The Heat Exchanger That Failed Twice in One Year

Why this matters

The first diagnosis on this job was right. The metallurgy call was right, the failure mechanism was named correctly, and the repair was the one most competent shops would have made. It came back in nine months.

That is a more useful case than a botched diagnosis, because there is nothing to correct in the technical work. What was missing was a question nobody had a habit of asking: not what killed the part, but what set the speed at which it was being killed. Those are two different investigations, and the second one only becomes obvious when the interval collapses.

Stop before you open it

The exchanger sits between an open evaporative loop and a closed loop, and opening it puts three hazards in front of you at once.

Isolate both sides, lock the isolation, and verify depressurized at a gauge on the section you are opening rather than elsewhere in the system. Stored pressure in a closed loop is hazardous energy under 29 CFR 1910.147. Take a temperature reading before breaking any joint, because hot water at pressure flashes as it exits and the scald reaches past the flange.

The tower-side water is the one that needs a control most people skip. Open evaporative water is warm and holds biological growth, and the hazard route is inhalation of aerosol rather than contact, so keep the draining low and slow with no spray, keep other trades and occupants out of the space during the work, and where aerosol cannot be avoided, respiratory protection belongs under a 29 CFR 1910.134 program.

Any acid cleaning of the plates is a separate hazard with its own entry: chemical splash goggles and face shield plus the glove class named in section 8 of the product's safety data sheet, and confirm from section 10 that nothing in the space carries hypochlorite, because tower chemical storage frequently does and an acid meeting hypochlorite releases chlorine gas.

The first failure and why the diagnosis was right

The exchanger was installed in March and ran until February of its sixth year, 71 months, when it began passing water between the two sides. It was pulled and opened.

The plates showed pitting on the tower side: small, deep, discrete penetrations with sharp edges, clustered rather than uniform, on stainless plates. That morphology is not general corrosion and it is not erosion, and on stainless in an open evaporative service it points hard at chloride. The tech called chloride pitting.

That call was correct and it was made on evidence rather than on habit. Pit morphology, the material, the side of the plate affected, and the service all agree. The library's cards on reading pitting against uniform and crevice attack cover the morphology read itself, so it is not re-derived here.

The repair was also defensible. The exchanger was replaced with the same specification. Upgrading the alloy would have been a guess without a chloride number, and specifying a more resistant alloy on a hunch is how a shop spends a customer's money on a variable it has not measured. The plates were replaced, the system was flushed, the unit went back in service in February.

The second failure, and the number nobody had computed

It failed again in November of the same year. Same mechanism, same side, same morphology. Nine months.

At this point most shops reach for one of three explanations: a bad batch of plates, an installation error, or a supplier who shipped the wrong material. All three are worth ruling out and none of them was true here.

The number that reframed the job took thirty seconds. The first unit ran 71 months. The second ran 9. That is a ratio of about eight to one on time to failure, on the same equipment, in the same service, on the same water source.

Be careful what that ratio does and does not say. It does not let you back out a rate law, and nobody should try: two data points cannot establish an exponent, and the relationship between chloride concentration and pitting rate on stainless is steep and depends on temperature, surface condition and crevice geometry. What the ratio does say is a great deal simpler and more useful. A collapse of that size is not drift. Drift produces a shorter second life, not a life one eighth as long. A ratio like that says a step change happened in the environment, and the job is now to find an event.

That is the whole pivot in this case. The diagnosis moved from the part to the rate, and the rate has a date.

Looking for the event

The second unit ran 9 months, all of it after February. The first unit ran 71 months, almost all of it before February. If the environment changed, the change is somewhere near the boundary, and if the first failure was also caused by the change rather than by age, the change is a little earlier than that.

The tower had a conductivity controller with a blowdown valve, and a treatment log. Three things were checked in this order, cheapest first:

The conductivity setpoint against the actual readings. The setpoint had not been touched. The recorded readings had. Through the fifth year, tower conductivity sat around 1,600 microsiemens per centimeter. From December of the fifth year onward, the recorded readings sat around 3,200.

The makeup conductivity. Around 400, unchanged, on the same municipal source. So cycles of concentration, which at steady state is the ratio of a conservative parameter in the system water to the same parameter in the makeup, had gone from about 4 to about 8. That ratio holds where conductivity is behaving conservatively, meaning nothing is removing dissolved solids except blowdown, which is a reasonable read on this system because there was no visible scaling to drop salts out of solution.

The blowdown valve. Partially blocked. It was opening on command and passing a fraction of its rated flow, so the controller believed it was blowing down and the tower was concentrating anyway.

Everything non-volatile in the makeup water had therefore roughly doubled in the tower loop, chloride included, starting in December of the fifth year. The first exchanger failed two months into that condition after 69 months of normal service. The second exchanger was born into it and lasted nine.

So the first failure was not end of life either. It was the same event, arriving at a unit that had already spent most of its useful thickness. The 71-month life looked like a normal service life and was actually a normal life plus a short exposure to a much harsher one, which is exactly the reading that made a like-for-like replacement look reasonable.

Why the water report did not catch it

The site had a water analysis in the project file from the year the system was built. It showed a chloride level well inside what the plate material tolerates.

It was accurate and it was useless, for a reason that generalizes past this job. A makeup analysis describes the water going in. An evaporative system does not run on its makeup, it runs on its makeup multiplied by the concentration factor it is holding, because water leaves as vapor and the dissolved solids stay. A report that says the chloride in the makeup is acceptable is only an answer at a stated number of cycles, and the number of cycles is set by the blowdown control rather than by the water.

Which means the instrument that would have caught this was not a lab report at all. It was the ratio between two conductivity readings anyone could take with a handheld meter in five minutes, at any point in those eleven months, and the historical record that would have shown when the ratio moved. The site's treatment log recorded doses and readings; it did not record equipment changes or control faults, and there was no entry anywhere marking the day the blowdown valve started passing less. The record needed to answer "when did this start" is a different record from the one that answers "what is the water," which is its own subject and its own card.

The fix, and the second decision it enabled

Two pieces of work, in order.

The blowdown valve was cleared and its actual passed flow verified against the controller's command rather than assumed from the controller's status light. The tower was brought back to its target cycles and held there for a period long enough to confirm on repeat readings, not on one.

Then the alloy question, which was now answerable. With a measured chloride figure in the tower water at the target cycles, and the operating temperature at the plate, the exchanger manufacturer can state whether the standard plate material is inside its chloride limit for that service or whether a more resistant grade is warranted. That is a specification the supplier owns, and it depends on temperature and on the specific grade, so it is asked rather than assumed. The important part is that the question now has real inputs. In February it did not.

The customer got one more thing in writing: the tower concentration ratio was added to the routine visit as two conductivity readings and a computed ratio, with a defined action if the ratio exceeds target. That is a reading that takes minutes and it is the one that would have turned an eleven-month blind spot into a same-week finding.

How they confirmed it

Confirmation was designed as a prediction rather than an inspection, because an inspection of a new exchanger tells you nothing for years.

The prediction: with cycles held at target, tower conductivity should sit near its previous long-run value rather than near double it, and it should hold there across a full seasonal swing including the highest evaporation weeks of summer, when concentration pressure is greatest. If the blowdown valve was the whole story, that holds. If something else is also concentrating the loop, summer is when it shows, because that is the end of the range where evaporation is highest. In the mild shoulder months, a partly disabled blowdown can look adequate simply because there is less to remove.

They logged the ratio at every visit through the following summer. It held at target. The exchanger was inspected at the end of that season and showed no new pitting.

The portable finding is the one that did not require any of this site's specifics: when a correct repair recurs on a much shorter interval, compute the ratio between the two intervals before doing anything else. It costs nothing, it distinguishes drift from a step change, and a step change has a date you can go find.

References

• 29 CFR 1910.147, control of hazardous energy, for isolation and depressurization before opening a heat exchanger
• 29 CFR 1910.134, respiratory protection program requirements, where work generates aerosol from open evaporative water
• Heat exchanger manufacturer documentation for plate material chloride limits at operating temperature
• See related: Why Makeup Water Is the Variable That Decides Everything; Pitting, Crevice and Uniform Corrosion; What an Evaporative Tower Does to the Water in It