| 10 Sep 2026
Fleet-scale evidence and a testable coupled mechanism for radial-raceway spalling in an offshore 5 MW three-row roller main bearing
Abstract. Standard rating-life calculations predicted sufficient durability, yet premature radial outer-raceway spalling recurred in a 203-turbine offshore 5/6.2 MW fleet. By 11 August 2026, eight bearings had been removed after 26–53 months and two additional units were stopped awaiting replacement. All eight dismantled bearings exhibited damage within the stationary 3–9 o’clock lower radial load zone, whereas axial-row damage was limited. Fleet-scale teardown evidence for offshore three-row cylindrical-roller main bearings is rarely reported in the peer-reviewed wind-energy literature. The present dataset therefore provides an unusual opportunity to establish a repeatable circumferential damage pattern and to organize competing mechanisms within an explicit evidence hierarchy. Roller-end breakage, gearbox-side cage contact, a measured full-power inner-to-outer-ring temperature difference of about 12 °C, and partly obstructed grease inlets provide mutually consistent, but individually non-unique, evidence for load redistribution and lubrication-access effects. Grease screening of 198 turbines identified Cu ≥ 5000 ppm in 21 units (10.61 %), Fe ≥ 5000 ppm in 10 units (5.05 %), and the combined condition Cu > 10000 ppm and Fe > 5000 ppm in five units (2.53 %); these thresholds are used only as project-specific risk flags, not as transferable diagnostic cut-offs. A nominal Hertz calculation gives a baseline maximum roller load of about 109 kN and contact pressure of 0.997 GPa. System FEA predicts 0.067° maximum raceway misalignment; across the 100 mm roller length this corresponds to an end-to-end geometric offset of about 0.12 mm, providing an order-of-magnitude structural basis for testing a moderate 20–50 % edge-pressure amplification range (Kedge = 1.20–1.50), while not constituting a calibration of Kedge. The combined evidence is most consistent with accelerated rolling contact fatigue in which the fixed load zone controls damage location, while thermo-structural load redistribution, restricted grease replenishment, debris and local case-condition variability can reduce fatigue margin. The principal contribution is therefore fleet-scale field evidence and an evidence-ranked, falsifiable engineering framework for offshore main-bearing failure assessment, rather than identification of a unique root cause or a new life-prediction equation.
First and foremost the work is extremely valuable because it is one of the very rare instances where a large dataset including failures is documented publicly, including aspects of the root cause analysis. This alone makes it worth a publication.
The key question is now if this is a scientific paper and a fit for this journal as is, and I think it is not quite there. In its present shape and especially considering the title, I think it would be a much better fit for a failure analysis journal. Then, the authors indicate the key novelty of the paper is a new method towards evaluating the causes, and I think if this would be coupled clearly with the 'fleet' aspect it would make for a solid scientific question that would also fit this journal. But in such a case, it would require a major overhaul, beginning with the title being more in the line of "Methods for investigating fleet-scale damages.." or something similar. Then, the introduction would have to have many more references on root cause analysis methods and examples of root causes analyses in wind turbines and other bearings to allow a clear distinction between the methods used there and the methods one would use when being able to use fleet data.
In my eyes, the first decision the authors have to take is thus this: Is this a singular analysis or a method development? If it is a single analysis, this journal might not be the best place for it. If it goes for methods, title and introduction need a major overhaul and it has a most pressing need to make a distinction between 'more singular' analyses and what the authors do in their work.
There are several minor things to reconsider in the writing, in the following just a few first comments on this:
Common main bearing concepts include spherical roller bearings and tapered roller bearings (double row are two single row) - could you elaborate on the reasoning behind introducing a three-row roller bearing in this design?
The authors state they cannot calculate failure probability, but the data seems to be there. I can only hazard the guess this is not calculated for legal reasons, but if the reasons are technical, here a suggestion: It seems hard to understand how the commissioning date is unknown for a turbine, and even if the reason for removing the bearing and the exact damage time are unknown, one could take the value as left-censored and the remaining operating turbines as right-censored values. It would also be worthwhile to understand if the bearing were designed with L10 or L1 and what was the target life of them (20 years, 25 years?)
For an example of integration of left-censored values into large bearing lifetimes, see Menck et al from 2026 (https://asmedigitalcollection.asme.org/tribology/article/148/10/102602/1233804/Rolling-Contact-Fatigue-Life-of-Large-Oscillating)
It remains unclear if the bearings removed from the turbines are within the 198 turbines which grease samples are taken into account.
In the light of the possible major overhaul of the entire paper I refrain from giving more detailed comments and questions, although it deeply saddens my bearing engineer heart. But perhaps there is another chance for a deeper look at this.