometimes he didn’t – and not knowing how critical it was – would then drop the radiator on the conveyor. (Likely he had been told but merely forgotten.) The impact of the drop would be passed through the misaligned braze bars, distorting the side rails and allowing decompression of the fin.

This was the missing element. After being made aware (again, more likely reminded) of how critical this was, along with an explicit instruction to place, not drop, the radiator, the issue went away. I did design plastic lead-ins and got them mounted on the machine as a trial. The theory was that an operator could put the braze bars into the lead-ins and literally drop them in; the lead-ins would align everything so that the operator could not do it incorrectly. It worked, but the operators didn’t like them so they removed them.

  1. When faced with a shift-dependent problem the very first thing to be considered should be operator error.
  2. Had this not been the root cause, some kind of environmental issue – would have been the next thing to be considered; since things like temperature, light level, and being aware/awake can be affected by shift schedules, even when people are accustomed to their off shift schedule.
  3.  The ideal solution, when faced with an operator-dependent problem, is to engineer it out.  But all the best efforts are naught if the operators don’t like the solution and remove it.

© 2014, David Hunt, PE