Roller Sliding in Wind Turbine Gearbox High-Speed-Shaft Bearings

4th Conference for Wind Power Drives 2019, Aachen, March 13, 2019

David Vaes, SKFJon Keller, National Renewable Energy LaboratoryPietro Tesini, SKFYi Guo, National Renewable Energy Laboratory

Photo by Jonathan Keller, NREL 49037

NREL/PR-5000-73428

© SKF Group

• Background

• Measurement setup

• Numerical model building and verification

• Transient load conditions

• Conclusions

Overview

Background

Photo from SKF

© SKF Group

Background: Premature Failures with White Etching CracksHS-S

HIS-SLS-P

• During recent years several countermeasures have been taken

• Since introduction of black-oxidising, no serial failure case reported by gearbox original equipment manufacturer

• Some failures still reported today by after market and end users:– Proper statistics are

missing.

Early cracks occur commonly within the

first 1–3 years of operational time

(<10% of the calculated rating life).

Photos from SKF

LS-P = Low-Speed PlanetHS-S = High-Speed ShaftHIS-S = High Intermediate-Speed Shaft

Illustration from SKF

© SKF Group

Premature Bearing Failures: Understanding the Drivers

‘Cyclic’ Stressesand Loading

Rolling Contact Fatigue

‘Accelerated’ Fatigue (Premature Spalling)

Higher ‘Stresses’ Lower ‘Material Strength’

WEC in bearing test exposed to additional

tensile stresses.

WEC in bearing test exposed to short heavy

loads.

White Etching Cracks (WECs) Occurrence in Bearings

Structural Stresses(Sub-Surface)

Frictional Stresses and Wear(Surface, Near-Surface)

Hydrogen(Surface and Sub-Surface)

WEC in larger bearing rolling contact fatigue.

WEC in bearing test exposed to water stand-

still corrosion.

WEC in bearing exposed to electrical

currents.

WEC in bearing test running under mixed

friction and slip.

WEC in bearing test that was hydrogen

charged.

Short Heavy Loads

Standstill Corrosion Electrical Current

Role of lubricant and tribochemistry

Low Film and Slip

Photos from SKF

© SKF Group

• The exact combination of drivers that explains the failures in wind gear units is not yet understood:- Limits of current solutions are not fully understood- A better understanding of critical operating conditions in wind gearboxes still

required• Simulations and measurements complete each other.

Critical Operating Conditions in Wind Gearboxes

Simulation• Requires a detailed set of

boundary conditions• Requires tuning of model

parameters• Disturbances are negligible• Possibility of in-depth

analysis of roller kinematics.

Measurement• Provides data in complex

operating conditions• Only requires to process the

measured signals• Limited number of output

parameters• Measurement disturbances and

input uncertainties.

Photo from SKF

Illustration from SKF

Illustration from SKF

Cons

ConsPros

Pros

Measurement Setup

Photo by Jonathan Keller, NREL 49037

© SKF Group

• Instrumentation focused on high-speed shaft, bearings, and lubricant:- Shaft speed

- Cage speed

- Roller speed

- Shaft torque and bending

- Stray current

- Bearing temperatures

- Air temperature and humidity

- Lubricant temperatures and moisture content

- LogiLube and Poseidon lubricant monitoring and routine oil samples

- SKF iMX8 system.

Winergy PEAB 4410.4 Gearbox and SKF Cylindrical Roller BearingsGearbox Instrumentation

Photo by Mark McDade, NREL 49050

Sliding

Source: Keller, J. and S. Lambert, Gearbox Instrumentation for the Investigation of Bearing Axial Cracking, NREL/TP-5000-70639, 2018.

© SKF Group

GE 1.5 SLE turbine:• Blade flap and edge bending

• Blade pitch angles

• Rotor azimuth and speed

• Main shaft torque and bending

• Active and reactive power

• Nacelle yaw

• Tower bending and torsion

• Wind vane offset

M5 met tower:• Air temperatures and humidity

• Wind speed and direction

And more…GPS time stamped.

Turbine and Meteorological (Met) Tower Instrumentation

Photo by Dennis Schroeder, NREL 21884

GE 1.5 SLEESS Mk 6 controller

M5 Met

Tower

Photo by Dennis Schroeder, NREL 49409

© SKF Group

Roller and Cage-Speed MeasurementCage-speed measurement:• Pin passage detected by

proximity sensor• One speed measurement per

cage revolution.

Roller-speed measurement:• Magnetized roller• Changing magnetic field detected

by coil next to the bearing• Position of magnetized roller

determined by cage pin.

Cage Pin

MagnetizedRoller

Photos by Jonathan Keller, NREL 40979 and 40981

Illustration from SKF

Numerical Model Building and Verification

Illustration from SKF

Illustration by NREL

© SKF Group

Bearing B

Bearing A

• Mean cage speed during each revolution is available

• Instantaneous roller speed is available but highly disturbed

• Operating bearing clearance is unknown (bearing inner ring temperature often not available)

Measurement Limitations and Processing

Processed Rolling Element (RE) Speed vs. Averaged at Stationary, Fully Loaded Operating Condition

Spee

d [rp

m]

Time [s]

Postprocessing of the measurement is necessary:

1. Select time intervals where the cage speed is constant

2. Use several cage revolutions to filter the disturbance of the roller speed

3. Select best measured intervals.

© SKF Group

Measurement Screening1. Systematic detection of all cage-speed plateaus:

Spee

d [r

pm]

Time [s]

2. Least-squares fit of a piece-wise approximation of the roller speed:

Azimuth [deg]

Spee

d [r

pm]

3. Systematic selection based on error:

Spee

d [r

pm]

Azimuth [deg]

Final selection, based on most interesting and diverse

operating conditions to increase the validity of the

semiempirical model.

rpm = revolutions per minute | s = seconds | deg = degree

© SKF Group

• At lower operating temperature, the rollers decelerate significantly more in unloaded zone

• Higher temperature lower viscosity less drag losses on rollers in unloaded zone slower deceleration

Effect of Temperature on Roller Speed: “Down Slope”

0 100 200 300 400

Cage position [deg]

1500

2000

2500

3000

3500

4000

Rol

ler s

peed

[rpm

]

RS

RE speed using averaging

Temp = 36°C0 100 200 300 400

Cage position [deg]

1500

2000

2500

3000

3500

4000

RS

RE speed using averaging

Temp = 57°C0 100 200 300 400

Cage position [deg]

1500

2000

2500

3000

3500

4000

RS

RE speed using averaging

Temp = 67°C

© SKF Group

• A significantly larger temperature difference is measured on bearing B than on bearing A:– Bearing B has much smaller radial clearance and a larger loaded zone than bearing A

• Slower deceleration of the rollers in unloaded zone at increasing temperature (lower oil viscosity)

• Drag losses increase with the size of the roller (larger projected surface of the rollers).

Effect of Temperature and Oil Viscosity – Overview

Tem

pera

ture

Diff

eren

ce[C

]

Time [min]

min = minute | µm = micrometers

Temperature Difference of Bearing RingsClearance as Function of

Temperature Difference of Bearing Rings

Bearing A

Bearing B

© SKF Group

The model is designed by two SKF proprietary software.

SKF Numerical Modelling

SKF BEAST SKF SimProExpert

QJ bearing replaced by nonlinear stiffness.

• Transient multibody dynamic solver

• Detail contact calculation, elastohydrodynamic layer lubrication

• Cage-roller interaction

• Drag losses not automatically modeled.

Linear rotational damping torque is applied to both the cage and the rollers.

Illustrations from SKF

© SKF Group

Analytical Model Predicts Roller and Cage Sliding

Roller dynamics model (analytical):• Harris roller dynamics model

Lubricant hydrodynamics model based on:• Bercea cage friction model• Dowson and Higginson lubricant model

0i o cQ Q F− + =

0ij o v cgF F F Q− + − =

12

ri o cg cg c

dM M DQ Jdωµ ωφ

− + =

Tangential

Radial

Torsional

Primary Governing Equations

Cage

Source: Guo, Y. and J. Keller. Forthcoming. Analytic Formulations of Rolling Element Bearing Sliding in Wind Turbine Gearboxes, Mechanism and Machine Theory.

Frictional Energy Loss

Roller Free Body Diagram

Illustration by NREL

© SKF Group

Parametric Studies To Verify Model ParametersExample in BEAST: Influence

of Rotational Damping

Spee

d [rp

m]

Azimuth [rad]

Example in Analytical Model: Influence of Temperature

Spee

d [rp

m]

Azimuth [rad]

Nms = newtonmeter-second | rad = radian

© SKF Group

Verification of Simulation ResultsSp

eed

[rpm

]

Azimuth [rad]

Torque = 7,930 Nm, TOR ~ 41oC, Toil = 45oCδB = 5𝜇𝜇m, δA = 95𝜇𝜇m

Azimuth [rad]

Torque = 7,930 Nm, TOR ~ 63oC, Toil = 63oCδB = 30𝜇𝜇m, δA = 145𝜇𝜇m

Azimuth [rad]

Torque = 9,520 Nm, TOR ~ 57oC, Toil = 41oCδB = 20𝜇𝜇m, δA = 145𝜇𝜇m

Input

Nm = newtonmeters | TOR = Outer ring temperature | Toil = Oil supply temperature | δ = Clearance

Roller Speeds Roller Speeds Roller Speeds

Measurements at Transient Conditions

kW = kilowatt | LSS = Low-speed shaft | HSS = High-speed shaft | kNm = kilonewtonmeter

Pow

er, T

orqu

e an

d Sp

eed

Drivetrain Conditions during Emergency Stop

© SKF Group

Transient Conditions – Emergency Stop

21:49:00 21:49:10 21:49:20

Mar 04, 2018

-500

0

500

1000

1500

2000

Power [kW]

Torque LSS [kNm]

HSS speed [rpm]

21:49:00 21:49:10 21:49:20

Mar 04, 2018

-1000

0

1000

2000

3000

4000

5000

Rol

ler s

peed

[rpm

]

Roller speed at emergence stop: RS

Measured

Theoretical

21:49:00 21:49:10 21:49:20

Mar 04, 2018

-1000

0

1000

2000

3000

4000

5000

6000

Rol

ler s

peed

[rpm

]

Roller speed at emergence stop: GS

Measured

Theoretical

• Torque oscillations at drivetrain 1st

eigenfrequency• Oscillations result in cage and roller dynamics• Rotor side more sensitive to torque oscillations

than generator side• Roller speed measurements unreliable at low

speed conditions.• At brake engagement roller speed reduces to

about 80% slip and accelerates back in about 1.5 seconds.

Rotor Side Generator Side

Pow

er, T

orqu

e an

d Sp

eed

Drivetrain Conditions

21:49:00 21:49:10 21:49:20

Mar 04, 2018

0

100

200

300

400

500

600

700

Cag

e sp

eed

[rpm

]

Cage speed at emergence stop: RS

Measured

Theoretical

21:49:00 21:49:10 21:49:20

Mar 04, 2018

0

100

200

300

400

500

600

700

Cag

e sp

eed

[rpm

]

Cage speed at emergence stop: GS

Measured

Theoretical

Rotor Side Generator Side

© SKF Group

Transient Conditions – LVRT (50% Drop for 300 milliseconds)

21:46:00 21:46:04 21:46:08 21:46:12 21:46:16 21:46:20

Mar 05, 2018

0

200

400

600

800

1000

1200

1400

1600

1800

Power [kW]

Torque LSS [kNm]

HSS speed [rpm]

21:46:00 21:46:04 21:46:08 21:46:12 21:46:16 21:46:20

Mar 05, 2018

2500

3000

3500

4000

4500

5000

5500

6000

Rol

ler s

peed

[rpm

]

Roller speed at LVRT: GS

Measured

Theoretical

21:46:00 21:46:04 21:46:08 21:46:12 21:46:16 21:46:20

Mar 05, 2018

1000

1500

2000

2500

3000

3500

4000

4500

Rol

ler s

peed

[rpm

]

Roller speed at LVRT: RS

Measured

Theoretical• Torque oscillations at drivetrain 1st

eigenfrequency after Low-Voltage Ride Through (LVRT)

• Load oscillations resulting in cage and roller dynamics

• Roller speed reduces to about 50% slip and accelerates back in about 0.5 seconds.

Rotor Side Generator Side

21:46:00 21:46:04 21:46:08 21:46:12 21:46:16 21:46:20

Mar 05, 2018

460

480

500

520

540

560

580

600

620

Cag

e sp

eed

[rpm

]

Cage speed at LVRT: RS

Measured

Theoretical

Rotor Side

21:46:00 21:46:04 21:46:08 21:46:12 21:46:16 21:46:20

Mar 05, 2018

500

520

540

560

580

600

620

640

660

Cag

e sp

eed

[rpm

]

Cage speed at LVRT: GS

Measured

Theoretical

Generator Side

Pow

er, T

orqu

e an

d Sp

eed

Drivetrain Conditions

© SKF Group

• Operating at low load results in much higher slip levels.

Roller Speed at “Curtailment”1500 kW 250 kW

Rotor Side

Generator SideInboard

Future Steps and Conclusions

Photo by Dennis Schroeder, NREL 49418

© SKF Group

• Repeat the procedure for the cage speed at low load. The proximity sensor is not affected by the same disturbance as the induction coil.

• Simulation of transient conditions (i.e., when measurement cannot be efficiently filtered from noise).• Use simulation results to evaluate critical conditions for the bearings (e.g., by power slip density or

cumulative frictional energy).

Ongoing Steps

Pow

er S

lip D

ensi

ty [W

/mm

2 ]

Azimuth [rad]

Spee

d [rp

m]

Azimuth [rad]

Roller Speed Frictional EnergyPower Slip Density

W/mm2 = watts per millimeters squared | J/s = joules per second

Azimuth [rad]

Fric

tiona

l Ene

rgy

[J/s

]

© SKF Group

• Measurement of roller and cage speed gives useful insight in the bearing kinematics at different operating conditions:– Low load/curtailment– Emergency stop– LVRT

• High roller slip and accelerations have been measured at these events• BEAST model has been built and shown to be able to accurately predict roller and cage

behaviour at different loads and temperature• Next steps:

– Apply and validate the models at special events– Evaluate the roller slip losses at special events.

Conclusions

© SKF Group

Thank you for your attention!Vaes, D., Y. Guo, P. Tesini and J. Keller. 2019. Investigation of Roller Sliding in Wind Turbine Gearbox High-Speed Shaft Bearings (Technical Report). NREL/TP-5000-73286. National Renewable Energy Laboratory (NREL), Golden, CO (US). http://www.nrel.gov/docs/fy19osti/73286.pdf

This work was authored by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the U.S. Department of Energy (DOE) under Contract No. DE-AC36-08GO28308. Funding provided by the U.S. Department of Energy Office of Energy Efficiency and Renewable Energy Wind Energy Technologies Office and CRADAs 16-608 with SKF GmbH and 17-694 with Flender Corporation. The views expressed in the article do not necessarily represent the views of the DOE or the U.S. Government. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for U.S. Government purposes.