Wind Turbine Gearbox Planet Bearing Failure Prediction Using Vibration Data, S. Koukoura, University of Strathclyde
Data Insights from an Offshore Wind Turbine Gearbox Replacement, A.K. Papatzimos, University of Edinburgh
Further investigation of the relationship between main-bearing loads and wind field characteristics, A. Turnbull, University of Strathclyde
Damage Localization using Model Updating on a Wind Turbine Blade, K. Schröder, University of Hannover
Data Insights from an Offshore Wind Turbine Gearbox Replacement
Alexios Koltsidopoulos Papatzimos1,2 Tariq Dawood2
Philipp R. Thies3
1Industrial Doctorate Centre for Offshore Renewable Energy (IDCORE), Edinburgh, EH9 3JL
2EDF Energy R&D UK Centre, London, CR0 2AJ
3University of Exeter, Cornwall, TR10 9FE EERA DeepWind 2018
15thDeep Sea Offshore Wind R&D Conference Trondheim, Norway
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference 1
1. Introduction and Motivation 2. Wind Turbine Gearbox Failures 3. Wind Turbine Gearbox Monitoring 4. Data Pre-processing
5. Failure Detection & Diagnosis 6. Data-Driven Models 7. Conclusions & Future Work
2 R&D UK Centre
Contents
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Introduction-EDF Group Offshore Assets
• Teesside Offshore Wind Farm
• 27 2.3MW turbines
• 1.5 km offshore
• 7-15m water depth
• Installation completed in June 2013
• Blyth Demonstrator Project
• 5 Vestas 8.3MW turbines
• Future assets
• Totalling 1.5GW
Motivation
• Gearbox replacement @ Teesside
• Gearboxes are designed to last for the lifetime of the asset- IEC 61400-4
• Majority of onshore and offshore wind turbines have a geared drivetrain
• Currently largest installed wind turbine (V164-8.0 MW) has a gearbox
• Early detection by OEM
• Reduce downtime
• Reduce component lead time
• Understand component reliability
• Perform future fault prediction and diagnosis
3 R&D UK Centre
1. Introduction and Motivation
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
4 R&D UK Centre
2. Wind Turbine Gearbox Failures
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
[1]
[2]
5 R&D UK Centre
2. Wind Turbine Gearbox Failures
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
[1]
[1]
Most common failure causes [3, 4]:
• Fundamental gearbox design errors
• Manufacturing or quality issues
• Underestimation of operational loads
• Variable and turbulent wind conditions
• Insufficient maintenance
6 R&D UK Centre
2. Wind Turbine Gearbox Failures
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Micropitting [9] Tooth breakage [10] Pitting [11]
Spalling [12] Scuffing [11]
Most common failure locations [4-8]:
• HS Bearing
• IMS bearing
• Planet bearing
Most common failure modes:
7 R&D UK Centre
2. Wind Turbine Gearbox Failures
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
• SCADA
• Temperature, pressure, vibration, current, rotational speed, etc.
• Timeseries
• CMS
• Vibration
• Sampling in time instances
• Pre-processed (Envelopes, FFTs, Cepstrum, RMS, etc)
• Oil Particle Counter
8 R&D UK Centre
3. Wind Turbine Gearbox Monitoring
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
3 stage planetary/ helical gearbox
Power Curve
9 R&D UK Centre
4. Data Pre-processing
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference SCADA Alarms + maintenance
log timestamps have been removed that include:
Yaw, Pitch, Generator, Electrical, Grid, Sensor failures, Environmental conditions, Maintenance operations
Filtering
Savitzky–Golay filter
Filtered Power Curve
Vibration Signal Vibration Signal
10 R&D UK Centre
4. Data Pre-processing
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Binned Power Curve
Filtering
Bins IEC 61400-1-22 Filtered Power Curve
11 R&D UK Centre
5. Failure Detection & Diagnosis
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Gearbox Oil Temperature vs Active Power Gearbox Oil Temperature vs (Rotor Velocity)^2 High Speed Temperature vs Active Power
SCADA Active Power vs Rotor Velocity
12 R&D UK Centre
5. Failure Detection & Diagnosis
Gearbox Oil Temperature Bins Gearbox Oil Temperature Bins
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
SCADA
13 R&D UK Centre
5. Failure Detection & Diagnosis
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Planet Bearing Envelope
Planet Bearing FFT
CMS
14 R&D UK Centre
5. Failure Detection & Diagnosis
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
CMS Planet Bearing Cepstrum RMS
Particle Counter
15 R&D UK Centre
5. Failure Detection & Diagnosis
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
SCADA
16 R&D UK Centre
6. Data-Driven Models
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Algorithm Specifications True Pos.
(Healthy)
True Positive Rate (Warning)
SVM Gaussian, Scale:0.26 97% 92%
Ensemble Bagged Trees, Split: 10, learners: 30 96% 91%
KNN Mahalanobis, NN=10 96% 92%
Decision Tree Gini's index, max number of splits: 400 95% 86%
SVM Quadratic, box constraint: 1 93% 81%
• “Healthy” state for data 4 months after replacement (orange)
• “Warning” state for data 4 months prior to replacement (blue)
CMS
17 R&D UK Centre
6. Data-Driven Models
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
• Not constantly monitored systems
• Automation of forecasting models
• Autoregressive model for RMS signal
• Predicted same slope for 26 out of 27 turbines
Conclusions
• Planet stage bearing spalling on a 3-stage 2.3MW turbine gearbox
• Similar studies investigated catastrophic gearbox failures
• Identify and diagnose the failure by using SCADA and CMS data
• Temperature readings
• RMS vibration
• Data driven models to predict future failures Future Work
• Further test the models in other failure modes and wind turbine models
• Investigate the environmental conditions’ impact on the results
R&D UK Centre
18
7. Conclusions and Future Work
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
[1] Carroll J, McDonald A and McMillan D 2016 Failure rate, repair time and unscheduled O&M cost analysis of offshore wind turbinesWind Energy196 1107-19
[2] Koltsidopoulos Papatzimos A, Dawood T and Thies PR 2017 An integrated data management approach for offshore wind turbine failure root cause analysisProc. 36th Int. Conf. on Ocean, Offshore and Arctic Engineering(Trondheim) vol. 3B (New York: ASME)
[3] Qiu Y, Chen L, Feng Y and Xu Y 2017 An Approach of Quantifying Gear Fatigue Life for Wind Turbine Gearboxes Using Supervisory Control and Data Acquisition DataEnergies101084
[4] Musial W, Butterfield S, McNiff B 2007 Improving Wind Turbine Gearbox ReliabilityEuropean Wind Energy Conference [5] Nejad AR, Gao Z, Moan T 2014 Fatigue Reliability-Based Inspection and Maintenance Planning of Gearbox Components in Wind Turbine DrivetrainsEnergy Procedia53248-57
[6] Nejad AR, Gao Z and Moan T 2015 Development of a 5MW reference gearbox for offshore wind turbinesWind Energy 196 1089–1106
[7] McVittie D 2006Wind turbine gearbox reliabilityOnline: https://goo.gl/gqLSv8
[8] Smolders K, Long H, Feng Y and Tavner PJ 2010 Reliability Analysis and Prediction of Wind Turbine GearboxesProc.
European Wind Energy Conference(Warsaw: EWEA)
[9] https://www.gearboxfailure.com/wp-content/uploads/2016/07/MicropittingOn_Misaligned_Gear.jpg [10]
http://media.noria.com/sites/archive_images/Backup_200101_Gear3.jpg?__hstc=108323549.07430159d50a3c91e72c280a 7921bf0d.1514764800116.1514764800117.1514764800118.1&__hssc=108323549.1.1514764800119&__hsfp=528229161 [11] https://www.novexa.com/en/intervention/gears/defects-treated.html
[12] http://www.rttech.com.au/wp-content/uploads/2010/06/mt6.jpg
R&D UK Centre
19
References
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
Alexios.Koltsidopoulos@edfenergy.com Acknowledgement:
This work is funded by the Energy Technology Institute and the Research Council Energy Programme as part of the IDCORE programme (grant EP/J500847).
R&D UK Centre
20
Questions
EERA DeepWind 2018 15th Deep Sea Offshore Wind R&D Conference
EERA DeepWind Conference 2018 18thJan 2018
A Turnbull1, E Hart1, D McMillan1, J Feuchtwang1, E Golysheva2and R Elliott2
1University of Strathclyde, Glasgow, UK
2Romax InSight, Nottingham, UK
Further investigation of the relationship between main-bearing loads and wind field characteristics
EERA DeepWind Conference 2018 2
Motivation
• Main-bearings seldom reach design life of roughly 20 years.
• Some failing after as little as 6 years [1].
• Reasons for this are still not fully understood.
• Cost associated with the repair is expensive.
• As we move further offshore, these effects are amplified due to cost of support vessels, weather and access restrictions.
EERA DeepWind Conference 2018 3
Research aims
1. Create a simple model which focuses on realistic input loads from which cause and effect can be easily separated.
2. Understand loading across wind turbine operating envelope and link this to wind field conditions.
3. Provide evidence to support claims that axial to radial load ratio is a key factor in main bearing failure.
EERA DeepWind Conference 2018 4
Aeroelastic model
• GH Bladed software used for aeroelastic wind turbine simulations.
• Wind field characteristics – 4 wind speeds (10, 12, 16, 20m/s) – 2 shear profiles (shear exponent 0.2, 0.6)
– 3 turbulence intensities (high, med, low as described in IEC standards [2] )
• 144 different wind fields to define operating envelope.
• Hub forces and bending moments extracted in all three degrees of freedom.
EERA DeepWind Conference 2018 5
Drivetrain model
• Drivetrain models generated for both double and single main bearing configuration.
• Separate model for radial and axial loads.
• Lengths and spring stiffness's determined by ROMAX Insight FEA modelling software for commercially available wind turbine of rated power around 2MW.
• Bearing type dependent on the configuration.
EERA DeepWind Conference 2018 6
Drivetrain model
EERA DeepWind Conference 2018 7
Results – Peak axial loads
EERA DeepWind Conference 2018 8
Results – Peak radial loads
EERA DeepWind Conference 2018 9
Results – Load ratio
Medium turbulence intensity and high shear
EERA DeepWind Conference 2018 10
Results – Load ratio
Effects of shear profile
EERA DeepWind Conference 2018 11
Results – Load ratio
Effects of turbulence intensity
EERA DeepWind Conference 2018 12
Conclusions
• Strong link between wind conditions and main bearing loads for both configuration – wind shear highest sensitivity factor.
• In general it can be observed that the double bearing configuration experiences a significant decrease in load ratio.
• Highest load ratio occurs in the single main bearing configuration in high shear and low turbulent conditions.
• With single main bearing configuration observed to fail more often, evidence suggests there could be link with load ratio.
EERA DeepWind Conference 2018 13
Potential impact of research
• Develop ways in which to bring the relationship into design stage when calculating component life, steering away from traditional methods of steady cyclic loading.
• Use relationship as a factor to support decision making of wind turbine type/configuration at particular site.
EERA DeepWind Conference 2018 14
Thank you for your attention, any questions?
S
Structural Change Identification at a Wind Turbine Blade using Model Updating
K. Schröder, S. Grove, S. Tsiapoki, C.G.
Gebhardt and R. Rolfes
EERA DeepWind’18, 18.01.18
DeepWind’18 18.01.18 2
C Content
I. Motivation
II. Optimization based model updating III. Rotor blade test
IV. Model updating at the rotor blade 1. Damage localization
2. Ice accretion V. Conclusion and Outlook
DeepWind’18 18.01.18 3
Motivation
Remote location
Rotor blades: costly and time-consuming repair
Ice accretion: - Risk of ice throw - Undesired loads
Localization and quantification of structural changes using model updating
DeepWind’18 18.01.18 4
Finite Element Model Updating
Damage event
DeepWind’18 18.01.18 5
Deviation between numerical model and measured data
Modal parameters
Transmissibility functions
• Eigenvalues
• Mode shapes
• Eigenvalues
• Mode shapes
Quantification of the „difference“ between model and measurement
DeepWind’18 18.01.18 6
Minimization of the deviation
• Nonlinear
• Constrained
• Nonconvex
• Several local minima Global optimization algorithm:
Simulated Quenching Local optimization algorithm:
Sequential Quadratic Programming
DeepWind’18 18.01.18 7
R
Rotor blade test
• Hammer excitation
• 12 measurement channels
DeepWind’18 18.01.18 8
• Hammer excitation
• 12 measurement channels
• Ice mass
• Damage
TTrailing edge bondline:
Spot of damage initiation Trailing edge – Pressure Side (outside)
Rotor blade test
DeepWind’18 18.01.18 9
N
Numerical Modeling
Rectangular Cross Section
Known: EI and mass
26 Timoshenko beam elements
Clamping at blade root
Material damping
DeepWind’18 18.01.18 10
N
Numerical validation
Stiffness reduction
DeepWind’18 18.01.18 11
N
Numerical validation–
Modal Parameters
Parameter number
Parameter number
DeepWind’18 18.01.18 12
Numerical validation – Transmissibility Functions
Parameter number
Parameter number
DeepWind’18 18.01.18 13
Ice accretion
4 steps
Variation of density
Optimization problem:
Step 3: 14,4kg at 32m-33m and 33m-34m
DeepWind’18 18.01.18 14
Ice localization – Modal Parameters
Correct Localizations in runs 1, 3, 7, 9 und 11
Verification using objective function value
Ice localization using modal parameters is possible
Parameter number
DeepWind’18 18.01.18 15
Ice quantification – Modal Parameters
StiffnessParameter 4 in %
Ice set (rotor blade mass in %)
0.1 0.3 0.6 0.9
DeepWind’18 18.01.18 16
Conclusion & Outlook
• Updating in numerical examples and for ice quantification successful
• Minimization using global two-step optimization algorithm
• No success for damage localization using measured data
• Modal parameters superior to transmissibility functionszationusususususuuuuuuuuuuuuuuuuuuuuuuuuuuuuusssssssssssssssssssssssssiniiiiiiiii g measururururuuuurrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrededededeeeeeeeeeeeeeeeeeeeeeeeeeeeeeedddddddddata trrrrrrranaaaaa smisssssssssssssssisisisisisisisisisisisisisisisssbibibibibibibibibibibbbbbbbbbbblilililililillliiiitytytytytytytytytyyyyyyyyfufufufufufufufufufufufufufufufufuuuuncncncncncncncncncncncncncncncncncncncnctictitititttttttttttttonooo s
• Investigate more advanced metrics for model updating
• Application to changing conditions (in situ) Conclusion
Outlook
DeepWind’18 18.01.18 17