According to a report by Windpower Engineering & Development, the cost to repair main bearings is one of the highest compared to other turbine systems ranging from $150,000 to $300,000. As fielded turbines age, the aggregated downtime has increased to more than 20,000 hours.
The presentation features:
- New capabilities in the DigitalClone Live software that predict early wear initiation in the main bearing raceways and rollers.
- How to understand the current health of the main bearing in a particular asset to mitigate damage prior to failure.
- How to reduce the cost of repair, currently ranging from $150K - $300K for each occurrence, and reduce the amount of downtime.
View the Webinar Recording at:
http://sentientscience.com/resource-library/videos/understanding-main-bearing-failures-mitigating-a-150k-300k-om-cost-2/
6. SentientScienceCorp.-Proprietary/PrivateLevel1
Main Bearing Wear Model
Main Bearing Wear Model
01
Main Bearing
Business Challenge
Understanding the cost and
risk associated with main
bearing failure.
02
Why Main
Bearings Fail
Outlining premature failure
and common failure modes.
03
The Main Bearing
Wear Model
A deep dive into Sentient’s
DigitalClone Live main
bearing wear model.
04
The DigitalClone
Solution
Deploying DigitalClone to
provide life extension
actions.
8. SentientScienceCorp.-Proprietary/PrivateLevel1
Why Investigate Main Bearings?
9
High Risk:
Main bearing failure often
transfers thrust loads to the
gearbox, which can lead to
catastrophic damage in the
gearbox.
High Cost:
Main Bearing Replacement
ranges from $150K - $300K. In
extreme cases can cause
upwards of $500K.
Uncertainty:
Alerted of main bearing failure after
SCADA temperature alarms are
going off.
Supply Chain Control:
Lack of control in supply chain due
to no lead time for main bearing
failures, causing unexpected
downtime.
Main Bearing Wear Model
9. SentientScienceCorp.-Proprietary/PrivateLevel1
Average Cost of Main Bearing Replacement
Main Bearing Wear Model
Cost Breakdown:
Challenge:
1. Limited understanding of how many
failures and where
2. Unaware of why the main bearing has
failed
Solution:
1. Life extension actions to avoid
unexpected downtime and mitigate
replacement costs
Category Cost
Downtime $10 -$30K
Labor $15K - $35K
Bearing Replacement $38K - $50K
Main Shaft Repair $50 - $120K
Average High Cost to Repair: $335K
Average Low Cost to Repair: $190K
12. SentientScienceCorp.-Proprietary/PrivateLevel1
Premature Failure in Main Bearings
Main Bearing Wear Model
Three Point Configuration
Spherical Roller Bearing
INNER RACE, OUTER RACE,
ROLLER
CAGE, GUIDE RINGS
High Radial load capacity
Accommodates misalignments
Require relatively high ratio of
radial to axial load
Increased sliding due to
Heathcote slip
Thrust Load
Bearing Design
Low RPM
Loss of Lubrication
Uneven Load Distribution
Surface Roughness
Lubricant Viscosity
High Pressure
High Sliding
Low Lambda
Low Lambda
13. SentientScienceCorp.-Proprietary/PrivateLevel1
Failure Modes in Main Bearings
Main Bearing Wear Model
Wear on Inner
Race
Micropitting
Macropitting
Spalling
Abrasive Damage Roller, Cage Crack
FAILURE PROGRESSION IN MAIN BEARINGS (manifests as early as 6 TO 10 YEARS)
Sentient’s Main Bearing Failure model identifies early manifestation of damage in Main Bearings and provides
recommendations to slow down damage progression, such that the Remaining Useful Life of the bearing can be extended
Asperity Plastic Deformation
Adhesive wear, Surface Fatigue
Wear Bands on Inner Race
Incubates in the vicinity of wear
bands
Cyclic shear stresses at
shallow depths below the
asperities
Loss of geometry due to
significant micropitting
High contact stresses, edge
loads
Debris dents the race and
rollers
Rapidly evolves and
accelerates damage
Exceptions from the typical failure progression is also observed. White Structure Flaking (WSF) caused due to material impurities, hydrogen
embrittlement, cage and guide ring wear leading to cage cracks, roller turning. The main bearings can fail as early as 4 to 5 years
15. SentientScienceCorp.-Proprietary/PrivateLevel1
Bearing Life Analysis – Conventional vs Prognostics
16
TRADITIONAL APPROACH – FEA/EMPIRICAL EQUATIONS
SENTIENT APPROACH – MATERIAL BASED PROGNOSTICS
Quick evaluations of design/material/lubricant change
Cost effective and scalable solution
Saves time and resources
Large number of outputs generated in short time
Assess effect of parameters which are experimentally
difficult
Deterministic
Bearing Life
Macroscale
wear model
(FEA)
Wear
Coefficient
Conduct
Experiments
Probablistic
Bearing Life
(L10/L50)
Wear
Location/Dep
th/Coefficient
Contact
Mechanics
Lubrication
Model
Material/
Physics based
Wear Model
16. SentientScienceCorp.-Proprietary/PrivateLevel1
Main Bearing Wear Model
DCL Main Bearing Model Strategy
WINDLOAD MODEL BEARING DYNAMICS MODEL WEAR MODEL
SCADA data
Turbulence Intensity
Windspeed
Wind Direction
Loads on Main shaft
Bearing Configuration Geometry
Bearing Clearances
Material Elastic modulus
Main shaft RPM
Contact Pressure
Sliding Velocity
Material : Hardness, Elastic Modulus, Ultimate
Strength
Surface topography : 𝑅 𝑞, 𝑅 𝑠𝑘, 𝑅 𝑘𝑢, 𝛽 𝑥, 𝛽 𝑦
Lubricant: Viscosity, PV Coeff
Main Bearing Risk Ranking [MTOD]
Critical locations
Life Extension Action
Wear Rate
Wear Coefficient
Inputs
Outputs
Local Contact Forces
Sliding velocities
Equivalent Radii
Overturning Moments
Thrust, Radial Load
17. SentientScienceCorp.-Proprietary/PrivateLevel1
Reverse Engineering – Geometry & Surface Topography
Main Bearing Wear Model
Sentient conducts an in-depth analysis of bearing samples received from the customer to enlist the input parameters required
for physics-based models (Geometry, Surface topography, Hardness, Microstructure)
Bearing
Sample
IR Roller OR
𝑹 𝒒 0.41 0.46 0.39
𝑹 𝒔𝒌 -0.93 -0.35 -1.32
𝑹 𝒌𝒖 4.28 3.25 5.23
𝜷 𝒙 23.93 25.54 12.43
𝜷 𝒚 533.51 136.45 9.60
Example of Surface Topography analysis
Example of Geometry measurements on inner race of an asymmetrical SRB
18. SentientScienceCorp.-Proprietary/PrivateLevel1
Reverse Engineering – Micro-hardness, Microstructure
Main Bearing Wear Model
Microhardness comparison of five different bearing suppliers/designs
Example of Microstructure Analysis (showing an inclusion and martensitic phase)
A first hand comparison of different suppliers can be
provided based on surface roughness, micro-hardness and
material cleanliness.
However, a more detailed assessment based on failure
models is required to draw definitive conclusions.
19. SentientScienceCorp.-Proprietary/PrivateLevel1
Bearing Analysis Tool (BAT)
Main Bearing Wear Model
rotor side generator side
• largely unloaded
upwind rollers
• intermittent contacts
continuous loading of
downwind rollers
sliding velocity
distribution
load distribution
Roller to Race contact details
Input to Wear Model
20. SentientScienceCorp.-Proprietary/PrivateLevel1
Location of Wear/Micropitting Bands
Main Bearing Wear Model
1
11
21
310
100
200
300
400
2
35
68
101
134
168
201
234
267
301
334
Wearmetric[N/s]
1
11
21
310
100
200
300
400
2
35
68
101
134
168
201
234
267
301
334
Wearmetric[N/s]
Downwind Inner Race
Downwind Outer Race
Samples received by SentientKotzalas & Doll, 2010
UW UWDW DWDWEarly Stage
Sentient’s main bearing model can predict the circumferential and axial
location of wear on the main bearing.
• The inner race is more likely to wear compared to the outer race
• The upwind side is unworn while the downwind side is severely worn
• Wear bands are seen as two distinct bands around the center of downwind
side race or at the center depending on bearing design and the loads
supported by the bearing
Wear Peaks
Wear Indicator = Pressure x Sliding Velocity
21. SentientScienceCorp.-Proprietary/PrivateLevel1
Outputs of Wear Model
Main Bearing Wear Model
Wear Rate (m/cycle) vs Pressure and
misalignment
Wear Coefficient vs Pressure and misalignment
Example wear outputs from the analysis of a 230/600 SRB Main Bearing
Wear evolution of interacting surfaces
23. SentientScienceCorp.-Proprietary/PrivateLevel1
Effect of Surface Roughness
Main Bearing Wear Model
𝑹 𝒒 = 𝟎. 𝟐 𝝁𝒎
𝑹 𝒒 = 𝟎. 𝟒 𝝁𝒎
Rq = 0.2
Rq = 0.4
5.39E-11
4.85E-09
4.66E-11
8.74E-09
0.00E+00
1.00E-09
2.00E-09
3.00E-09
4.00E-09
5.00E-09
6.00E-09
7.00E-09
8.00E-09
9.00E-09
1.00E-08
0.2 0.4
WEARRATE(M/CYCLE)
ROUGHNESS
Grease1 Grease2
Comparison of worn surfaces
Contact Index (Red indicates asperity contact)
Comparison of Roughness and Lubricant
Contact Pressure Profiles
𝑹 𝒒 = 𝟎. 𝟒 𝝁𝒎
𝑹 𝒒 = 𝟎. 𝟐 𝝁𝒎
Surface Roughness has a major impact on the
wear rate. Depending on lubricant viscosity and
relative velocity, increased asperity contacts
cause increase in wear rates
24. SentientScienceCorp.-Proprietary/PrivateLevel1
Effect of Hardness
Main Bearing Wear Model
H=2.5 GPa (24 HRC)
WR = 2.26 x 10-6 m/cyc
H=5 GPa (49 HRC)
WR = 1.12 x 10-6 m/cyc
H=7 GPa (60 HRC)
WR = 8.16 x 10-7 m/cyc
Hardness is an important property which affects wear rate. Wear Rate reduces with increasing hardness in agreement
with experiments.
y = 4E-10x2 - 7E-08x + 4E-06
R² = 0.9996
0.0E+00
5.0E-07
1.0E-06
1.5E-06
2.0E-06
2.5E-06
0 10 20 30 40 50 60 70 80
WearRate(m/cyc) Hardness (HRC)
H=9 GPa (67 HRC)
WR = 6.234 x 10-7 m/cyc
25. SentientScienceCorp.-Proprietary/PrivateLevel1
Effect of Temperature
Main Bearing Wear Model
𝑇 = 70 𝑂
𝐶, 𝑊𝑅 = 6.971 𝑥 10−8
𝑚/𝑐𝑦𝑐
𝑇 = 50 𝑂
𝐶, 𝑊𝑅 = 3.791 𝑥 10−8
𝑚/𝑐𝑦𝑐
50 𝑂 𝐶
70 𝑂
𝐶
50 𝑂
𝐶
70 𝑂
𝐶
RPM = 20 RPM = 3
Temperature can increase due to dry contact and/or environmental
conditions.
A linear increase in temperature causes an exponential decrease in viscosity
which can be detrimental to the two surfaces in contact.
At higher RPMs, the contacts are more sensitive to temperature because the
preexisting oil film is being degraded. At lower RPMs, the oil film is non-
existent and therefore the contacts are less sensitive.
Comparison of worn surfaces
Contact index showing the interaction between RPM and temperature
26. SentientScienceCorp.-Proprietary/PrivateLevel1
Effect of Main Shaft RPM
Main Bearing Wear Model
RPM = 5
RPM = 15
FILM THICKNESS PROFILE ASPERITY CONTACTS
ASPERITY CONTACTSFILM THICKNESS PROFILE
230/600 Bearing
𝜂 = 0.54,0.42,0.34,0.17 𝑃𝑎. 𝑠
𝜆 =
ℎ
(𝑅 𝑞1
2
+𝑅 𝑞2
2
)
𝜆 < 1: Boundary Lubrication
LAMBDA RATIO
Lower main shaft RPMs (<5 RPM) can be very detrimental to raceway and rollers as the bearing operates in boundary
lubrication regime.
27. SentientScienceCorp.-Proprietary/PrivateLevel1
Wear Modeling - Elastic Plastic Effects
Main Bearing Wear Model
𝜎
𝜖
𝑆 𝑦
𝐸
𝑀
Although the maco-scale/theoretical contact
pressure observed in bearings is maximum
of 1.5 GPa, asperity interactions due to
surface roughness cause contact pressure
to reach as high as 3 to 5 GPa. To
accommodate for such higher pressures, an
elastic-plastic constituent model is required.
Comparison of contact pressures observed for an elastic-plastic model vs purely elastic model
ELASTIC 𝑆 𝑦 = 1.5 𝐺𝑃𝑎 (Unhardened Steel)
𝑆 𝑦 = 7𝐺𝑃𝑎 (Hardened Steel, 60 HRC) 𝑆 𝑦 = 9𝐺𝑃𝑎 (Hardened Steel, 65 HRC)
28. SentientScienceCorp.-Proprietary/PrivateLevel1
Wear Modeling – Variable Coefficient of Friction
Main Bearing Wear Model
𝑅 𝑞 = 0.17𝜇𝑚, 𝜇 = 0.3571 – 0.408 𝑅 𝑞 = 0.0017𝜇𝑚, 𝜇 = 0.3571 – 0.3573
𝑅 𝑞 = 0.017𝜇𝑚, 𝜇 = 0.3571 – 0.3626 𝑅 𝑞 = 0.5𝜇𝑚, 𝜇 = 0.3571 – 0.559
0.3
0.35
0.4
0.45
0.5
0.55
0.6
0 0.1 0.2 0.3 0.4 0.5 0.6
COF
RMS Roughness
COFmin COFmax
DMT
Coefficient of friction is not assumed to be a
constant but changes according to the geometry of
the contacting asperities and surface energy of the
material based on the DMT model of adhesion
Depiction of variable coefficient of friction at asperity contacts
31. SentientScienceCorp.-Proprietary/PrivateLevel1
Actions Provide Life Extension & Value Across All Health States
Hot Oil Flush, Re-grease
Major Component
Exchange
Axially Realign the Bearing Races
Monitor & Manage Main Bearing Loads
Recommendations to improve reliability and minimize downtime:
1. Perform grease sampling, flush (Hot Oil) and re-grease on an annual basis
1. Measure for changes in axial displacement on an annual basis and consider methods to axially re-align the bearing races. (The available
methods to re-align the races of the mainshaft bearing include re-shimming the bearing and/or the housing and warrant further discussion)
2. Consider blade pitch and rotor thrust measurements and optimization methods to assure blade pitch consistency and rotor thrust values are
not in excess of OEM specification (to minimize axial displacement forces)
32. SentientScienceCorp.-Proprietary/PrivateLevel1
Average Cost of Main Bearing Replacement/Savings
Main Bearing Wear Model
Cost Breakdown:
Life Extension Actions:
1. Planning replacement
during low wind season
2. Ensuring MB availability
3. Reducing MB repair costs
4. Ensuring replacement
component has optimal
life
5. Supplier trade-off and
impact on asset life
Category Cost % Savings
Downtime $10 -$30K 0 - 75%
Labor $15K - $35K 10 - 25%
Bearing Replacement $38K - $50K 10 - 25%
Main Shaft Repair $50 - $120K 5 - 15%
Average High Cost to Repair: $335K Up to ~$62K
Average Low Cost to Repair: $190K
INTRO –
Started with by predicting the GBX life – biggest issue we were seeing in the market
Now we are expanding our offerings starting with the main bearing wear model
We will then move into blades, pitch bearings, generators and other components.
Currently - 32 Main bearing projects – where we are projecting failure rates in the specific machines
We are looking for additional projects for our validation testing if anyone is interested, id be happy to help.
Recording - link
So, how are we different?
We provide life extension actions for specific assets – MATERIAL SCIENCE APPROCAH .
Which allows us to predict early crack initiation for critical components.
Give customers data they need to make better operating decisions ahead of time.
We have been around since 2001 where we did R&D for several years. For the past 3 years, we have been very focused on the servicing the wind industry for both operators and suppliers globally.
Talk about different industries.
Aerospace
Rail
Industrial
Thank you that feedback was helpful and will allow us to tailor our presentation today.
High Cost: Main Bearing Maintenance is often overlooked during annual services . Most operators put a higher focus on GBX failures, ironically MB failures can be most costly because repairs can never be done uptower.
High Risk: Main bearings often take the load off of the GBX. If you knew the MB was failing ahead of time, you could repair to avoid risk to GBX. This occurs when Main Bearing fails to support thrust loads causing misalignment. Sustible to failure after the MB fails
Supply Chain Issue – Main bearings are very complex bearing. Supply chain needs time to create these bearings. If we could do a better job at planning how many need to be replaced, we could reduce unplanned downtime.
Uncertainty: Most operators are only aware of MB failures after SCADA temperature alarms are going off – too late.
When working with our operators to help extend the life of their assets, we found that MB failures are a significant portion of unplanned O&M costs.
Helping to mitigate these costs with the wear model by providing life extension recommendations.
Dr. Arnab G-hosh
Arnab Ghosh, obtained his PhD from Purdue University
His research experience is in computational modeling of surface initiated tribological phenomenon
At Sentient, he is responsible for the development of wear models
Contribute to theoretical development, digital experiments
Sentient applies physics-based models to individual turbines
Understand different turbine health and specific components health within the turbine.
From this we provide component recommendation.
Once we understand the RUL of the component, we provide specific recommendations to extend the life of the asset.
Depending on how much life we see left in the component, we provide different recommendations.
Summarize
High cost of main bearing replacement – impacting unplanned O&M budgets
With the MB wear model, we identify what turbine needs help.
We digitally model different maintenance activities.
For example, we can help you understand the ROI for lubrication treatments that can help push out failures.