How to Read Corrosion and Name Its Cause
Why this matters
There are perhaps a dozen corrosion mechanisms a field-service tech meets, each with a different corrective action, and only one of them is fixed by replacing the corroded part with the same part. Get the mechanism wrong and the repair is a subscription: the shop returns on the same interval forever, the customer stops believing the diagnosis, and nobody ever writes down enough for the next tech to do better.
What follows is built around the record you leave behind. The observations are worth making because they discriminate between mechanisms, and they are worth writing down because the evidence is gone the moment the part is cut out. Fill the record on site, in this order, and the mechanism usually names itself from the completed fields.
Before you touch the corrosion
Do not probe, scrape, or tap a corroded pressure boundary while it is under pressure. Probing is how a weep becomes a jet, and a pitted or graphitized wall can be a small fraction of its nominal thickness with no outward sign. Isolate the system, relieve the pressure, and verify at a gauge or an open bleed point before any tool touches the metal. That isolate-lock-tag-verify sequence is what 29 CFR 1910.147 requires for stored mechanical and fluid energy. For corrosion inside electrical equipment, the controlling standard is different: de-energize under 29 CFR 1910.333(b)(2) and prove dead live-dead-live (NFPA 70E-2021, 120.5) before the enclosure is opened.
Know the coating before you grind or heat anything. Cadmium plating and chromate conversion coatings release toxic fumes when ground or torched, and lead solder does the same. Where the coating is unknown on older equipment, treat it as hazardous: no torch, and mechanical removal only with dust capture and respiratory protection selected and fit tested under 29 CFR 1910.134.
The record: nine fields, filled on site
Write these before anything is cleaned. Each one exists because it eliminates mechanisms.
1. Location and orientation. Where on the component, and which way does it face? Note gravity (top, bottom, side), flow direction if there is flow, and what is above it. A pattern that follows gravity is a wet-surface problem. A pattern that follows flow is a velocity problem. A pattern that follows neither is chemical or galvanic.
2. Metals present, including neighbors. Name every metal within the wetted or damp zone, including fasteners, brackets, hangers, and the fitting on the other side of the joint. A single metal cannot corrode galvanically. Two dissimilar metals plus an electrolyte can, and the one that suffers is the less noble of the pair.
3. Geometry of the attack. General thinning across a broad area, discrete pits, a groove, a crack, or a change of color with no dimension loss. This field alone eliminates more mechanisms than any other.
4. Product: color, texture, adherence. Flaky orange, white powder, green crust, black sludge, a hard tubercle, or nothing at all. Note whether it wipes off or is bonded.
5. Distribution. Continuous, banded at a height, confined to a joint or under a gasket, downstream of a fitting, or a single isolated point. Distribution is what turns one observation into a pattern.
6. Electrolyte source and its chemistry, if known. Process water, condensate, rain, wash-down, or humidity. Where an analysis exists, capture chlorides, pH, and hardness rather than the whole report.
7. Temperature at the location, measured. Not setpoint, not nameplate. Several mechanisms switch on with heat.
8. Flow and velocity, if there is flow. Pipe size, measured or design flow, and any restriction, pump, or partly closed valve upstream.
9. Age, history, and what was done last time. Installed when, failed when, what was replaced, and what changed about the system just before the failures started.
How the fields name the mechanism
| Mechanism | The fields that identify it | The corrective action |
|---|---|---|
| Uniform / general | General thinning, one metal, continuous distribution, product covers everything | Coating, material upgrade, or accept and schedule replacement |
| Galvanic | Two dissimilar metals in contact, damage concentrated on the less noble one at the joint, clean elsewhere | Isolate the metals (dielectric fitting, isolating washer) or change one of them |
| Pitting | Discrete pits, clean metal between, chlorides present, often stainless or aluminum | Change alloy, remove the chloride source, control deposits |
| Crevice / under-deposit | Damage confined under a gasket, washer, deposit, or lap, none on open surface | Eliminate the crevice or the deposit; cleaning alone recurs |
| Erosion-corrosion | Grooves or horseshoe pits pointing downstream, just past an elbow, valve, or burr; high velocity | Reduce velocity, upsize, deburr, remove the restriction |
| Cavitation | Spongy honeycombed area downstream of a pressure drop, often at a pump or a throttled valve | Fix the suction condition or the throttling, not the metal |
| Microbiologically influenced | Discrete pits under hard tubercles or black slimy deposits, sometimes a sulfide smell, stagnant or intermittent service | Treat the water, restore flow, eliminate dead legs |
| Dealloying | Colour change with shape retained: brass turned pink and porous, cast iron soft and grey | Replace with a resistant alloy; the affected part has no strength left |
| Stress corrosion cracking | Fine branching cracks, little product, tensile stress plus chlorides plus heat | Remove one leg of the trio; treat as replacement |
| Concentration cell | Attack at the wet edge of a partly wetted surface, or under a wet insulation lap | Eliminate the trapped moisture path |
| Stray current | Rapid localized attack near a grounding or bonding fault, out of proportion to the environment | Trace and correct the electrical fault first |
Two of those are worth a caution because they get confused. Pitting and crevice corrosion look identical once you clean the part, and the discriminator is field 5: crevice attack stops abruptly at the edge of the gasket, washer, or deposit that created it, while pitting scatters across an open surface. Clean the part before you record the distribution and you have destroyed the distinction.
A completed record
Third pinhole leak in 14 months on a hot water recirculation return in a small commercial building. Two previous repairs were pipe-section replacements with identical tube.
- 1. Location and orientation: outer radius of a horizontal elbow and the first few inches of tube downstream of it. Nothing on the upstream side.
- 2. Metals present: copper tube, copper fittings, brass valve bodies elsewhere in the loop, steel hangers with isolators intact. No dissimilar-metal contact at the leak.
- 3. Geometry: shallow horseshoe-shaped pits with the open end of each horseshoe facing downstream, plus a general polished groove. Not discrete round pits.
- 4. Product: essentially none inside. The interior is bright and scoured. Small green crust outside at the leak only, which is the water that came out.
- 5. Distribution: confined to the downstream side of directional changes throughout the return leg. Straight runs are clean.
- 6. Electrolyte: domestic potable water, no unusual chemistry reported.
- 7. Temperature: measured 143 F at the return.
- 8. Flow: 3/4 in type L copper, inside diameter 0.785 in, measured 8 gpm through the return.
- 9. History: leaks began after the recirculation pump was replaced with a larger one during an unrelated repair.
The velocity calculation. The tube's inside area is 0.785 in diameter, so 0.484 square inches, which is 0.00336 square feet. At 8 gpm, the volumetric flow is 0.0178 cubic feet per second, and dividing gives a velocity of about 5.3 ft per second. Commonly cited design limits for copper tube are in the region of 5 to 8 ft per second for cold water and drop sharply for hot water, often quoted around 2 to 3 ft per second above about 140 F, because heat accelerates the same attack. Those limits belong to the tube manufacturer and the plumbing code in force, and they are the values to cite in a report, but the shape of the guidance is not in dispute: a hot recirculation return at 5.3 ft per second is roughly double the commonly quoted ceiling for its temperature.
The verdict. Fields 3, 5 and 8 together name erosion-corrosion, and nothing else fits. Pitting from chlorides would scatter across straight runs and would not orient to flow. Galvanic attack needs a dissimilar-metal pair that field 2 says is not there. Microbiological attack leaves deposits and this surface has none.
The correction, with the arithmetic. Getting velocity to 2.5 ft per second in the existing tube means reducing flow to 2.5 divided by 5.3, times 8 gpm, which is about 3.8 gpm, a reduction of roughly 53 percent. That is a balancing valve or a pump speed change, not a pipe job. Upsizing to 1 in type L, inside diameter 1.025 in, gives 0.825 square inches or 0.00573 square feet, and the same 8 gpm then runs at about 3.1 ft per second. Worth noting plainly rather than letting it pass: 3.1 is still above the 2 to 3 band, so upsizing one size on its own does not clear the guidance at this flow. Balancing the flow does, and doing both is belt and braces.
What it cost to keep replacing tube. Three repairs at roughly 2.0 hours each is about 6.0 hours of shop time, all of it recovery work on a system that was never going to stop leaking, plus whatever the water reached. Rebalancing the loop is well under an hour. Both figures are the same currency, technician hours on site, and the ratio is on the order of 12 to 1 in favor of fixing the velocity.
What would have changed the answer. If the pits had been round and scattered rather than horseshoe-shaped and directional, the record would point at water chemistry and the pump would be innocent. If field 9 had shown no pump change and the leaks had been there since installation, the likely cause moves to undeburred tube ends at the fittings, which produces the same mechanism from a different origin and is fixed at the joint rather than at the pump. And if the loop had run at ambient temperature, 5.3 ft per second sits inside the commonly cited cold-water range and the whole finding weakens.
Making the record hold up
- Photograph the interior surface before you cut, with a scale and an arrow for flow direction. A photo of a cut pipe with no flow reference cannot be re-read by anyone, including you, in six months.
- Keep the failed section, labeled with flow direction and date. If the next failure looks different you need the first to compare against, and laboratory examination is only possible on a piece that still exists.
- Sanity check your mechanism against field 9. If your named cause has been present since day one and the failures have not, something changed and you have not found it yet.
- State a confidence level in writing. "Erosion-corrosion, high confidence, based on pit orientation and measured velocity" is defensible. "Corrosion" is not, and the difference matters when the customer asks why the last two repairs did not hold.
- Name the correction in terms of the mechanism, not the part. A work order that says "replace section" invites the same repair next time.
References
- 29 CFR 1910.147, control of hazardous energy, for isolating and verifying stored fluid or mechanical energy before probing a corroded pressure boundary
- 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5, for de-energizing and proving dead before opening electrical equipment to inspect corroded terminations
- 29 CFR 1910.134, respiratory protection, for the selection and fit testing required when mechanically removing an unknown coating
- Copper tube manufacturer documentation and the plumbing code in force for the design velocity limits that apply to a specific installation
- See related: Reading Rust and Corrosion Patterns; How Each Common Metal Tells You It Is Failing; How to Tell Erosion From Corrosion