Improving Component Reliability Through

Performance and Condition Monitoring Data

Analysis

Shawn Sheng
NREL

Wind Farm Data Management

& Analysis North America

March 25-26, 2015

Houston, TX
Photo by Dennis Schroeder, NREL 21883

NREL/PR-5000-64027
NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC.
Outline
 Introduction
• Reliability
• Reliability-Critical Turbine
Subsystems/Components
• Typical Failure Modes
 Data Analysis
• Performance Monitoring
• Condition Monitoring
 Case Studies
• Main Bearings
• Gearboxes
• Generators
 Concluding Remarks
• Summary
• Future Opportunities DOE 1.5 MW Turbine/PIX17245

NATIONAL RENEWABLE ENERGY LABORATORY 2

Introduction
Reliability [1]
 One Definition of Reliability:
• The probability that an asset or component will
perform its intended function without failure for a
specified period of time under specified conditions.

 Metrics: choose one to track reliability

improvements
• Mean time to repair or replace (MTTR) or mean time
between maintenance (MTBM)
• Mean time between failure (MTBF): primarily
repairable
• Mean time to failure (MTTF): primarily nonrepairable
• Total downtime
• Mean downtime
NATIONAL RENEWABLE ENERGY LABORATORY 4
Reliability of Turbine Subassemblies: Old Statistics [2,3]
Failure/turbine/year and downtime from two large surveys of land-based European wind turbines over 13 years

Electrical System
LWK Failure Rate, approx 5800 Turbine Years
Electrical Control WMEP Failure Rate, approx 15400 Turbine Years

Other LWK Downtime, approx 5800 Turbine Years

Hydraulic System WMEP Downtime, approx 15400 Turbine Years

Yaw System

Rotor Hub

Mechanical Brake

Rotor Blades

Gearbox

Generator

Drive Train

1 0.75 0.5 0.25 0 2 4 6 8 10 12 14

Failure/turbine/year Downtime per failure (days)

• The Wissenschaftliches Mess-und Evaluierungsprogramm (WMEP) database was

accomplished from 1989 to 2006 and contains failure statistics from 1,500 wind turbines.
• Failure statistics published by Landwirtschaftskammer Schleswig-Holstein (LWK) from
1993 to 2006 contain failure data from more than 650 wind turbines.
NATIONAL RENEWABLE ENERGY LABORATORY 5
Outage Duration for Different Components: New Statistics [4]
 Mechanical: Yaw Systems, Mechanical Brakes, Hydraulic
Systems, Rotor Hubs, Drivetrain
 Electrical: Sensors, Electrics , Control Systems
Scope of Discussion

Average Turbine Outage Duration

for Failures >1 Hour (days)

1 to 2 MW
<1MW
0
Gearbox Generator Blades Mechanical Electrical Others

NATIONAL RENEWABLE ENERGY LABORATORY 6

Typical Failure Modes: Gearbox Bearings

Micropitting Spalling

Scuffing
Axial Cracks Fretting Corrosion

Photo Credit: Robert Errichello, GEARTECH; Andy Milburn, Milburn

Engineering; Gary Doll, University of Akron; and Ryan Evans, Timken
NATIONAL RENEWABLE ENERGY LABORATORY 7
Typical Failure Modes: Gearbox Gears

Bending Fatigue Scuffing

(intermediate-stage pinion) (high-speed-stage pinion)

Micropitting Fretting Corrosion

Photo Credit: Rainer Eckert, Northwest Laboratory and Bob Errichello, GEARTECH
NATIONAL RENEWABLE ENERGY LABORATORY
Typical Failure Modes: Main Bearings

Debris Damage
Micropitting Roller End Thrust

Edge Loading Center Guide Ring Wear

Cage Failure

Photo Credit: Richard Brooks, Timken

NATIONAL RENEWABLE ENERGY LABORATORY 9
Typical Failure Modes: Generators

Magnetic Wedge Loss

Contamination

Electric Arc Damage Fluting

Photo Credit: Kevin Alewine, Shermco Industries; Gary
Doll, University of Akron; and Ryan Evans, Timken
NATIONAL RENEWABLE ENERGY LABORATORY 10
 Typical Failure Modes: Lubrication [5]
 Temperature  Foreign Materials
• Overloading • Wear particles
• Over greasing • Improper filtration
• Wrong viscosity • Poor lube storage
• Improper cooling methods
 Moisture • Poor lube
equipment storage
• Hot operation, then
shutdown  Viscosity
• Improper seals • Temperature
• Additive depletion • Oxidation
• Improper • Moisture/chemicals
vent/breather • No/lack of additives
device
• Leaking cooling
system
Photo Credit: Bill Herguth, SGS Herguth and Art Miller, EDF
NATIONAL RENEWABLE ENERGY LABORATORY 11
 Summary
 Major turbine components have
diverse and complex failure
modes.
 Wind turbine reliability
improvement is not a simple task:
• The number of
subsystems/components a turbine
has
• The modes of each
subsystem/component may fail

Tehachapi Pass Wind Farm, CA/NREL 18453

• The challenges with identifying root
causes for each failure mode
 Terminology challenge:
• Definitions of failure modes
• Definition of “failure” for different
subsystems/component. Illustration Credit: Jon Leather, Castrol
NATIONAL RENEWABLE ENERGY LABORATORY 12
Data Analysis
Performance Monitoring Based on SCADA [6]

NATIONAL RENEWABLE ENERGY LABORATORY 14

 Performance Monitoring
 Classification of Measured Parameters [7]:
• Wind parameters: e.g. speed, deviation
• Performance parameters: e.g. power output, rotor speed, blade pitch
angle
• Vibration parameters: e.g. tower acceleration, drivetrain acceleration
• Temperature parameters: e.g. bearing and gearbox temperature
 Grouping of Control System Status Report [8]:
• Status codes: e.g. error, warning
• Operating states: e.g. brake, start, yaw, pitch
 Analysis:
• Correlate different groups of parameters (e.g. power and wind),
develop models for normal operational states, and use these models
to identify abnormal scenarios
• Conduct statistical analysis of events (e.g. status codes) experienced
by turbines at a wind plant
• Investigate measured parameters under the same operating state
NATIONAL RENEWABLE ENERGY LABORATORY 15
Performance Monitoring [6,9]
 Benefits:
• Readily available and no need of investment in dedicated
condition monitoring (CM) instrumentation
• Beneficial for identifying outliers by looking at key
performance parameters or status codes
• Can call attention to turbines identified as outliers that may
need further inspection.

 Drawbacks:
• May not be straightforward in pinpointing exact damaged
subsystems/components (e.g. bearings or gears inside
gearboxes)
• Many false alarms due to varying loads experienced by
turbines
• Does not meet full turbine CM needs, such as fault
diagnosis
NATIONAL RENEWABLE ENERGY LABORATORY 16
Condition Monitoring Based on Dedicated
Instrumentation [6,10]

 Illustrated:  Additional:
• Blade Root Loads • Acoustic Emission
• Vibration • Electrical
• Shock Pulse Method
• Oil
• Thermography
NATIONAL RENEWABLE ENERGY LABORATORY 17
Condition Monitoring with Drivetrain as a Focus
 Raw Signal Examples:
• Strains, accelerations, acoustic emissions
• Oil debris counts, oil condition measurements
• Currents, voltages

 Feature (Condition Indicator) Examples:

• Preprocessing: filtering
• Time-domain: peak, root mean square
• Frequency-domain: gear meshing frequencies and sidebands, bearing
fault frequencies

 Typical Diagnosis:
• Trending or rate of changes of features or condition indicators
• Appearance of frequency components corresponding to certain faults
or abnormal modulation of signal spectra
• Violating thresholds set for certain features

 Typical Prognosis:
• Data-driven models: regressions, neural networks
• Empirical or physics-based models: crack propagation by Paris Law

NATIONAL RENEWABLE ENERGY LABORATORY 18

Condition Monitoring
 Benefits:
• Capture of high frequency dynamics normally not achievable
with a typical SCADA system
• Identification of more failure modes occurred to turbine
subsystems or components
• Capability in pinpointing exact damaged
locations/components
• Enable condition or reliability-based maintenance

 Drawbacks:
• Additional investment required for instrumentation and
monitoring service
• Dedicated resources on data analysis and interpreting
results

NATIONAL RENEWABLE ENERGY LABORATORY 19

Improved Component Reliability
 Performance and
condition monitoring
data analyses are two
tools to achieve fault
diagnosis

 Immediate impacts on
O&M actions leading
to improved turbine
availability, an indirect
measure of reliability

 Root cause analysis, if

conducted to identify
faults, and addressing
the root causes can
lead to direct
improvement in
component reliability

NATIONAL RENEWABLE ENERGY LABORATORY 20

Case Studies
Main Bearings: Performance Monitoring [11]

Performance monitoring diagram

All data: power vs. wind speed Filtered data: power vs. wind speed
22
 Main Bearings: Performance Monitoring [11]

Power vs. wind speed color coded by air density Cumulative sum of residual for a turbine that was
fault-free through 12-month test period
 Modeling turbine power using both
wind speed and air density reduced
root mean squared error by 16%
 Fault free: cumulative sum of residual
oscillates about a value of zero
 Temperature trending: typically
Cumulative sum of residual for a turbine with a reliable for failure identification but
main bearing failure and replacement may be too late to save the bearing.
23
Main Bearings: Condition Monitoring [12]

24
Main Bearings: Condition Monitoring [12]
 Vibration analysis based
on accelerometers is
feasible but may
present challenges.

25
Gearboxes: Performance Monitoring [13]
 High-speed shaft (HSS) ratio: HSS
torque to HSS rpm
 Model developed based on normal
operation
 Thresholds established based on a
certain allowable false alarm rate
 Two angles: response and residual
 Abnormal: outside of the established
thresholds

26
Gearboxes: Condition Monitoring [14]
1. Completed dynamometer run-in test
2. Sent for field test: experienced two oil losses (root cause)
3. Stopped field test
4. Retested in the dynamometer under controlled conditions

High-Speed Stage Gear Damage

Photo by Lee Jay Fingersh, NREL 16913 Photo by Robert Errichello, NREL 19599

NATIONAL RENEWABLE ENERGY LABORATORY 27

 Gearboxes: Vibration Analysis [15]
 Intermediate-speed shaft
sensor
 Dynamometer test of the
same reference gearbox
(left) indicated healthy
gearbox behavior

 Various condition
 Dynamometer retest of the indicators can be
damaged gearbox (right) defined to ease the
indicated abnormal behavior fault diagnostics
• More side band frequencies process.
• Elevated gear meshing
frequency amplitudes
NATIONAL RENEWABLE ENERGY LABORATORY 28
 Gearboxes: Vibration Analysis [16]

 Inspected by EDPR 11/11/2014 – HS GS Bearing Axial

Cracks
 NOTE: Previously inspected June 2014 with no
findings
 HS bearings replaced 12/16/2014

Photo Credit: Adam Johs, EDPR

NATIONAL RENEWABLE ENERGY LABORATORY 29
Gearboxes: Oil Debris Monitoring [17]

Oil debris during the test of a healthy test gearbox Oil debris during the test of a damaged test gearbox

 Damaged gearbox shed debris much faster:

• Left (healthy gearbox): about 1 particle per hour
• Right (damaged gearbox): 70 particles per hour

 Caution:
• Rely more on the averaged particle generation rates than those
calculated in real time
NATIONAL RENEWABLE ENERGY LABORATORY 30
 Gearboxes: Oil Sample Analysis [18]
 Results: dynamometer test of the reference gearbox
• Particle counts: important to identify particle types

• Element identification

Reference Limits Analysis Results

NATIONAL RENEWABLE ENERGY LABORATORY 31
 Generators: Performance Monitoring [19]
 Autoregressive Neural Network
Model:
• Inputs: stator temperature;
power output; nacelle
temperature; and ambient
temperature
• Output: stator temperature
• Error: modeled temperature vs.
Time series of the stator temperature on a generator measurements

 Damage: generator replacement

 Model Performance:
• Accuracy: ±1°C
• First alarm violation: 59 days
ahead
• Second alarm violation: 48 days
ahead
Averaged prediction error for the autoregressive neural
network model

32
Generators: Condition Monitoring [20]

GROUNDING THROUGH
PEAK AT 2x ELF THE BEARING
(7200 CPM)

 Electric signature analysis

based on numerical simulations
or small-scale test rigs not
much on utility-scale turbines.
33
Summary: Performance Monitoring
 Most data analysis techniques can identify abnormal
behaviors

 Nonlinear modeling approaches may be more accurate

for wind turbine applications. Neural networks-based
models are hard to generalize

 Wind speed, power output, and various temperatures

are the main parameters investigated. Combining wind
speed with air density can improve modeling accuracy

 Temperature is typically a reliable indicator of

component failure but may not provide enough lead
time to save the monitored component.
34
Summary: Condition Monitoring
 Most condition monitoring data analysis techniques
can help pinpoint specific subsystems/components
with faults

 Vibration analysis appears to be the most widely

investigated and reported technique. It can monitor
the health of most drivetrain, and even turbine,
subsystems/components

 Oil debris counting results are easier to interpret and

provide unique information on gearboxes (typically the
only oil-lubricated subsystem in a wind turbine)

 Shock pulse method may be more effective for the

low-speed stage in wind turbine.
35
Concluding Remarks
Summary
 Improving turbine component reliability is not a simple task:
• Complexity of turbines operating in harsh environments
• Diverse subsystem/component failure modes with inconsistent
definitions
• Identification of and addressing root causes is time consuming or
challenging
• Performance and condition monitoring data analysis can help

 Performance monitoring data analysis:

• Readily available measured parameters and status codes
• Initial screening to identify abnormal turbine behaviors
• Not enough to meet full turbine condition monitoring needs

 Condition monitoring data analysis:

• Covers more failure modes than typical performance monitoring data
analysis
• Pinpoints damaged locations/components and enables condition based
maintenance
• Requires additional investment for instrumentation and resources for
data analysis or results interpretation
37
Future Opportunities
 Field application feasibility study of various data
analysis techniques; if not yet feasible but deemed
beneficial, investigate enabling approaches

 Data analysis or modeling work that enables remaining

useful life estimation of turbine
subsystems/components

 Fusion with additional data streams to improve

operation & maintenance practices, reduce loads and
extend life of turbine subsystems/components

 Conduct root cause analysis, when feasible and

economical, address root causes in the field, and
provide feedback to subsystem/component suppliers
for reliability improvement of future products.
38
References
1. [Link]
2. Feng, Y. and Tavner, P. (2010). “Introduction to Wind Turbines and Their Reliability & Availability,”
presented at the European Wind Energy Conference, Apr. 20-23, Warsaw, Poland.
3. Tavner, P., Spinato, F., van Bussel, G.J.W. and Koutoulakos, E. (2008). “Reliability of Different
Wind Turbine Concepts with Relevance to Offshore Application,” presented at the European
Wind Energy Conference, Mar. 31 – Apr. 3, Brussels, Belgium.
4. Chamberlain, K. (2015). “WEU Onshore Asset Optimization & Reliability Benchmarking Report
2015.”
5. 50 Lubrication Failure Modes. [Link]
6. Yang, W., Tavner, P. J., Crabtree, C. J., Feng, Y. and Qiu, Y. (2014). "Wind Turbine Condition
Monitoring: Technical and Commercial Challenges," Wind Energy, 17(5): 673-693.
7. Wang, K. S., Sharma, V. S. and Zhang, Z. Y. (2014). “SCADA Data based Condition Monitoring of
Wind Turbines,” Advances in Manufacturing, 2:61-69.
8. Castellani, F., Garinei, A., Terzi, L., Astolfi, D., Moretti, M. and Lombardi, A. (2013). “A New Data
Mining Approach for Power Performance Verification of An Onshore Wind Farm,” Diagnostyka,
14(4): 35-42.
9. Sheng, S. and Yang, W. (2013). “Wind Turbine Drivetrain Condition Monitoring - An Overview,”
NREL Report No. PR-5000-58774.
10. García Márquez, F. P., Tobias, A. M., Pinar Pérez, J. M. and Papaelias, M. (2012). "Condition
Monitoring of Wind Turbines: Techniques and Methods," Renewable Energy, 46, October: 169-
178.
11. Butler, S., Ringwood, J. and O'Connor, F. (2013). “Exploiting SCADA System Data for Wind Turbine
Performance Monitoring,” In: Proceedings of the 2013 Conference on Control and Fault-Tolerant
Systems, Oct. 9-11, Nice, France.
39
References
12. Brooks, R. (2013). “Condition Monitoring Methods for Wind Main Shaft and Gearbox Planetary
Sections,” Wind Energy Workshop at the Annual Conference of the Prognostics and Health
Management (PHM) Society, Oct. 14-17, New Orleans, LA.
13. Yampikulsakul, N., Byon, E., Huang, S., Sheng, S. and Yu, M. (2014). "Condition Monitoring of Wind
Power System With Nonparametric Regression Analysis," IEEE Transactions on Energy Conversion,
29(2): 288-299.
14. Sheng, S. and Yang, W. (2013). “Wind Turbine Drivetrain Condition Monitoring – An Overview,”
ASME Turbo Expo 2013, San Antonio, TX, June 3-7.
15. Sheng, S. (2011). “Investigation of Various Condition Monitoring Techniques Based on a Damaged
Wind Turbine Gearbox,” 8th International Workshop on Structural Health Monitoring 2011
Proceedings, Stanford, CA, September 13-15.
16. Johs, A. (2015). “EDPR North America – Condition Monitoring System Retrofits, An Operator’s
Experience,” presented at the NREL Gearbox Reliability Collaborative Annual Meeting, Golden,
CO.
17. Sheng, S. (2015). “Wind Turbine Condition Monitoring, Reliability Database, and O&M Research
Update,” NREL Report No. PR-5000-63868.
18. Sheng, S. (2011). “Investigation of Oil Conditioning, Real-time Monitoring and Oil Sample Analysis
for Wind Turbine Gearboxes,” presented at the 2011 AWEA Project Performance and Reliability
Workshop, January 12–13, 2011, San Diego, CA.
19. Schlechtingen, M. and Ferreira Santos, I. (2011). "Comparative Analysis of Neural Network and
Regression based Condition Monitoring Approaches for Wind Turbine Fault Detection,"
Mechanical Systems and Signal Processing 25(5): 1849-1875.
20. Clark, D. (2011). “Condition Monitoring and SCADA Data for Performance Improvements,” Wind
Turbine Condition Monitoring Workshop, Broomfield, CO, Sep. 19-21.
40
Thanks for Your Attention!
Special thanks go to the U.S. Department of Energy, the condition
monitoring and O&M research partners!

NREL’s contributions to
this presentation were
funded by the Wind and
Water Power Program,
Office of Energy Efficiency
and Renewable Energy,
the U.S. Department of
Energy under contract No.
DE-AC02-05CH11231.
The authors are solely
Photo by HC Sorensen, Middelgrunden Wind Turbine Cooperative, NREL 17855
responsible for any
omissions or errors
[Link]@[Link]
contained herein.
303-384-7106
41