1 of 35

Wind Turbine Maintenance Management

OSIsoft Data Management

Senior Project Presentation

Advisor:

Participants:

1

Dr. Javad Seif

Daniel Davila, Robert Garcia

Xavier Reyes, Young Woo

2 of 35

2

Improve maintenance through

data monitoring

Describe current situation

Observe wind farm’s performance

Design SPC control chart for monitoring

Use MRP to proactively maintain wind farm

3 of 35

Define

What Is the Main Idea?

Data management company

Australian energy company

Develop wind turbine maintenance plan for client

3

4 of 35

Define

Who Exactly Are We Helping?

AGL Energy Ltd.

One of Australia’s largest energy companies
Generation capacity of 11 GW
OSIsoft customer for 10+ yrs
OSIsoft helps monitor, operate, and optimize its fleet of generation assets

Photo Source: AGL, Macarthur Wind Farm

4

5 of 35

Define

What Is the Goal?

Photo Source: OSIsoft

Prepare a maintenance plan framework for AGL

Knowns:

5 clusters of 10 turbines (50 total)
Two years of data (2018-2019)
13 different sensors per turbine

5

(-) Downtime

→ (+) Energy Output

→ (+) Revenue

→

6 of 35

Define

Outline

Conduct failure analyses
Model turbine availability
Simulate energy output for validation
Develop reliability model
Create MRP for spare parts
Perform cost analysis
Design control chart for monitoring critical turbine attributes
Craft and present recommendations

6

7 of 35

Measure

Data Handling with Python

7

Transition Matrix

Correlation Matrix

8 of 35

Measure

Failure Analysis

Photo Source: Enel Green Power

8

Accurately model wind turbine availability by investigating:

Causes of turbine failure
Changes in operational state (“OK”, “TurbError”)
State of turbine during both short- and long-term
Turbine uptime across whole system

9 of 35

Measure

Failure Modes

Photo Source: Chiu Yi-tung, Taipei Times

9

Common causes of turbine failure:

Damaged blades
Bearings failure in gearbox, generator

Turbines experience frequent failure within 20-year lifespan
Harsh weather causes significant wear
Routine maintenance key to reducing failure

Temperature / vibration readings, material quality inspections

10 of 35

Measure

Fault Tree Diagram

10

A very sensitive system!

11 of 35

Measure

How Reliable Are Wind Turbines?

Photo Source: BBC

11

Turbines prone to failure

Components rely on one another

Damage mostly occurs externally

Environmental damage (i.e. weather)
Too much wind

Existing reliability data limited

Most turbines installed < 20 years (the average turbine lifespan)

12 of 35

Analyze

Calculating Wind Turbine Reliability

12

R₃

Electrical System

Control System

T-Structure

Generator

Hydraulic System

Brake

Gearbox

Convertor

Rotor

Main Shaft

Pitch System

Reliability =

(R₂× R₈× R₅× R₁₀× R₆× R₃× R₄× R₁₁× R₁₂× R₁) × [1 - (1-R₇) × (1-R₉)]

Yaw System

R₄

R₁₁

R₁₂

R₁

R₃

R₉

R₆

R₁₀

R₂

R₈

R₅

13 of 35

Analyze

Markov Chain

p₁₁ ≡ Turbine remains operational

p₁₂ ≡ Turbine fails

p₂₁ ≡ Turbine is repaired

p₂₂ ≡ Turbine remains unoperational

13

Definition:

Turbine Markov Chain:

14 of 35

Analyze

Steady-state

Notes:

n = 15 days

S₀ ≡ Initial state matrix (assume “OK”)

14

15th State Matrix:

→ 95% → OK, after 15 days

→ 5 % → TurbError, after 15 days

Definition:

15 of 35

Analyze

Steady-state Distribution

Equations for obtaining steady-states (solving system of equations)

15

Setup:

Steady-state Distribution:

→ 95% → OK, in long run

→ 5 % → TurbError, in long run

16 of 35

Analyze

How Reliable Are Wind Turbines?

16

Example calculations with Cluster 1 turbines
Time Range t = 7 days (weekly)
Repair times rounded as integers → ± 0.5% uptime

Asset_Id	MTBF (Days)	MTTR (Days)	MTTF (Days)	Failure Rate λ	λt	Reliability	Uptime %
cluster1.turb1	30	1	13	0.03	0.23	79%	97%
cluster1.turb10	23	1	13	0.04	0.30	74%	96%
cluster1.turb2	24	3	3	0.04	0.29	75%	89%
cluster1.turb3	21	0.6	7	0.05	0.33	72%	97%
cluster1.turb4	21	1	13	0.05	0.33	72%	95%
cluster1.turb5	31	0.6	13	0.03	0.23	80%	98%
cluster1.turb6	25	0.4	13	0.04	0.28	76%	98%
cluster1.turb7	18	0.7	13	0.06	0.39	68%	96%
cluster1.turb8	30	0.8	13	0.03	0.23	79%	97%
cluster1.turb9	25	0.2	13	0.04	0.28	76%	99%

17 of 35

Analyze

Turbine Availability

17

Average cluster* uptime ≈ 96%
Average turbine reliability ≈ 65%
Least reliable clusters: 2, 3
Need to investigate failure causes

Vibration likely to contribute to failures

*One cluster comprises 10 turbines

18 of 35

Analyze

Simulation of Expected Energy Output

Considerations

All 50 turbines in parallel
Failure rate λ derived from MTTF
Ran simulation for 30 days and for 12 cycles to improve accuracy

18

19 of 35

Analyze

Simulation Results

Results

≈ 29 days of generation per turbine, on average
≈ 2 days of downtime per turbine, on average
Expected monthly output of 1.4B kW
Expected 93M kW lost to failures monthly

19

20 of 35

Improve

Proactive Solution

20

Monitor:

Turbine condition through data

Correct

Prevent and predict failure
Reduce failure rate

MRP

For spare parts

Cost saving

21 of 35

Improve

Proactive Maintenance

Photo Source: Rodolfo C Saavedra and Biswanath Samanta, Wind Engineering, Vol. 39 No. 6

21

Vibration a significant attribute to monitor

Excessive vibrations linked to higher probability of failure

Caused by extreme weather, rigid turbine structure, aerodynamic imbalances
Need to check if vibration data is in control

22 of 35

Improve

Bill of Materials (BOM) for Material Requirement Planning (MRP)

22

Modeled first two levels of BOM:

Level 0: Turbine

Level 1: Turbine Subsystems

23 of 35

Improve

Forecasting Failures

23

Forecast number of failures the farm would experience eight weeks in the future:

Constant forecasting most accurately represented data
Average of ≈19 (18.17) failures per week used for Master Production Schedule

24 of 35

Improve

Master Production Schedule

Lead time contains both shipping time and repair time
No Safety Stock or Initial Inventory
Failures rounded up to 19
Planned 2 months ahead

Assumptions

24

25 of 35

Improve

Expected Failure Costs From MPS

Formulated cost equation
Research yielded highly variable price points
OSIsoft can input true costs into equation

Formulation

25

26 of 35

Improve

Cost Analysis using Gearbox Maintenance Example

~55%

Cost reduction!

26

Proposed Solution (Proactive) Maintenance

Repair gearbox bearings at first sign of issues (vibration monitoring, oil check)
Bearings replacement on standby
Zero turbines have bearing failures
$4,100,000 (20-year EOL)

Traditional (Reactive) Maintenance

Repair / replace gearbox after damage from bearings has already occurred
No safety stock available
20 turbines have bearing failures
$9,100,000 (20-year EOL)

Note: Gearbox replacement cost average of $350,000. Gearbox bearing repairs costs average $10,000 (every 5 years, first sign of problems). 25% of gearboxes will need replacement every 10 years. 76% of replacements due to bearings issues.

To get a better idea of having some safety stock ready in the event of downtime, let’s do a cost analysis of what our proposed solution could look like in one subsystem, like the gearbox. In many cases, maintenance is more “reactive”; repairs don’t really occur until a failure happens and the turbine goes down. Usually, there isn’t any safety stock for smaller components like the bearings because inventory is more focused on things like the blades, which are easier to detect any issues. This often leads to a turbine being down for more than a day so the company can order a new bearing. Our solution, on the other hand, aims to investigate problems ahead of time by closely monitoring the drivetrain vibration and doing routine oil testing for the gearbox. And just in case the bearings need to be replaced, our proactive solution involving the Master Production Schedule will allow AGL to have a new set of bearings ready to go same day, meaning same day repair. In AGL’s case of 50 turbines, we determined that with OUR solution, we can potentially eliminate most gearbox-related failures versus reactive maintenance, which will continue having bearing issues because the problem isn’t being treated since all they’re doing is waiting until a gearbox fails before replacing it, which easily costs hundreds of thousands of dollars for just 1 turbine. By doing proactive maintenance, we’ve found that AGL can reduce gearbox-related costs by nearly 55%. Now if we had more info from AGL for the other subsystems to do a deeper analysis like we did here for the gearbox, we’d be able to continue reducing costs across the board and easily save millions across the 20 year lifespan for their turbines.

Things that changing failure rate affects:

Cost of maintenance crew
Cost of spare parts on hand

$11.3M value includes initial investment costs

Can simply remove estimated costs since haven’t implemented. Can just qualitatively describe each solution and its pros/cons, expected outcomes

27 of 35

Turbine Vibration Distribution Graph & Probability Plot

Control

27

28 of 35

Vibration Control Chart

Control

28

Average: 0.066 m/s²
N = 558 data points
n = 31 point sample size
m = 18 subgroups
Cycle Pattern

Process not in control

Units of vibration

29 of 35

Vibration Status

Control

29

≈ 96% of process in control

≈ 4% not in control

≈ 1-2% are abnormal

Potential causes for outliers:

Braking during high speed winds
Rigid tower vibrations
Loose components

30 of 35

Process Capability

Control

30

LSL	0
USL	0.5
Cₚ	1.62
Target Value	0
Cₚₘ	0.995

Cₚₖ	0.43
Z	1.29
% Conforming	80.22%
DPM	197,760
Defect Rate	27.5/day

Control and maintain 0 vibration
Process is capable of meeting specification
Expecting ≈ 28 abnormal vibration per day

31 of 35

Conclusion

Data-driven Results

Build a maintenance plan that allows for safety stock of failure-prone components (gearbox, bearings, blades, generators) annually to reduce turbine failure rates.

Wind turbines are moderately efficient (> 90% uptime)
Clusters 2, 3 low uptime requires investigation
96% of vibration data in control
Early maintenance via monitoring and routine testing significantly more cost-efficient than replacing when turbine damage is done

Photo Source: Great Big Story (YouTube)

31

32 of 35

Recommendations

Preparing AGL for Success

Provide true maintenance, replacement costs of turbine and components
Consider triannual maintenance schedule
Opt for new vibration-dampening towers (if replacing)
Invest in additional sensors, monitoring changes in condition (i.e. vibration, temperature, etc.)
Dynamically adjust pitch, yaw according to weather, limiting blade damage and reducing failure

Photo Source: JHU Hub

32

33 of 35

References

Blewett, D. (2021, December 20). Wind turbine cost: Worth the million-Dollar price in 2022?

Weather Guard Lightning Tech. Retrieved April 19, 2022, from https://weatherguardwind.com/how-much-does-wind-turbine-cost-worth-it/

Dao, C., Kazemtabrizi, B., & Crabtree, C. (2019). Wind Turbine Reliability Data� Review and impacts on levelised cost of energy. Wind Energy, 22(12),� 1848–1871. https://doi.org/10.1002/we.2404

Dean, D. (2008). Wind turbine mechanical vibrations ... - national wind watch. Retrieved April

20, 2022, from https://docs.wind-watch.org/Dixie-Dean-2008_Soil-vibration.pdf

Gowdar, R. D., & Mallikarjune Gowda, M. C. (2016). Reasons for wind turbine� generator failures: A multi-criteria approach for Sustainable Power Production.� Renewables: Wind, Water, and Solar, 3(1).� https://doi.org/10.1186/s40807-016-0029-1

How much do wind turbines cost? Windustry. (2016). Retrieved April 19, 2022, from

https://www.windustry.org/how_much_do_wind_turbines_cost

Kahrobaee, Salman, and Sohrab Asgarpoor. “Risk-Based Failure Mode and Effect Analysis for

Wind Turbines (RB-FMEA).” 2011 North American Power Symposium, 2011,

https://doi.org/10.1109/naps.2011.6025116.

33

34 of 35

References

Mein, S. (2020, May 11). Top 3 types of wind turbine failure. Firetrace International. Retrieved

April 19, 2022, from https://www.firetrace.com/fire-protection-blog/wind-turbine-failure

Nissi. (2022, March 30). How much energy does a wind turbine produce? " 2022. Mission New

Energy. Retrieved April 19, 2022, from https://www.missionnewenergy.com/how-much-energy-does-a-wind-turbine-produce/

Ozturk, S., Fthenakis, V., & Faulstich, S. (2018). Failure modes, effects and� criticality analysis for wind turbines considering climatic regions and comparing� geared and direct drive wind turbines. Energies, 11(9).� https://doi.org/10.20944/preprints201807.0602.v1

Rajesh Kumar Reddy, P., Maheshwara Rao, & K., Bala Kishore, P. (2015) Wind Turbine Pitch And� Yaw Control. International Journal of Science, Technology & Management, 4(1).� http://www.ijstm.com/images/short_pdf/1426229491_643.pdf

Saavedra, Rodolfo C., and Biswanath Samanta. “Noise and Vibration Issues of Wind Turbines

and Their Impact – A Review.” Wind Engineering, vol. 39, no. 6, 2015, pp. 693–702,

https://www.jstor.org/stable/90007104 . Accessed 20 Apr. 2022.

Wind Turbine. Wind turbine - Energy Education. (n.d.). Retrieved February 2, 2022,� from https://energyeducation.ca/encyclopedia/Wind_turbine

34

35 of 35

Thank You!

35