Comments

Overall remarks

It is a well-written paper with enough material to form a substantial contribution to the pitch bearing literature, particularly as there is a dearth of papers about three row roller bearings. As it is a methodology paper, it should be relatively easy to follow the procedure adopted which it is. The flow of the methodology has been clearly illuminated by the well-separated chapters. The reasoning for the methodology has been illuminated properly with enough validation done for each aspect of the method. There are some minor issues which need to be corrected highlighted in the next section.  Since the authors validated this method with its relevant assumptions for a specific wind turbine, it may also be good to describe or postulate if this method will work in general for wind turbines with three row roller bearings as well (as much as possible within the limits of confidentiality).

Specific comments

1. Line 7: The modification of the bearing is mentioned but never discussed further (except in the conclusion). Please either include a line or so in the main text about this. Even if it is confidential, possibly mention this.
2. Line 81: The choice of 21 slices for the FEM model is not explained. Why was this value chosen? Is it standard within the extension?
3. Line 83: “contacting roller and raceway” may be a better formulation.
4. Line 88: Unless confidential, possibly mention if the split happens in the inner or outer ring and approximate location. I see a split in Figure 1; this could be highlighted there, too.
5. Line 198: The reference “Stammler et al.”: Please date these references as there are multiple from the same author.
6. Line 316-318: It is mentioned that the Reusner algorithm gives a more accurate load distribution than the FE simulation. It is mentioned later on about it being due to the Hertzian nature of contact algorithm (“For these 72 simulations, the contact pressures of all rollers were determined using a non-Hertzian contact algorithm based on Reusner (1977). Unlike the Hertzian rolling elements used in FE, the non-Hertzian algorithm allows for detection of edge stresses and is more accurate in general.”). To make it clearer, the hertzian nature of the FEM model can be explained before this part of the text.
7. Line 517: About the bending moment and resulting structural deformation playing a larger role in the loading of the radial row than the radial load itself: How exactly does structural deformation happen in this case? What is a possible reason for this? Can there be numbers put to this (if possible)?
8. Line 527: By a simplified approach like in Eq. 21, do you mean to say that only the axial load (ignoring the Fr mentioned in Eq.21) is to be considered here?  Also, it could also be clarified then that since it is a roller bearing, the 10/3 exponent is to be used. Also, for the sake of clarity, is Ca of a single axial row to be used for life?

General remarks:

Well explained and clear structured paper.

The graphics are too small and difficult to read. 

Some sources only given with name, no further information (year etc.) à e.g. “see Stammler et. al.”

Will the use of a full 3D model of the bearing and the time series allow to predict the circumferential location of the rolling contact fatigue failure?

Data Quality "Fair" due to the confidentiality requirement.

In detail:

Line 7: In comparison to ball bearings, the calculation of roller bearings [..] leading to significantly more degrees of freedom.

A Ball bearings also have many degrees of freedom, if modelled with balls having realistic kinematic behaviour (instead of springs and fixed contact angle) since the contact angle ball to raceway can differ from ball to ball and is strongly load dependent.

Line 8: For this paper, the bearing is modified slightly: What does modified slightly mean? Please describe more detailed, if possible.

Line 56-58: Does the roller profile used for the calculation differ between the "confidential" turbine pitch bearing and the more "generale" bearing? If so, how is the influence onto the "normalized" results?

Line 65: But is it possible to get the ISO life calculations quantitively reliable?

Line 81-83: Can you explain why 21 springs are used in the FE model, whereas ISO/TS16281 requires min. 30 slices in their roller slice model?

Line 88: [..] This split of the ring is considered in the model and the surfaces are connected to each other by an internal frictional contact. [..]  Axial clamping force? Bolt preload? Influence bolt preload on roller preload?

Line 166-167: Influence of pitch angle might only become more important if the blade model differs circumferentially from its mechanical properties. How is the blade root stiffness modelled, uniform or non-uniform, anisotropic?

Line 220: again the question about blade root model and physical properties around circumference - uniform or non-uniform

Line 296-298: Is there a difference for small back and forward motion compared to full rotations in calculating dynamic equivalent load and bearing lifetime? If so, further explanation would help. Have time steps with negligible rotation been ignored, if so what is the minimum pitch angle to be considered?

Line 310: why 21 laminae in FE compared to min. 30 as per ISO 16281

Figure 14: Graphics of convergence analysis in not convincing compared to line 349. The graphics do not show why 30 lamina is a suitable choice.

Line 405: What is the expected difference in rating life between analysis done on time steps compared to load bins?

Line 516: What impact will the radial row, at 10-15 times the life of the axial raceways, have on the combined rating life?

Line 519-520: Will radial load be added to eq. ax. load Pa and compared with axial load capacity Ca? Further explanation required for formula.

Line 529: Further explanation required, also in relation to formula above. Normally ball bearings have higher k factor on moments compared to axial rows of 3 row roller bearings if calculating axial and radial raceways separately on 3 row roller bearings.

Line 566: What would be a high edge stress?

Line 581: What is the expected failure mode of a three-row roller bearing