Steam Turbine Failures: Seven Ways a Turbine Dies and How to Catch Each One Early

A steam turbine is one of the most durable machines industry has ever built. Rotors run for thirty or forty years. Casings outlive the engineers who bolted them down. And yet turbines are wrecked every year, in power stations and sugar mills and refineries, by failures that left a paper trail for months before the event.

That is the frustrating thing about steam turbine failures: very few of them are genuinely sudden. A blade that liberates at 3,000 rpm spent weeks or months growing the crack that released it. A bearing that wipes had contaminated oil long before the whitemetal let go. A rotor destroyed by water induction was fed by a drain line that had been holding condensate for years. In almost every post-mortem we have been part of, somebody had the evidence in hand and did not know what they were looking at.

This article is the field guide. It covers the seven failure modes that account for the overwhelming majority of turbine damage, and for each one it gives the mechanism, the early warning signs, and what actually prevents it. It is the companion piece to our steam turbine overhaul guide, which covers what happens once the machine is opened. This one is about what puts it on the crane in the first place.

Blade Failures: Fatigue, Erosion and the Things Steam Carries

Blades are the most highly stressed components in the machine and the most common source of forced outages. They fail in four distinct ways, and telling them apart matters because the fixes are completely different.

High-cycle fatigue is the classic. Every time a blade passes a nozzle it receives a small aerodynamic pulse, tens of thousands of times a minute. If a blade row's natural frequency sits close to one of those excitation frequencies, the resonant stress grows until a crack initiates, usually at the root fixing, a lacing wire hole or a tenon. The crack then propagates with each cycle until the remaining section fails in overload, and a released blade at speed takes its neighbours with it. The insidious part is that a well-designed machine only becomes resonant when something changes: a repair blade with slightly different mass, deposits that alter stiffness, or sustained operation at an off-design speed.

Water droplet erosion lives in the last low-pressure stages, where the steam is 8 to 12 percent wet and blade tip speeds approach 500 metres per second. Coarse water droplets, too heavy to follow the steam path, strike the leading edges of the rotating blades and progressively chew away the inlet edge. It is a slow, predictable wastage, which is why manufacturers fit stellite shields to the leading edges. It becomes a failure when erosion is left to undercut the shield or when prolonged low-load operation, which pushes the wetness up, accelerates it beyond the inspection interval.

Solid particle erosion works from the other end of the machine. Hard magnetite scale exfoliates from superheater and reheater tubing, mostly during starts, and arrives at the HP and IP first stages as high-velocity grit. It blunts nozzle trailing edges and rounds blade profiles, costing efficiency long before it threatens integrity. A machine that loses heat rate after every outage, without any deposit chemistry to explain it, is usually eating its own boiler scale.

Deposits change blade natural frequencies, unbalance rotors, and roughen aerodynamic surfaces. They deserve their own section, because they are not really a turbine problem at all.

Early warnings for the blade family: a step change or new peak in the vibration spectrum, particularly at blade pass frequencies; stage efficiency drift; and, for LP erosion, visual evidence at every opportunity the condenser is open. Prevention is equally specific: keep blade repairs to like-for-like mass and profile, control low-load hours on the LP end, manage boiler-side chemistry to limit exfoliation, and treat every liberated fragment found in a drain or strainer as a mandatory investigation, not a curiosity.

Four distinct blade failure mechanisms that get lumped together as blade damage. Each has its own location, cause and prevention. Source: MIMAH engineering analysis; ASME and industry practice.

Steam Path Deposits: A Chemistry Problem Wearing a Turbine Costume

When a turbine is opened and the blading is coated in white or grey scale, the instinct is to blame the machine. The machine is innocent. Deposits are the boiler and feedwater chemistry arriving at the one place in the cycle where pressure and temperature fall far enough for dissolved impurities to come out of solution.

The mechanism is straightforward. Impurities in the steam, silica and sodium salts chief among them, are carried in solution at boiler pressure. As steam expands through the machine its capacity to hold them collapses, and each species precipitates at the stage where its solubility runs out. Silica characteristically plates out across the intermediate and LP stages. Copper, picked up from copper-alloy feedwater heater tubing, travels at high load and deposits in the HP section, where even thin films throttle the first stages and can cost a large unit several megawatts of capacity.

The numbers that keep a steam path clean are brutally small. IAPWS guidance on steam purity for turbine operation puts normal operating values in the parts-per-billion range: silica held to around 10 ppb in steam, sodium and chloride each to around 2 ppb. The companion IAPWS guidance on volatile treatment of the steam-water circuit sets the feedwater and boiler water regimes that make those steam values achievable. If your chemistry programme cannot demonstrate those numbers continuously, the turbine is accumulating deposit, and no amount of mechanical maintenance will stop it.

The performance signature of fouling is one of the most readable in the business. At constant steam flow, first-stage pressure climbs as deposits narrow the nozzle areas. Section efficiencies, measured by enthalpy drop, drift down. Maximum achievable load falls at valves-wide-open. A unit trending all three is fouling, and the trend even tells you roughly where: rising first-stage pressure points at the HP inlet, while degraded LP performance with stable HP pressures points at silica further downstream. Chloride and sulphate deserve special mention because their failure mode is not efficiency but corrosion: they concentrate in the phase transition zone of the LP turbine and drive stress corrosion cracking and corrosion fatigue in blade roots and discs, which is how a chemistry excursion becomes a blade liberation two years later.

Seeing first-stage pressure creep or unexplained megawatt loss? Our maintenance and asset management team can trend your unit's performance signature and tell you whether you have a fouled steam path before anyone opens a casing.

Bearing and Lubrication Failures

Journal bearings are simple, forgiving and utterly dependent on the few hundredths of a millimetre of oil film between shaft and whitemetal. Nearly everything that kills them arrives through the oil.

Contamination is the leading cause. Water gets into turbine lube oil continuously through gland steam leakage along the shaft, and a system that cannot strip it out ends up with emulsified oil, corroded journals and degraded film strength. Particulate from a dirty top-up, a failing filter or wear debris scores the soft whitemetal. The early warnings are cheap to collect: routine oil analysis showing rising water content, rising particle counts or a climbing wear-metal trend is a bearing failure announced months in advance.

Oil film instability is the more theatrical failure. In a lightly loaded or worn journal bearing, the oil wedge itself can begin driving the shaft around the clearance in a self-excited orbit. On a spectrum it appears as a distinct subsynchronous peak at a little under half running speed, and if the machine speeds up until that frequency coincides with the rotor's first critical, the motion locks on and amplitudes become destructive within seconds. The signature is unmistakable to anyone who has read our article on vibration analysis for rotating equipment: whirl tracks just below half of running speed, whip refuses to climb any further. Prevention is a bearing design and loading question, which is why converting a bearing type or changing oil viscosity without analysis is gambling.

Loss of lubrication is the one nobody gets a second chance at. A tripped oil pump with a failed standby, a blocked line, or a drained header on coastdown will wipe every bearing in the machine in less time than it takes to read this paragraph, and a wiped thrust bearing lets the rotor move axially into the stationary blading. This is why the emergency oil pump, the DC pump and the low-pressure trip are tested items, not installed decorations. If you cannot produce the date of your last successful DC oil pump autostart test, that is a finding.

Rotor Problems: Bows, Cracks and the Cost of a Hurried Start

The rotor is the component you can least afford to damage, and most rotor damage is inflicted in the few hours around starts and stops.

Bowing comes in two grades. A temporary thermal bow forms whenever a hot stationary rotor is allowed to sit still: the top of the shaft stays hotter than the bottom, the shaft arcs upward, and any attempt to roll it in that state produces heavy 1x vibration and, if forced, a rub. This is the entire reason turning gear exists, and a hot machine belongs on barring until metal temperatures say otherwise. A permanent bow is what you get when the warnings are ignored: a hard rub locally heats the shaft, the heated fibre yields in compression, and the rotor comes off turning gear with a set that only a repair shop can remove. The early warning is the eccentricity reading before steam admission. Operators who roll off high eccentricity because the dispatcher is impatient are manufacturing a rotor repair.

Thermal transient damage is subtler. Thick HP rotor sections heat slowly; their surfaces heat fast. Every aggressive start puts the surface into compression against the cold bore, every fast cool reverses it, and the accumulated low-cycle fatigue eventually initiates cracks at stress concentrations. Chasing differential expansion alarms and honouring the manufacturer's ramp rates is not conservatism, it is fatigue-life accounting.

Cracking is the end state, and a transverse rotor crack is one of the few faults that can destroy a machine outright. The vibration signature is a change, often gradual, in 1x and especially 2x behaviour as the crack opens and closes with each revolution, which is one reason a good baseline spectrum is worth more than any single reading. The other line of defence is non-destructive examination of bores, surfaces and blade attachment areas at major openings, which is a core scope item in our steam turbine overhaul guide.

Rotor problems in escalating order. A hurried start is not a time saving; it is a withdrawal from the rotor's fatigue account. Source: MIMAH engineering analysis.

Gland and Seal Degradation: The Quiet Efficiency Thief

Labyrinth seals do their work with fine knife-edges running at fractions of a millimetre from the shaft. One rough start, one rub, one bowed-rotor roll, and those clearances open permanently. Nothing breaks, nothing trips, and the machine runs on for years. It simply runs worse.

The costs stack up in three places. Interstage leakage lets steam bypass blading, taking work out of every affected stage. Shaft-end leakage at the HP end wastes live steam. And at the LP end, worn glands let air leak inward, degrading condenser vacuum, which penalises the entire cycle. A unit with tired seals commonly gives away one or two percent of heat rate, invisible on any given day, expensive over a running year.

The early warnings are all performance and utility trends rather than vibration: rising gland steam consumption, a gland steam condenser running hotter than history, falling condenser vacuum with a clean condenser, and section efficiencies drifting without deposit chemistry to blame. The prevention is partly operational, since most seal damage happens during badly managed starts, and partly an overhaul decision: measuring and restoring seal clearances, or upgrading to retractable or brush seal designs where the duty justifies it, is among the highest-return work in any major outage.

Control and Protection Failures: The Overspeed Problem

Everything above damages a turbine. Overspeed is the failure that removes it from the map. If a turbine at full load loses its coupled load, generator breaker opening being the classic case, and the steam valves fail to close, the rotor accelerates at a startling rate. Destructive overspeed, where discs and blade roots burst under centrifugal load, sits not far above the trip setting, which is why every layer of this protection is treated as safety-critical.

The layers are the governor, which should catch a load rejection and control speed without a trip; the overspeed trip, typically set around ten to eleven percent above rated speed; and the stop and control valves that must actually swing shut when told. The commonest defeat of all three is mundane: valve stem sticking. Oxide builds on stems and bushings of valves that sit at one position for months, and a valve that has not moved in a year cannot be assumed to move at all. This is why valve exercising, partial-stroke testing where fitted, and full trip testing on a defined schedule exist. A trip test is not an inconvenience to production; it is the only evidence that the last line of defence works. Any turbine that cannot show a current, dated, passed overspeed and valve-closure test record is running on faith.

The early warnings here are operational rather than mechanical: sluggish valve response during routine load changes, governor hunting, valves that need more actuator force than history, and trip tests that pass marginally or late. Every one of those is the protection system asking for maintenance while it can still be scheduled.

Not sure when your protection was last genuinely proven? Our industrial engineering services cover turbine controls and protection reviews, trip test witnessing and root cause analysis when something has already gone wrong.

Water Induction: The Turbine Killer

Steam turbines tolerate steam. They do not tolerate water. Slugs of condensate entering a machine at speed do damage in three ways at once: thermal shock as water quenches hot metal, distorting casings and bowing rotors; mechanical impact as incompressible water hits blading designed for vapour; and axial thrust as flooded stages push the rotor against its thrust bearing, sometimes hard enough to wipe it and drive rotating parts into stationary ones. A serious water induction event can write off a rotor in seconds, which is why it has its own dedicated standard.

The water rarely comes from anywhere exotic. ASME TDP-1, the industry's recommended practice on preventing water damage to steam turbines, catalogues the usual suspects: feedwater heaters flooding back through extraction lines when a tube fails or level control is lost, attemperator sprays passing when they should be shut, boiler carryover during drum level excursions, gland sealing steam systems, and, most commonly of all, drain lines that are blocked, mispiped or simply left shut. TDP-1's answer is defence in depth: non-return valves and power-assisted isolation on extraction lines, heater high-level protection that actually trips, attemperator block valves, and a drain system operated with discipline rather than habit.

The early warnings are there for operators who look. A widening temperature difference between the top and bottom of a casing metal is the signature of water lying in the bottom half. Extraction line metal temperatures falling toward saturation, water detected at low points, unexplained vibration during load changes, or a heater level alarm that gets acknowledged rather than investigated all belong on the same list. The prevention is ninety percent operating discipline: drains proven open for every start, heater protection tested, and the crew trained to treat a water indication as a trip decision, not a note in the log. Cold reheat and extraction piping does not forgive optimism.

Water induction is the turbine killer, and the warnings are visible to operators who look. A widening top-to-bottom casing metal temperature difference is water lying in the bottom half. Source: ASME TDP-1, prevention of water damage to steam turbines; MIMAH engineering practice.

Frequently Asked Questions

What is the most common cause of steam turbine failure? Blade-path problems are the most frequent source of forced outages, with fatigue and erosion leading. But the distinction worth making is between frequent and fatal: water induction and overspeed cause a small share of events and a large share of write-offs. And underneath a surprising fraction of "mechanical" blade and disc failures sits steam chemistry, because chloride and sulphate carried into the LP turbine drive the corrosion fatigue and stress corrosion cracking that mechanical inspection later finds.

How can I tell my turbine has deposits without opening it? Trend three things at comparable steam conditions: first-stage pressure at a known flow, section efficiencies by enthalpy drop, and maximum load at valves-wide-open. Rising first-stage pressure with falling capability is the classic fouling signature, and the location of the efficiency loss tells you roughly which stages carry the deposit. The steam and saturated-steam chemistry record then tells you what the deposit is likely to be before a single bolt comes off.

What does water induction actually do to a turbine? Three things, usually together: thermal shock that distorts casings and bows the rotor, impact damage to blading from incompressible water, and axial thrust overload that can wipe the thrust bearing and let rotating parts contact stationary ones. Damage ranges from permanent casing distortion to a scrapped rotor. Prevention through drain discipline and extraction line protection, per ASME TDP-1, is orders of magnitude cheaper than any repair.

How often should overspeed and trip protection be tested? Follow the manufacturer's schedule and your insurer's requirements, but the principle is fixed: protection that has not been tested must be assumed not to work. Valves should be exercised regularly to prove they move, and full functional trip tests done at defined intervals and after any protection system work. The units that suffer destructive overspeeds are almost never the ones with a current test record.

Can a bowed rotor be straightened? A temporary thermal bow, caught early, needs nothing more than time on turning gear until eccentricity returns to baseline. A permanent bow from a hard rub is repair-shop territory: controlled thermal straightening, machining, or in bad cases a new shaft section. The economics are so lopsided that the hours spent on barring before a hot restart are the cheapest insurance in the plant.

Are fixed overhaul intervals still the right approach? Opening a healthy machine on a calendar costs money and, worse, introduces its own risks, since reassembly errors are a real failure cause. The alternative is condition-based intervals built on the evidence this article describes. That said, condition-based does not mean never: protection testing stays on the calendar regardless, and inspection intervals for known damage mechanisms should be set by engineering assessment, not hope.

Overhaul When the Evidence Says So, Not When the Calendar Does

Look back across the seven modes and a pattern emerges. Almost every one of them announces itself through the same handful of channels: the vibration spectrum, the performance trend, the steam and oil chemistry record, and the protection test log. A plant that maintains those four disciplines has continuous sight of blade condition, deposits, bearings, seals and rotor behaviour, which is precisely the evidence needed to run overhaul intervals on condition rather than superstition. Operators who once opened machines every five or six years by default now routinely and safely stretch intervals well beyond that, because the machine is telling them it can. The reverse also holds: the same evidence will call a machine in early, before a growing fault chooses its own outage date. Vibration acceptance and monitoring criteria for large turbine generators are standardised, and we have set out how to apply them in our guide to ISO 20816 vibration limits. Steam system fundamentals, for readers building the wider programme, are covered well in the US Department of Energy's steam system sourcebook.

This is the work MIMAH does. Across four decades of team experience in turbine, generator and rotating equipment work, from steam turbine overhauls in Nigeria to root cause analysis at White Nile Sugar in Sudan, the consistent finding is the one this article opened with: turbines almost never fail without warning, they fail without anybody reading the warnings.

Running a steam turbine on hope and a calendar? Talk to our engineering team. We will read your unit's vibration, performance and chemistry evidence, tell you which of these seven modes is live in your machine, and give you an overhaul scope and interval built on what the turbine is actually saying.