How to Tell Erosion From Corrosion

Why this matters

Metal disappearing from a component has two fundamentally different explanations. Corrosion is chemistry: the metal reacts and leaves a product behind. Erosion is mechanics: fluid, particles, or collapsing vapor bubbles knock metal away and leave nothing behind but clean surface. They look similar in a photograph and they respond to completely opposite corrections. Answer corrosion when the truth is erosion and you will specify a more resistant alloy, wait for it, pay a harder part to do the same job, and watch it fail on nearly the same schedule. Answer erosion when the truth is corrosion and you will slow the flow down on a system that had nothing wrong with its flow.

The discrimination is not difficult, but it depends on observations that are destroyed by cleaning the part and by taking it out of the system without noting which way the fluid was going.

Step 0: Before you get close to it

Isolate and depressurize before you open anything, and verify at a gauge or an open bleed that the pressure is actually gone. A component thinned by either mechanism has less wall than its nameplate suggests, and it fails without warning when disturbed. That isolate, lock, tag and verify sequence is what 29 CFR 1910.147 requires for stored fluid and mechanical energy.

If part of your diagnosis is listening to a running machine, protect your hearing while you do it. Cavitation and high-velocity flow are diagnosed partly by sound, and the sound is loud. OSHA's occupational noise standard, 29 CFR 1910.95, sets the action level for a hearing conservation program at an 8-hour time-weighted average of 85 dBA and the permissible exposure limit at 90 dBA, and standing beside a cavitating valve is comfortably in that territory even if your total exposure for the day is not.

Step 1: Fix the flow direction and the energy source before removing anything

Mark the flow direction on the component with a marker or a piece of tape before it comes out, and note what is immediately upstream: a bend, a valve, a pump, a reducer, an orifice, a burr at a cut. Note whether the system runs continuously or intermittently.

Everything in the following steps depends on relating damage position to flow, and once the part is on a bench there is no flow direction any more. This is the observation people most often lose and the one they most often need.

Step 2: Read the surface texture, before cleaning

This single observation separates the two families most of the time.

• Corrosion leaves a product. Oxide, scale, crust, powder, sludge, tubercle. Even when the product has washed away, the underlying metal is dull, etched, or granular.
• Erosion leaves bare metal. Polished, scoured, sometimes with a satin sheen, and conspicuously clean compared to its surroundings. In a system where everything else has a film, the eroded area is the shiny part.
• Cavitation leaves a spongy, honeycombed, pinholed texture. It looks like the metal was eaten from inside by something with very small teeth, and the edges of the damaged zone are sharp rather than blended.

Wash the part before you look and you lose the product, which is the corrosion evidence, and you polish the surroundings, which is the contrast the erosion evidence depends on.

Step 3: Locate the damage relative to the flow

• Erosion sits at impingement points: the outer radius of a bend, the first inches downstream of a restriction, the leading edge of a vane or a blade, the point where a jet lands, and immediately past any burr or protruding gasket that trips the flow.
• Cavitation sits just downstream of a pressure drop, in the recovery zone rather than at the restriction itself, because that is where the vapor bubbles collapse. Damage on the seat is throttling wear; damage on the body or the pipe an inch or two past the seat is collapse.
• Corrosion follows chemistry and gravity, not geometry. It concentrates where fluid sits, where two metals meet, under deposits, and at the bottom of horizontal runs, and it is entirely indifferent to whether a bend is nearby.

A damage pattern that maps neatly onto flow geometry is mechanical. A damage pattern that ignores flow geometry is chemical.

Step 4: Look for direction in the damage itself

Erosion is directional and leaves that direction in the metal: grooves, scallops, wave patterns, and horseshoe-shaped pits with the open end pointing downstream. Run a fingernail across the damage. Directional texture means something was moving.

Corrosion pits are not directional. A round pit with an undercut mouth is chemical. A crescent with a tail is not.

Cavitation is the exception that proves the value of step 3: its damage is not directional either, which is why position matters so much for identifying it.

Step 5: Ask what is in the fluid

Solids turn a marginal velocity into an erosive one. Sand or grit after a main break, weld slag or cutting swarf left at commissioning, scale spalling off upstream, filter media that escaped, or a media bed that broke through. If a system that ran clean for years began eroding, something changed and there is a good chance it is now carrying solids.

Air is the other one. Entrained air is a two-phase flow and it erodes far more aggressively than the same velocity of liquid alone, so a system that started drawing air at the pump suction or through a failed vent gets erosion that its design velocity does not explain.

Step 6: Check the operating point against the design

Get the numbers rather than the impressions. Flow rate and pipe size give velocity. Inlet and outlet pressure at a valve give the drop across it. Suction pressure and fluid temperature give whether a pump has margin above the fluid's vapor pressure. Every mechanical mechanism above is an operating-point problem, so if the operating point matches the design and has never changed, look harder at chemistry.

Step 7: Change one variable and watch

The confirming test for a mechanical mechanism is that it responds to a mechanical change immediately. Reduce the flow, open the throttled valve and take the drop somewhere else, restore the suction condition, or filter the fluid, then re-inspect on a short interval. A chemical mechanism does not care.

The mirror is equally decisive and is the trap in the worked case below: a chemical mechanism responds to a material upgrade, and a mechanical one barely does.

The upgrade that proved the diagnosis wrong

A building's pumped loop keeps destroying the trim in one control valve. The shop's first read is corrosion: the damaged surface is rough and pitted, the water is warm, and corrosion is the familiar answer. They specify a harder, more corrosion-resistant trim, wait out the lead time, and install it.

The result that reframed the problem. The original trim was lasting about 9 months. The upgraded trim lasted about 11 months, an improvement of roughly 1.2 times. If chemistry had been removing the metal, a genuinely more resistant alloy should have changed the interval far more than that. A small improvement from a much harder material is the signature of a mechanical mechanism: hardness slows erosion a little, and resistance to a chemical attack that is not happening does nothing at all.

What the second inspection found, done properly. Flow direction marked before removal. The seat itself showed modest wear. The severe damage was in the valve body and the first two inches of pipe downstream of the seat, spongy and honeycombed with sharp edges, and the surrounding metal was bright and clean rather than filmed. Nothing about the pattern followed gravity or joint geometry; all of it sat in the pressure recovery zone.

The operating point. The valve was throttled hard to balance the loop: inlet 65 psig, outlet 15 psig, a drop of 50 psi across one seat, on water at 140 F, whose vapor pressure is about 2.9 psia. Expressed as a ratio, the drop of 50 psi against the inlet absolute pressure less the vapor pressure, which is 65 plus 14.7 minus 2.9, or 76.8 psi, gives 50 divided by 76.8, about 0.65. The threshold at which a given valve begins to cavitate depends on its own pressure recovery characteristic, which is in the valve manufacturer's data and is not a number to carry between valve types, but a ratio of two-thirds across a single seat is deep into the region where a recovering valve will flash and collapse.

The correction. Stage the drop instead of concentrating it. Two restrictions in series, each taking 25 psi, run 65 to 40 and then 40 to 15. The first now sees 25 divided by 76.8, about 0.33, roughly half the previous ratio, and the second sees a drop from a lower inlet pressure but with a much smaller step. Alternatives that address the same root cause are raising the downstream pressure, resizing the valve so it is not operating nearly closed, or reducing the pump head so there is less pressure to burn off in the first place. Fitting a still harder trim is the one option that treats the symptom.

What would have changed the conclusion. If the damage had been on the seat faces and the sealing surfaces rather than downstream, this is ordinary throttling wear and the fix is a valve suited to modulating service. If the damaged area had carried a product layer and the surrounding metal had been equally dull, chemistry is back in play and the water needs testing. And if the fluid had been carrying grit, the same downstream position could be plain particle erosion, which is fixed by filtration rather than by staging the pressure drop.

Confirming which one you named

• Compare against a sheltered surface in the same fluid. Every system has metal that sees the fluid but not the flow energy, such as the inside of a dead leg or the back of a flange face. If that metal is attacked too, chemistry is present. If it is pristine while the impingement zone is scoured, the mechanism is mechanical.
• Check whether identical components in the same loop are equally affected. Corrosion attacks the whole wetted system to some degree. Erosion and cavitation attack the specific locations where the energy is, and leave the rest untouched.
• Section the worst area and look at the damage depth profile. Erosion thins from the surface in a smooth contour. Corrosion undercuts and leaves overhanging edges and subsurface cavities.
• Re-inspect on a short interval after a mechanical change, not a long one. Mechanical mechanisms respond within weeks. Waiting a full season to see whether a flow change helped wastes the fastest confirmation available.
• Write the mechanism into the work order, with the operating numbers that support it. "Cavitation, valve dropping 50 psi of 65 psig inlet at 140 F" survives a change of tech. "Valve trim worn" invites the same replacement next year.

References

• 29 CFR 1910.147, control of hazardous energy, for the isolation and verification required before opening a pressurized component
• 29 CFR 1910.95, occupational noise exposure, for the 85 dBA action level and 90 dBA permissible exposure limit that govern working beside noisy equipment
• Valve manufacturer data for the pressure recovery characteristic that sets the cavitation threshold of a specific valve; this is device-specific and is not transferable between valve types
• Pump manufacturer data for required suction head, where cavitation is suspected at the pump rather than at a valve
• See related: How to Read Corrosion and Name Its Cause; Pitting, Crevice and Uniform Corrosion