Your 6 MW back pressure turbine has read 2.9 mm/s for two years. This morning the panel shows 7.4 mm/s and it is still climbing, alarm is set at 7.1 and trip at 11.2, and the shift manager wants to know whether the machine can hold load until the Sunday shutdown forty hours away. That single decision, taken with data or without it, is worth tens of millions of naira in either direction.
Most steam turbine high vibration causes fall into a small number of classes, and each class announces itself differently. Unbalance does not behave like a rub, a rub does not behave like oil whirl, and misalignment looks nothing like either once you have phase information in front of you. The difficulty on most captive plants is not that the fault is exotic. It is that the only number anyone holds is an overall amplitude on a panel, which answers almost none of the questions that matter.
Every figure in this guide is an engineering reading, not a price. Cost depends on scope, outage length and parts lead time, so treat any specific figure as something to confirm in a technical proposal against your machine.
The Number on the Panel Is the Least Useful Piece of Data
Overall vibration amplitude tells you that something changed. It does not tell you what changed, where, or how fast it is getting worse.
Four other things carry the diagnosis:
- Frequency content, meaning how much energy sits at running speed, at twice running speed, at harmonics or below running speed
- Phase, the angular position of the heavy spot against a once-per-revolution reference
- Direction, whether the energy is radial or axial
- Rate of change, which on a turbine often matters more than the absolute level
There is also a measurement question that catches many sites out. Casing seismic pickups read velocity in mm/s on the bearing housing, while proximity probes read shaft relative displacement in micrometres peak to peak inside the bearing. A rotor can be moving a great deal inside a heavy casing that barely registers on a seismic pickup, which is why an overall casing number can look reassuring while a bearing is being destroyed.
Evaluation criteria are set out in ISO 20816-2 for land based steam turbines and generators above 40 MW, and in ISO 20816-3 for the industrial machines that cover most captive units in Nigeria. Both make the point that panel readings obscure: evaluation of change matters as much as evaluation of level. Your OEM alarm and trip settings, fixed against that specific rotor and bearing design, are what actually govern.
If your machine has moved outside its normal band and the only record you hold is a daily logsheet figure, that is where the work starts. Request a technical proposal and an engineer will review the vibration history and the instrumentation you have before anything is quoted.
The Main Steam Turbine High Vibration Causes and How Each One Announces Itself
Unbalance from deposits or blade erosion. Energy concentrated at running speed, phase stable, amplitude rising with the square of speed. The usual driver is deposit build up from poor steam purity, or metal loss from wet steam erosion in the last stages of a condensing machine. It develops over months, which is what separates it from most other causes.
A step change at running speed. A sudden jump rather than a ramp points at material leaving the rotor: a blade tip, a shroud segment, a coupling bolt or a balance weight. This is one of the few signatures that justifies coming off load immediately rather than continuing to observe.
Thermal bow. Running speed energy with phase that wanders as the rotor soaks, present on run up and settling out after slow rolling. It comes from uneven cooling of a stationary hot rotor, and it is nearly always a turning gear discipline problem rather than a rotor problem.
Rub. Running speed energy with unstable phase, plus harmonics at two and three times speed and sometimes sub-harmonics at half or one third of running speed. Amplitude often rises and then falls as the rub wears clearance, which is why a falling number after a spike is not good news.
Misalignment, soft foot, piping strain and foundation defects. Strong second order energy with a real axial component and a phase difference across the coupling. The underrated version on steam turbines is piping strain, where seized spring hangers or a failed anchor let steam line thermal growth push the casing out of position between cold and hot condition. Cracked grout and loose holding down bolts show a similar picture and are common on machines that have been through several bearing jobs.
Bearing and oil film faults. Sub-synchronous energy between roughly 0.38 and 0.48 times running speed is oil whirl, an instability of the fluid film itself. Clearance opened by babbitt wear, oil thinned by dilution or heat, and a lightly loaded journal all push a bearing towards it, as set out in our note on turbine bearing babbitt failure.
Thermal Bow and Rubs: The Causes That Punish a Fast Restart
Thermal bow is the cheapest fault on this list to prevent and one of the most expensive to ignore. When a hot rotor is left stationary the upper half stays hotter than the lower half, the shaft bows, and the next start puts a bent rotor through its critical speeds.
Handling it is a discipline question. Turning gear must run through the entire cooldown, and if it is out of service the machine needs a full cooldown before restart plus a slow roll with a vibration record on the way back up. Cutting that short to recover load is how a recoverable bow becomes a permanent one.
The step from bow to rub is short. Once the bowed rotor contacts a gland or oil seal you get local heating at the contact point, which deepens the bow, which increases the contact. That feedback loop can bend a rotor beyond field correction inside a single start attempt.
Worked example (hypothetical): 1.5 MW back pressure unit. It trips on high vibration on every hot restart within about twenty minutes of shutdown, while cold starts stay clean. The signature is running speed energy with phase that moves through the run up, plus second and third order harmonics above 2,400 rpm.
Suppose the turning gear motor burnt out seven months earlier and was never replaced, so the rotor was left stationary while hot after every stop. Motor repair and control wiring, plus a gland seal strip replacement across a four day outage, is a modest job set against what came next.
The delay is the expensive part. Over two months, lost production hours at the plant’s own contribution margin add up to many times the cost of the repair that would have fixed it when the motor first failed. Putting a defensible figure on that exposure is covered in plant downtime cost per hour.
Bearing and Oil Film Faults: The Steam Turbine Vibration Causes That Build Slowly
Journal bearings on a steam turbine rarely fail suddenly. They lose clearance control, and vibration follows the oil before it follows the metal.
Three site conditions push a healthy bearing towards instability. Viscosity falling, through thermal degradation, condensate dilution or the wrong grade going in at a top up. Water ingress, which on a condensing machine usually traces back to gland sealing or a gland condenser running at poor vacuum. And clearance opening through wear, which cuts the load capacity of the film at the same eccentricity.
Oil whirl and oil whip are worth separating. Whirl sits at a fraction of running speed and tracks with speed as you change it. Whip locks onto the rotor’s first critical frequency and stays there as speed rises, and it is the more dangerous because the rotor is being driven at a resonance.
Worked example (hypothetical): 4.5 MW condensing turbine with a waste heat boiler. It develops a sub-synchronous component at about 0.43 times running speed, roughly six weeks after a lube oil top up during a hurried outage. Overall casing vibration has risen only modestly, from 3.4 to 5.1 mm/s, so the machine is left running.
Oil analysis settles it. Viscosity has dropped about 18 percent below grade and water content is near 900 ppm, traced to gland sealing steam and a gland condenser holding poor vacuum. Two journal bearings are later found with babbitt wiping and clearance well outside drawing tolerance.
Oil replacement and flush, gland condenser correction and re-babbitting of two journals over nine days runs to several times the cost of the routine sample that would have caught the dilution a fortnight in, on the schedule described in turbine lube oil analysis. Bearing work of this kind needs a proper shop, measured clearances and a qualified engineer on site.
When to Trip a Steam Turbine on High Vibration
This is the question plant managers actually need answered, so here is a plain response.
Bring the machine off load without waiting for the trip setpoint when any of these appear:
- A step change in amplitude rather than a ramp
- Vibration rising together with bearing metal temperature
- A new sub-synchronous component
- A change in axial position or differential expansion alongside the vibration
- Audible rubbing or grinding from the casing
Each of these describes damage in progress, and the machine will not improve while you watch it.
You can usually plan a controlled shutdown, rather than tripping, where the rise has been gradual over weeks, the signature is a stable running speed component with steady phase, bearing temperatures are normal and the level is inside the OEM alarm band. That is the deposit unbalance case, and it is generally safe to run to a scheduled outage window with monitoring frequency increased.
One rule has no exception. Do not raise a vibration trip setting to keep a machine on load. Those setpoints are protection values fixed by the OEM against that rotor, its criticals and its bearing design, and moving them converts a bearing repair into a rotor replacement with a lead time measured in months.
Be honest about what cannot be judged from a control room. Deciding whether a rotor with a confirmed rub can run to a planned outage needs run out measurements, bearing inspection and a rotordynamic view of the machine. That is specialist work, it needs an outage window, and no responsible answer to it comes over the phone.
Worked example (hypothetical): 6 MW back pressure unit. It drifts from 2.8 to 6.9 mm/s over eleven months, running speed dominant with stable phase and normal bearing temperatures. Steam sampling finds cation conductivity and sodium well outside limits, with carryover from a boiler drum whose separators were damaged during an earlier tube repair.
In that case you would run on to the planned outage with weekly phase referenced readings rather than tripping. Blade path cleaning and an in situ trim balance across five days bring vibration back down to 3.1 mm/s. The turbine was never the fault, and without correcting the boiler water treatment the deposits would return within a year.
What Diagnosis and Correction Cost
The table maps signature to likely cause class and the first diagnostic step. Cost depends on scope, outage length and parts lead time once the diagnosis is confirmed, and our cost guide sets out how that is built, or request a technical proposal for your machine.
| Vibration signature | Likely cause class | First diagnostic step |
|---|---|---|
| Running speed dominant, stable phase, rise over months | Deposit or erosion unbalance | Steam purity sampling, in situ trim balance |
| Step change at running speed during operation | Lost blade, shroud or coupling component | Controlled shutdown, borescope and run out check |
| Unstable phase on hot restart only, harmonics present | Thermal bow progressing to gland rub | Turning gear check, soak and slow roll record |
| Twice running speed with axial component | Misalignment, soft foot, piping strain or grout failure | Cold and hot laser alignment, hanger and anchor survey |
| Sub-synchronous at 0.38 to 0.48 times speed | Oil whirl from clearance, viscosity or light loading | Oil analysis, bearing clearance measurement |
| Non-synchronous locking at the first critical | Oil whip, bearing instability | Rotordynamic assessment, bearing geometry review |
| Amplitude rise with bearing metal temperature rise | Babbitt distress or oil starvation | Oil analysis, bearing lift and inspection |
Two patterns are worth noticing. The cheapest rows are the housekeeping ones, turning gear discipline, oil condition and alignment, and those are the faults most often left until they become rotor damage. The expensive rows cost what they cost because of the outage window and parts lead time, not the labour.
The first correcting action is usually a monitoring one. Take a phase referenced baseline on every bearing at a defined load and steam condition, then repeat it monthly and log it. Six readings in, the trend answers the run or stop question far better than any single amplitude ever will, at a fraction of the cost of one avoidable shutdown.
If a machine has moved outside its normal band and you need the signature read before the next outage is planned, book a plant assessment and an engineer will read it with you. Correction work sits under steam and gas turbine overhaul, and alignment, balancing and bearing scopes on driven equipment fall under rotating equipment services. We scope the diagnosis first and quote the correction afterwards, in writing, against what the data shows.
Frequently Asked Questions
What vibration level is too high for a steam turbine?
There is no single figure that applies across machines, because the acceptable level depends on rotor design, bearing type, speed and measurement location. ISO 20816-2 and ISO 20816-3 give evaluation zones as general guidance, but your OEM alarm and trip settings are what govern the machine. As a working rule, a doubling from a stable baseline deserves investigation even when the absolute level still sits inside the alarm band.
Can we keep running a turbine that is above its vibration alarm?
Sometimes, and only with the right evidence in hand. A gradual rise with stable phase, normal bearing temperatures and no axial position change can often be run to a planned outage with increased monitoring. A step change, a new sub-synchronous component or rising bearing metal temperature is a different situation and calls for a controlled shutdown rather than continued observation.
Does balancing fix most steam turbine vibration problems?
No, and this is the most common and most expensive mistake we see. Balancing corrects unbalance, so it helps with deposits and blade loss but does nothing for misalignment, rubs, bearing faults, piping strain or foundation defects. Balancing a machine whose real fault is a rub or an unstable bearing wastes the outage and usually leaves the vibration back where it started within weeks.
How long does a proper vibration diagnosis take?
A phase referenced survey on a running machine typically takes one to two days on site, including readings at several loads and a run down record where the machine can be stopped. Oil sampling and analysis add roughly a week for laboratory turnaround. Where the survey points to bearing or rotor damage, confirming it needs an outage window and casing opening, which is planned work rather than something completed during a site visit.