Pump Reliability & Predictive Maintenance (PlusSmart) · 11 min read · Updated November 2026
Mechanical Seal Failure RCA: Reading the Seal, Finding the Cause, Fixing the System
Most seals do not fail because they are bad seals. They fail because the pump ran dry, dead-headed, cavitated, vibrated, or pumped something the seal was never chosen for. A root-cause method that starts with the failed part in your hand and ends with a change to the system — with the evidence table we use on every seal we are asked to replace.
A mechanical seal is two lapped faces — one rotating with the shaft, one stationary — pressed together by springs or a bellows and separated by a fluid film a few micrometres thick. The film lubricates and seals at the same time. Anything that removes it (dry running), boils it (flashing at the faces), contaminates it (solids), or disturbs the faces (shaft deflection, vibration, misalignment) destroys the seal. So when a seal fails in months instead of years, the question is never "which seal brand is better" but "what removed the film". The seal itself usually tells you.
Step one: read the evidence
| What you see on the failed seal | Most likely cause | Confirm with | Remedy |
|---|---|---|---|
| Faces heat-checked (fine radial cracks), burnt, blued; elastomers hardened | Dry running or flashing at the faces — dead-head, loss of prime, hot fluid near vapour pressure | Drive/DCS trends: power drop at speed, zero flow periods; seal-chamber temperature | Dry-run and dead-head protection in the control; Plan 21/23 cooling or Plan 32 flush; FlexICE DLC faces that tolerate transient dry running |
| Wear track uneven, off-centre or wider on one side | Shaft deflection or runout, misalignment, pipe strain, operation far from BEP | Vibration spectrum (1×, 2×, axial), dial-indicator runout, flow vs BEP | Align, relieve pipe strain, return pump to 70–120 % BEP, check bearings |
| Faces scored, grooved, eroded | Solids between the faces — slurry, sand, scale, product crystals | Fluid analysis, flush flow check | Hard face pairs (SiC/SiC, TC), Plan 32 clean flush, dual seal with barrier fluid |
| Crystals or hard deposits on the atmospheric side | Product crystallises or coke-forms when the normal weep dries | Fluid properties | Plan 62 quench (steam/water), dual seal Plan 52/53 |
| O-rings swollen, cracked, extruded | Elastomer incompatible with fluid or temperature; pressure beyond rating | Fluid composition, temperature logs | Change elastomer (EPDM, FKM, FFKM) or use metal bellows with no dynamic elastomer |
| Springs clogged, rotating face stuck | Solids packing in the spring area; wrong seal design for the fluid | Inspection | Stationary-spring design, bellows seal, flush |
| Faces chipped at the edges | Mechanical shock — water hammer, cavitation, slugging | Pressure trends, vibration broadband | Ramp control, check-valve review, NPSH correction |
| Carbon face blistered | Oil or viscous fluid absorbed into carbon then forced out by heat | Fluid type | Non-carbon face pair (SiC/TC), lower face load |
Step two: correlate with operating data
The seal evidence gives a hypothesis; the data confirms it. For a pump on a VFD, export speed, current, torque and power for the weeks before the failure; for any pump, export discharge and suction pressure, flow if available, and the start/stop log. What you are looking for:
- Dead-heading: at steady speed, power falls to the shut-off value and head rises; often coincides with a valve sequence or a level control holding the discharge closed.
- Dry running: torque and power collapse while the pump runs; often at batch end, sump drawdown or loss of prime.
- Run-out / cavitation: power above the curve, broadband vibration, "gravel" noise; often when suction level is low or a parallel pump stops.
- Chronic off-BEP: most hours spent below 70 % or above 120 % of BEP flow — the pump is simply the wrong size for how it is used now.
- Starts per hour and ramps: frequent starts with fast ramps produce pressure transients that chip faces and loosen pipework.
PlusSmart Analytics does this correlation across a fleet from historian and drive data; for a single pump an engineer with the trends and the seal in hand reaches the conclusion in an hour.
Step three: fix the system, then the seal
The order matters. Changing the seal without changing the cause buys a few months. The system fixes, in order of frequency we see them:
- Protection in the control. Dead-head, dry-run and run-out protection derived from drive data (ReliaPumpIQ) — no flow meter needed — interlocked with sequences so a valve cannot close on a running pump. This single step is why an iron-ore port went from continuous seal failures to none since 2015.
- Operating point. Trim the impeller, change the speed range, or balance parallel pumps by flow so each stays in its preferred region.
- Alignment and pipe strain. Laser alignment after every intervention; vibration trending (i-ALERT) to catch drift.
- Seal support plan. Verify the API 682 plan is right for the fluid and that it actually works (flush flow present, reservoir level maintained, cooler not fouled).
- Seal selection. Only now: face materials for the fluid and solids, elastomers for chemistry and temperature, bellows for hot duties, split cartridge designs where in-place replacement saves a crane and a day.
Documenting so it does not come back
Every seal failure should leave three records: the photographs of the faces and elastomers, the trend export for the preceding two weeks, and the chosen cause and action in the CMMS work order — "seal replaced" alone is useless. Within a year the plant has a seal standard by duty and a list of pumps whose seals never fail any more. We review failed seals and the operating data at no charge for pumps that fail repeatedly; send the photographs and the trends to PlusSmart Vietnam.
Need engineering help with this?
Our engineers apply this to real projects every week — selection, specification, commissioning. Send your question or a part number.
Quick answers
What information do you need to select a replacement seal?
Pump model and shaft diameter, fluid with composition and solids, temperature, seal-chamber pressure, speed, the current seal type and arrangement, and photographs of the failed seal with a short description of how it failed.
Will a better seal survive dry running?
Better faces survive longer — FlexICE diamond-like-carbon faces tolerate transient dry running and poor lubrication far better than plain carbon/SiC — but no seal survives sustained dry running. Fix the cause with dry-run protection and use the better face as the margin.
Why do parallel pumps fail seals on only one pump?
Equal-speed control makes the pump with the lower curve (the worn one) carry less flow and drift towards dead-head while the newer pump does the work; its seal overheats while the HMI shows both running normally. Flow balancing corrects it.