Bearing Science and Failure Analysis for Wind Turbines: Reading Damage, Rating Life and the Limits of Both

Calculated Bearing Life and Field Failure

Bearing engineering works from two kinds of knowledge that rarely meet on the same page. The first is the rating calculation, a statistical model of rolling contact fatigue built on defined assumptions about the material, the load, the lubricant and its cleanliness. The second is the failed part itself, which records what actually happened in the steel, the lubricant and the surrounding structure. The revised NREL design guideline for pitch and yaw bearings makes the gap between them explicit: of the damage modes it lists for raceways, rings, cages, seals, bolts and grease, only rolling contact fatigue and static overload of the raceways, core crushing, structural fatigue and static overload of the rings, and fatigue and static overload of the bolts can be calculated, while wear, cage damage, seal separation and grease degradation cannot (Stammler et al., 2024). The same guideline cautions that some damage modes trigger others, so that grease degradation can lead to raceway wear and the effective countermeasure lies upstream of the damage that is finally observed.

Broader estimates point the same direction, though each needs its limits stated. Malloy Wind's earlier technical guidance cites general bearing industry estimates that only 3 to 5 percent of failures originate in classical load-cycle material fatigue and roughly half in contamination (Malloy Wind, 2018); those figures describe bearings in general, not a wind fleet, and the population behind them is not documented. Peng et al. (2022), reviewing the wind bearing lubrication literature, reported secondary estimates that about 70 percent of bearing failures are linked to lubrication. The wind-specific field data are more cautious still: Hart et al. (2023) found spalling the most frequently reported main bearing damage and could not separate wear-induced, surface-initiated and subsurface-initiated fatigue. Read together, these sources suggest that classical subsurface fatigue, the mechanism the rating was built around, accounts for a minority of early wind bearing replacements, with surface-initiated fatigue and lubrication-related damage overlapping in much of the rest, which is why failure analysis carries more weight in wind than in most industrial applications.

The Rating Calculation and What It Leaves Out

Basic and Modified Rating Life

ISO 281 defines the basic dynamic load rating, C, as the constant load a population of identical bearings can carry for one million revolutions with 90 percent surviving, and the basic rating life, L10, as the life that 90 percent of such a population is expected to reach (ISO, 2007). Harris et al. (2009), in the original NREL design guideline, set out the familiar relation, L10 = (C/P)^p in millions of revolutions, with the exponent p equal to 3 for ball bearings and 10/3 for roller bearings and P the dynamic equivalent load that reduces the real combination of radial, axial and moment loads to a single steady value. The modified rating life, Lnm = a1 × aISO × L10, adjusts that figure for a reliability other than 90 percent through a1, which falls from 1.00 at 90 percent reliability to 0.64 at 95 percent and 0.25 at 99 percent, and for lubrication, contamination and the fatigue load limit Cu through the single factor aISO (Harris et al., 2009). The lubrication term rests on the viscosity ratio, kappa, the actual oil viscosity at operating temperature divided by the viscosity the bearing needs, and the contamination term on a factor eC that runs from 0 for heavily contaminated oil to 1 for very clean oil. ISO 76 governs the separate static load rating, C0, which limits the load a stationary bearing can carry before raceway indentation becomes unacceptable (ISO, 2006).

Sources: Harris et al. (2009); ISO (2006, 2007). The exact definition wording is in the paid text of each standard, and the explanations above are paraphrases.

Where the Calculation Stops

The scope of ISO 281 is the first limit. It addresses fatigue and leaves wear, corrosion and electrical erosion outside the calculation (ISO, 2007), and Zweiffel et al. (2024) made the same observation for gearbox bearings, noting that fatigue-based standards such as DIN/ISO 281 do not consider damage from electrical current. The second limit is sensitivity. Keller and Guo (2022), rating a reference pitch bearing for a 1.5 MW research turbine, found a basic rating life of 21,850 to 37,350 hours that fell to about 3,175 hours once the life modification factor for lubrication under slow oscillation was applied, and they described that factor as a large design uncertainty. A follow-up analysis by Keller et al. (2026), applying the revised 2024 design guideline to a reference 1.9 m pitch bearing, moved the estimate a long way: the original method's modified rating life of about 0.36 years became an estimated service life of 12.1 years for one bearing at a 10 percent replacement rate, or 5.8 years for the set of three bearings on a rotor. Compared against 533 pitch bearing replacements across 8,630 turbines of 1.5 to 1.85 MW in an EPRI field database, where the observed life to 10 percent replacement was 17 to 20 years, the authors judged even the updated method conservative, noting that a service life factor near 9, against the 3 they applied, would have been needed to match the field. The numbers are different measures built on different assumptions and are not directly comparable, but the progression shows how much of a pitch bearing calculation rests on factors the field has only recently begun to calibrate. Kenworthy et al. (2024) reached the opposite result for main bearings: rating lives under ISO 281 and IEC 61400-1 stayed well above 20 years for both a 230/600 and a 240/630 spherical roller bearing under expected conditions, which led them to conclude that the method does not account for the main bearing failure rates reported in 1 to 3 MW turbines and to recommend the more detailed analysis now standardized as ISO 16281. In one application the calculation is so sensitive to its lubrication factor that its answer is uncertain by an order of magnitude, and in another it is comfortably safe for a bearing the field replaces at roughly a quarter of turbines by year 20 (Hart et al., 2023). Neither result makes the calculation useless, but both suggest it describes one failure mode among several.

The standards themselves are moving. ISO 281:2007 remains the published edition, with a revision advancing through the ISO draft stages during 2026, and ISO 16281:2025 has replaced the earlier technical specification as the method for a modified reference rating life that accounts for the load distribution inside the bearing, tilt and internal clearance (ISO, 2025). For gearboxes, IEC 61400-4 Edition 2.0, published April 15, 2025, replaced the 2012 edition; it was written with turbines above 2 MW in mind, ties reliability across gear, bearing and turbine standards, adds a drivetrain load clause, replaces fixed prescriptive tests with a verification process built on failure modes and references ISO 281, ISO 15243 and ISO 16281 (International Electrotechnical Commission [IEC], 2025). The shift toward verification by failure mode is a notable one, since it brings the damage vocabulary described below into the design standard itself. Its companion, IEC TS 61400-4-1:2026, sets out a method for calculating the design reliability of gearboxes from failure modes that have publicly available calculation methods, and it states its own limit plainly: because some failure mechanisms seen in the field have no accepted theoretical model, the calculated reliability cannot currently predict the failure probability observed in the field (IEC, 2026). That admission, written into a design standard, is perhaps the clearest statement of the gap this guide describes.

A Common Language for Damage: ISO 15243

ISO 15243:2017 classifies in-service bearing damage by its appearance and likely cause into six main failure modes, each with sub-modes, and defines damage as visible harm to the running surfaces or structure, failure as the state in which a bearing can no longer perform its intended function, and the failure mode as the manner in which it fails (ISO, 2017). The value of the classification is practical: it separates what is seen on the part from the cause that is inferred, and it gives an operator, a service provider and a manufacturer the same names for the same features. The 2017 edition replaced the 2004 edition, so photographic guides written to the older classification remain useful for their pictures while their headings may differ. The NREL pitch and yaw guideline adopts the ISO 15243 definition of a damage mode as the sequence of origin, progression and final inoperability, with one deliberate change: it treats fretting corrosion and false brinelling as wear, where ISO places them under corrosion (Stammler et al., 2024).

Sources: ISO (2017) for the classification and clause numbers; wind examples from Kotzalas and Doll (2010), Malloy Wind (2021a, 2021b, 2022a, 2023c) and the companion guides.

Several of the most consequential wind failure terms do not appear in ISO 15243 at all. White etching cracks are subsurface crack networks bordered by altered microstructure that appears white when etched, white etching areas are the altered material itself, and axial cracking describes the cracks that run across the raceway in the shaft direction, often as the visible end of that process (Kotzalas & Doll, 2010; Malloy Wind, 2021b). These are research and industry terms, and a glossary or failure report that uses them should label them as such, beside the nearest ISO mode, which is usually rolling contact fatigue or fatigue fracture.

The Mechanisms Behind the Modes

Film Thickness and the Lambda Regime

Lubrication contributes to a large share of wind bearing damage, and much of that damage begins in the film itself. Kotzalas and Doll (2010) explained that a film too thin to separate the residual surface texture of roller and raceway, expressed as a lambda value below 1, allows asperity contact through the film, and that micropitting follows when that contact combines with sliding under load. The modeling work of Hart et al. (2022) gave the regime a scale for a 1.5 MW main bearing, finding that a contact inlet temperature of 35 °C marked the transition from elastohydrodynamic to mixed lubrication and that close to 80 percent of operating points were mixed by 40 °C, with grease starvation pushing the share higher. Lambda and the ISO 281 viscosity ratio, kappa, describe related conditions by different measures: lambda compares calculated film thickness with combined surface roughness, while kappa compares the actual operating viscosity with a reference viscosity, so a lambda below 1 does not by itself establish a kappa below 1. Both point toward thin-film operation, and in that regime the surface damage that follows is outside what the fatigue calculation sets out to predict.

Sliding: Heathcote Slip, Skidding and Smearing

Sliding converts a thin film into damage. In spherical roller bearings the roller and raceway curvatures differ, so only two points across each roller roll without slip, an effect Kotzalas and Doll (2010) termed Heathcote slip; sliding under heavy load produces adhesive wear and peeling, and the debris it sheds drives abrasive wear as it recirculates in the grease (Malloy Wind, 2023b, 2023c). In gearboxes the sliding arises differently, as rollers in lightly loaded high-speed bearings fail to reach rolling speed and skid, smearing the raceway (Malloy Wind, 2021b). Both are adhesive wear in ISO 15243 terms, and both are invisible to a rating calculation that assumes pure rolling.

Oscillation, Fretting and False Brinelling

Bearings that oscillate or stand still under vibration fail by a third route. Kotzalas and Doll (2010) described fretting and false brinelling developing when small movements or structure-borne vibration squeeze lubricant from the contact and remove the oxide film, and Malloy Wind distinguishes the resulting evenly spaced marks from true brinelling, which is plastic deformation from static overload (Malloy Wind, 2022a). The revised NREL guideline has refined how the risk is assessed, dropping the older dither and critical angle criteria as unsupported by test results and favoring the oscillation amplitude ratio x/2b, the movement compared with the width of the contact, together with a starvation number that combines it with grease properties (Stammler et al., 2024). This is the damage mode that dominates pitch bearings and appears in parked gearboxes and in generator bearings that sit through vibration while stopped.

Hydrogen and White Etching Cracks

White etching cracks remain the most debated mechanism in wind bearing science. Kotzalas and Doll (2010) reviewed evidence linking white etching areas to high shear stress near non-metallic inclusions, to roller and raceway sliding and to hydrogen diffusing from the lubricant into the steel, and Malloy Wind's synthesis proposes a compatible chain in which sliding breaks down the film, local heat and pressure release hydrogen from water or additives, and the hydrogen weakens the steel beneath the surface under ordinary cyclic load (Malloy Wind, 2021b). That chain is a hypothesis, one of several under study; transient high stresses from grid events, sliding severity and electrical effects are also implicated, and no single sequence has been shown to explain every case (Kotzalas & Doll, 2010; Sanchez et al., 2024). Sanchez et al. (2024) added that, under applied current, the occurrence of white etching cracking depended on the oil as well as the current, which suggests that the lubricant chemistry, the electrical environment and the steel all belong in the same investigation. Field reports from main bearings, where Hart et al. (2023) listed axial cracking among the damage types recorded, indicate that the mechanism is not confined to gearboxes.

Electrical Current

Electrical erosion follows its own physics, and not all damaging current takes the same path. In discharge currents, common mode voltage builds on the shaft until the lubricant film breaks down and arcs, melting micro-pits that accumulate into frosting and fluting; circulating currents arise instead through inductive coupling and flow through the shaft and both bearings without needing voltage to build on the shaft first, which is why grounding alone may not stop them (Guo et al., 2026; Malloy Wind, 2021a). No single safe current density has been established, since laboratory studies have found arc damage at densities as low as 0.01 A/mm² (Sanchez et al., 2024), and the TU Darmstadt application guide offers working bands, above 1 A/mm² likely destructive, 0.1 to 1 A/mm² possibly destructive and below 0.1 A/mm² probably harmless, as screening guidance (Gemeinder & Weicker, 2021). Joshi (2019) added that a bearing voltage of only tens of volts can impose a field of 30 to 40 V/µm across a thin film, and that insulating coatings and ceramic rolling elements act capacitively at high frequency, so the protection they offer depends on the current in a given machine.

Material, Heat Treatment and Residual Stress

Steel composition and heat treatment strongly influence how well a bearing tolerates each of these mechanisms. Through-hardened bearings, typically AISI 52100 at about 1.0 percent carbon, carry neutral to tensile residual stress at the surface, which, combined with fit and operating stress, favors cracking; case-hardened bearings start from low-carbon steel, diffuse carbon and in carbonitriding nitrogen into the surface and retain compressive surface stress over a tougher core, with higher retained austenite that produces shallower debris dents (Malloy Wind, 2018). Kotzalas and Doll (2010) located the greatest tensile residual stress in large induction-hardened rings at the boundary between case and core, where cracks can initiate under heavy static load. Steel cleanliness matters as much as chemistry, because inclusions are where subsurface fatigue and white etching cracks begin, and one manufacturer, NSK, now screens bar stock ultrasonically before forging so that a flawed section can be cut out (Hall & Saxum, 2025). Case carbonitrided steels such as NSK Super-TF combine high surface hardness, compressive residual stress and a tough core (Malloy Wind, 2022b). Surface treatments sit on top of these choices: black oxide converts the surface to improve oil retention and chemical stability, and DLC coatings add a hard, dissimilar layer to resist adhesion, each with limits documented in the companion guides (Malloy Wind, 2020, 2023a).

Failure Analysis in Practice

The Investigation Sequence

Root cause analysis on a wind bearing follows a sequence that Malloy Wind has applied since its first wind work on generators in the late 1990s: collect the failure history, the application record and the work the operator has already done; obtain at least one failed sample; clean, tear down, mark and inspect it; and look below the reported failure mode before specifying any change in geometry, material, heat treatment or coating (Carter, 2022). Some bearing types return to the manufacturer's laboratory and others are dismantled in Malloy Wind's Sioux Falls shop, and the findings, the measurements and the site feedback are shared with the manufacturer to shape the replacement design (Carter, 2022). For larger and newer platforms the investigation extends to metallurgy, metrology and raceway tracing (Hall, 2026), and for pitch bearings it can require pulling the filling plugs and balls or, when the bearing has seized, cutting it in half (Hall & Saxum, 2024).

The reason for the sequence is that the reported failure is often not the whole failure. Malloy Wind has described gearbox bearings returned for one failure mode that showed more below the surface under laboratory inspection, and blade bearings replaced for one mode that revealed others once opened (Carter, 2022). Manufacturing detail can be the deciding variable, as in a site where two brands of identical dimensions and rated capacity ran side by side and only one failed, a difference Malloy Wind traced to raceway hardening and bolt hole corrosion protection (Hall & Saxum, 2024). A replacement specified from the visible symptom alone risks answering a mechanism that was not the cause, or introducing a new weakness while correcting an old one.

Reading the Evidence With Care

The damage on a failed part has to be read with the same caution as any other evidence. Gemeinder and Weicker (2021) showed fluting on a hybrid bearing, which cannot carry current through its ceramic rolling elements, produced instead by mechanical vibration at standstill, and they noted that two failure causes can act at once, so lubricant contamination, overload and thermal damage should be excluded before current is blamed. Hart et al. (2023), reviewing main bearing damage reports across 15 GW of capacity, found spalling the most common recorded damage and read the accompanying surface damage as evidence that much of it began at the surface, while stating plainly that the data could not yet separate wear-induced, surface-initiated and subsurface-initiated fatigue. Spalling, in other words, is an end state that several of the ISO 15243 modes share, and the origin has to be found in the surrounding features, the subsurface sections and the operating history.

What to Collect Before the Bearing Leaves the Turbine

The most useful evidence is often lost before the bearing reaches a shop. Purged grease and oil samples, taken before the replacement and compared with a baseline from earlier service, record wear metals and contaminants that the cleaned part no longer shows. Elemental analysis typically reports 10 to 15 elements in parts per million, with iron, copper alloys from cages and silicon among the readings that matter (Hall, 2026); water content, grease condition and degradation need separate test methods chosen for the sample and the application. Peng et al. (2022) cited one study's proposed criterion of replacing grease whose dropping point has fallen by more than 30 °C or whose iron content exceeds 5,000 ppm, a useful reference point that should not be read as a universal threshold. Photographs of the raceway, the bolt holes and the grease in place, the turbine's SCADA temperature and vibration history, the relubrication record and any shaft voltage measurements complete the record, and Hart et al. (2023) identified site conditions and relubrication history as the variables most worth adding to future failure analysis.

Sources: Carter (2022); Gemeinder and Weicker (2021); Hall (2026); Hart et al. (2023); Peng et al. (2022).

Where to Look First

The damage seen on a failed bearing does not settle its cause, but it does suggest where an investigation should start. The table below gives an initial direction for each common finding, drawn from the mechanisms above, and is meant to shape the first questions, not to stand in for a root cause conclusion.

Sources: the mechanisms and sources discussed in this guide.

In Finality

Research and field experience point to the same conclusion, and it shapes every replacement decision in the companion guides. The rating calculation remains the necessary starting point, but its scope is rolling contact fatigue under defined assumptions, and the bearings that fail early in wind turbines fail largely through lubrication, sliding, oscillation, hydrogen, current and contamination (Hart et al., 2023; Kotzalas & Doll, 2010; Stammler et al., 2024). ISO 15243 gives those failures a shared vocabulary, and the move in IEC 61400-4 toward verification by failure mode, together with the candid limits written into IEC TS 61400-4-1, suggests that vocabulary is becoming part of design as well as diagnosis (IEC, 2025, 2026; ISO, 2017). Failure analysis supplies what neither the calculation nor the classification can, the evidence of what happened in a specific bearing on a specific turbine, and the most defensible upgrades are those specified after that evidence has been read, with its limits acknowledged, against the mechanism it shows.

Tech Notes on Bearing Science and Failure Analysis

• Through Hardened vs Case Hardened Bearings
• White Etching Cracking Failures Explained
• What is Heathcote Slip?
• Main Bearings: What is peeling damage?
• Brinelling vs False Brinelling
• Electrical discharge in generator bearings
• What is Black Oxide?
• Are DLC coatings no longer cutting it?
• Bearing Internal Clearance
• Forthcoming: ISO 15243 failure modes with wind examples; Reading a grease analysis report; How to ship a failed bearing for analysis; ISO 281, ISO 16281 and what changed

Other Advanced Guides: Blade Bearings, Main Shaft Bearings, Gearbox Bearings, Generator and Pitch Motor Bearings, Repower and Retrofit.

Frequently Asked Questions

Requesting a Failure Analysis

Call 605.336.3693 or email wind@malloyelectric.com with the turbine model, the bearing position and designation, and a description of what failed. The team will advise on what to collect before removal, how to ship the bearing and whether the investigation is best done in Malloy Wind's Sioux Falls shop, a service provider's shop or the manufacturer's laboratory. Before the bearing comes out:

• Photograph the seals, any leakage path, cracks and the mounting in place.
• Take a grease or oil sample and keep it with the bearing.
• Leave the grease in the bearing and do not clean or reassemble it.
• Record the turbine model, bearing position and designation.
• Pull the SCADA, vibration and relubrication history, and any shaft voltage readings.

References