Follow the water. It leaves the circulating pump, travels carbon steel headers and branch lines, passes the fin-fan bundle where the air side rejects the heat, and returns through the generator cooler where copper tubes sit inches from the stator. Every metre of that path is a different metal in the same electrolyte. Water flushing moved the loose product and left the tubercles, because flushing addresses velocity, not chemistry.
Where does the rust come from, and why did the water flush leave it?
Carbon steel in oxygenated, untreated water forms a porous ferric oxide layer that grows into tubercles: hard, domed deposits with an anaerobic, acidic, chloride-concentrated cell underneath. The tubercle is not the corrosion, it is the roof over it. Pitting continues under the cap at a rate the bulk water chemistry never predicts, which is why 30-year-old service water piping fails as through-wall pits rather than general wall loss.
Water flushing shears off what is already loose. It cannot dissolve magnetite or hematite, and the shear stress at normal loop velocities is nowhere near enough to break a mature tubercle free. Worse, a plain flush that does dislodge fragments sends them straight into the smallest passages on the loop, which are the generator cooler tubes.
The second mechanism is the metal couple itself. Carbon steel is anodic to copper and to 90/10 Cu-Ni. Wherever steel and copper share metallic continuity and a common water path, the steel carries the corrosion current. There is a subtler version that survives even a perfect dielectric break: copper dissolved anywhere upstream plates out on bare steel as metallic copper, and each deposit becomes a microscopic cathode driving a pit. Any cleaning chemistry that strips copper without complexing it will reliably seed this in the steel piping.
Check before going further: pull a spool piece or open a flanged joint on the worst-served branch and photograph the ID. Tubercles with black, wet, foul-smelling product underneath mean under-deposit and possibly microbiological attack, and the loop needs a biocide step, not just an acid.
What metals are actually on the loop?
Nothing gets specified until the metallurgy list is complete, including the parts nobody thinks about: pump casings and impellers, valve trim, gaskets, instrument tailpieces, strainer baskets, expansion tank internals, and the fin-fan bundle tubes and headers. Fin-fan bundles are supplied in carbon steel, in copper, and in aluminium-finned Cu-Ni; the fin material and the tube material are frequently different alloys.
| Loop component | Typical material | Acid tolerance | What to confirm before cleaning |
|---|---|---|---|
| Interconnecting piping | Carbon steel | Good with inhibited acid | Remaining wall thickness by UT at pits and low points |
| Generator cooler tubes | Copper / Cu-Ni | Limited; needs azole inhibitor | Tube OEM's approved cleaning chemistry and maximum temperature |
| Fin-fan bundle | Varies by bundle | Depends on tube alloy | Nameplate or bundle datasheet, not assumption |
| Pump wetted parts | Cast iron / bronze | Poor for aluminium or galvanised | Any galvanised or aluminium part must be removed or bypassed |
| Gaskets and seals | EPDM / NBR / graphite | Chemistry dependent | Elastomer compatibility with the chosen acid and inhibitor |
Aluminium and galvanised steel anywhere on the circuit rule out most mineral acid cleans outright. If either is present, the loop gets a chelant or an organic acid, or it gets mechanically cleaned.
Check: the material list is complete when every wetted item has an alloy written against it and the copper cooler manufacturer has confirmed an acceptable cleaning agent, concentration, temperature and maximum exposure time in writing.
Can inhibited acid clean carbon steel without attacking the copper?
Yes, with two conditions: the copper is protected by an azole film, and the copper is not in the circuit while the heavy iron load is coming off.
Isolate the generator cooler and jumper it with a temporary spool. The piping carries orders of magnitude more oxide than the cooler does, and it is the piping that will shed particulate. Clean the piping first, dump and rinse, then clean the cooler separately at lower strength on its own circulating rig with a fine-mesh filter, or backflush it and inspect. This single decision removes most of the risk in the job.
| Chemistry | Iron removal | Copper risk | Notes |
|---|---|---|---|
| Citric acid, ammoniated | Good on soft oxide | High | Ammonia attacks copper alloys and can crack brass. Rule it out. |
| Citric acid, non-ammoniated | Moderate | Low with azole | Slower, mild on thinned wall, easy effluent handling |
| EDTA / chelant | Good, holds iron in solution | Low | Near-neutral pH, forgiving on mixed metallurgy, higher cost |
| Inhibited hydrochloric | Aggressive, fast on tubercles | Requires copper complexing agent | Only on carbon-steel-only circuits with verified wall thickness |
| Sulfamic acid | Moderate | Low with azole | Common where chloride must be excluded |
Add tolyltriazole or benzotriazole to any solution that will contact copper. Add a copper-complexing agent to any strong mineral acid so dissolved copper stays in solution instead of plating on steel. Chloride-bearing acids get a hard limit on residual chloride at the rinse stage, because chloride left in a stagnant leg restarts pitting immediately.
Check: before dosing, ultrasonic-thickness the pipe at elbows, low points and every known repair patch. Tubercles are load-bearing on badly pitted pipe. Removing them will open leaks that the deposit was plugging, and that is normal, not a failure of the procedure. Budget for pipe replacement.
How do you run the circulation clean?
- Drain the loop, remove or bypass the generator cooler, any aluminium or galvanised component, control valves with soft trim, and instruments with dead legs. Fit the temporary spool.
- Fit a temporary bag or cartridge filter and a temporary strainer at the pump suction. Fit sample points on the return leg.
- Water-rinse until the effluent runs clear, to remove what the earlier flush has already loosened.
- If deposits were wet and odorous, run a biocide and disperse step first, then dump and rinse.
- Fill with the selected chemistry at the supplier's concentration and temperature, with inhibitor and, where required, copper complexing agent. Circulate at a velocity high enough to keep solids entrained through horizontal runs, not just enough to register flow.
- Stop when iron concentration plateaus across two consecutive samples. Do not run to a clock time.
- Dump hot, then rinse until the drain water is clear and pH is back near the makeup water value.
Vent every high point continuously. Air pockets create a stagnant zone that gets neither cleaned nor rinsed and becomes the first pit after commissioning.
Check: the clean is finished when the iron curve is flat, the rinse water is clear, and a borescope or a reopened spool shows grey metal instead of orange product.
What goes back in the loop?
Acid-cleaned carbon steel is bare and highly reactive. It will flash-rust within hours of contact with aerated water, so neutralisation and passivation follow the rinse immediately, with no overnight standing full of raw water.
Neutralise with an alkaline solution to raise pH, then passivate. A fin-fan cooled circuit is closed, not evaporative, so the water inventory is small and treating it is cheap. That is the difference between this loop and an open service water system that reforms rust no matter what is done to it.
| Parameter | Target for a closed steel/copper loop | Reason |
|---|---|---|
| Makeup water | Demineralised or condensate | Low chloride and low hardness; no scale, no pitting driver |
| pH | Mildly alkaline, capped to protect copper | Steel passivates high, copper degrades if pushed too far |
| Ferrous inhibitor | Nitrite-borate or molybdate | Maintains the passive film on carbon steel |
| Copper inhibitor | Tolyltriazole / benzotriazole | Films the copper and stops copper release and replating |
| Ammonia and amines | Excluded | Stress corrosion cracking of copper alloys |
| Dissolved oxygen | Minimised; tight expansion tank | Oxygen ingress is what restarts the whole cycle |
| Glycol | Only for freeze protection, industrial grade | Reduces heat transfer and raises pumping load; automotive silicate packages can drop out and foul |
Break the metallic continuity where steel joins copper. Dielectric unions or isolating flange kits at each transition remove the direct galvanic path. They do not remove the electrolyte path, which is why the azole in the water still matters.
Find and eliminate the oxygen ingress before refilling: a partially drained expansion tank, an air-permeable hose, a leaking pump seal drawing air on suction, or a vent left cracked open. A closed loop that keeps consuming inhibitor is a closed loop that is not closed.
Check:Inhibitor residual within range and pH stable means the passivation took.
Does the cleaning hold?
Baseline the loop the day it returns to service and compare against it, not against memory.
- Record generator cooler water-side differential pressure and cooling water flow at a known load. A rising differential at constant flow is fouling returning.
- Record the approach temperature across the generator cooler: cooling water inlet against hot gas or hot air outlet at a fixed load. Approach creep is the earliest fouling indicator, ahead of any alarm.
- Record the same two numbers across the fin-fan bundle, with ambient air temperature logged alongside so the comparison is valid.
- Trend total and dissolved iron monthly. A flat, low iron trend proves the passive film is intact; a climbing trend means active corrosion somewhere on the circuit.
- Install corrosion coupons, one carbon steel and one copper, in a flowing return leg. Pull and weigh them after the first 30 to 90 days.
- Keep the temporary filter in circuit through the first weeks and inspect the element. Orange fines mean the passivation is incomplete; black magnetite fines settling out is normal shakedown.
The job is proven when the coupon weight loss on both metals sits at the low rates your water treatment supplier specifies for a closed inhibited loop, and the cooler approach temperature at reference load is unchanged from the day of restart.
Frequently Asked Questions
Can I acid clean the carbon steel piping with the generator cooler still in circuit?
No. Isolate the cooler and jumper it with a temporary spool. The piping releases the bulk of the iron oxide and the cooler tubes are the finest passages on the loop, so any dislodged solids collect there. Clean the cooler separately, at lower strength, with the tube manufacturer's approved chemistry.
Does a dielectric union stop galvanic corrosion between the steel pipe and copper tubes?
It removes the metallic path, which is the dominant driver, but the water remains an electrolyte. Fit the isolating unions or flange kits at every steel-to-copper transition and keep a tolyltriazole or benzotriazole residual in the water to film the copper and stop copper ions plating onto bare steel.
Can I use automotive radiator cleaner and 50% coolant on a generator cooling loop?
Only as a stopgap on a small circuit with known metallurgy. Automotive packages are formulated for aluminium and cast iron, some carry silicates that drop out and foul heat transfer surfaces, and 50% glycol reduces heat transfer and raises pumping load. Use industrial inhibited glycol, and only where freeze protection is actually needed.
Will removing the tubercles cause the old pipe to leak?
On piping that has already had through-wall pits, expect it. The deposit is plugging thinned sections. Ultrasonic-thickness the elbows, low points and previous repair areas before dosing any acid, and have replacement spools and welders available during the clean.
Does the rust come back if I clean but do not treat the water?
Yes, and quickly. Freshly acid-cleaned carbon steel flash-rusts in aerated water within hours. Neutralise, passivate and refill with demineralised water carrying a ferrous inhibitor plus an azole for the copper, then close off the oxygen ingress that let the film break down originally.