Damage Detection of Operating Wind Turbine Blades and ... · PDF fileDamage Detection of Operating Wind Turbine Blades and Towers Using Signature Distances ... NREL 5MW reference o

Damage Detection of Operating Wind Turbine Bladesand Towers Using Signature Distances

Dylan [email protected]

Wind Energy CenterDepartment of Mechanical and Industrial Engineering

University of Massachusetts Amherst

October 16, 2013

D. Chase (Wind Energy Center) Damage Detection of Operating Wind Turbines October 16, 2013 1 / 17

Damage Detection

Structural Health Monitoring

The process of implementing a damage detection and characterizationstrategy for engineering structures.

Levels of detection:

1. Occurance of Damage2. Amount of Damage3. Location of damage

Other questions:

1. How many sensors are needed?2. Where should the sensor be located?


Damage Detection





Other questions:



Damage Detection





Other questions:



Research Summary

Original idea: apply wavelet signature method to turbines

I Pattern matching detects and isolates damageI Relies on healthy baseline signalsI Detects shape changes in the residual

Problem: Baselines hard to establish due to varying:I Wind conditionsI Initial conditionsI Loading

Solution: Methods were developed which are uniquely affected by thedamage

I Tower: shutdown maneuver is analyzed in the time domainI Blade: normal operation is analyzed in the frequency domain


Research Summary

Original idea: apply wavelet signature method to turbinesI Pattern matching detects and isolates damageI Relies on healthy baseline signalsI Detects shape changes in the residual





Research Summary


Problem: Baselines hard to establish due to varying:

I Wind conditionsI Initial conditionsI Loading




Research Summary






Research Summary






Codes

FAST: Nonlinear dynamic wind turbine model includes:I Dynamics of the: tower, blades, generator, gearboxI Control systems: pitch, yaw, brakesI Aerodynamics and hydrodynamics

MODES: Generates blade and tower mode shapes

TURBSIM: Generates full field turbulent wind files

WaveLab: Open source MATLAB toolbox


Codes






Codes






Codes






Codes






Model

NREL 5MW reference offshore windturbine

63 m blades

80 m tower

Pitch controlled

Full field turbulent wind

Distributed parameter systems

Damage = a reduction in stiffness

Outputs: Accelerometer locations

Stiffness Values: Stiffness is specified

Parameters: Damage is modelled


Methodology

5 wind profiles are considered

For each wind profile healthy and faulty conditions are simulated

The healthy turbine in Wind 1 is the baseline

Signature distances are calculated by comparing the baseline to theother signals

Result is healthy and faulty signature distances

The faulty cases consistently result in a larger signature distanceindicating that damage is detectable


What are wavelets?

0.2 0.4 0.6 0.8 1−0.2

−0.1

0

0.1

0.2 Haar Wavelet

0.2 0.4 0.6 0.8 1−0.3

−0.2

−0.1

0

0.1

0.2 D4 Wavelet

0.2 0.4 0.6 0.8 1−0.2

−0.1

0

0.1

0.2 C3 Coiflet

0.2 0.4 0.6 0.8 1−0.2

−0.1

0

0.1

0.2 S8 Symmlet

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

1

2

3

4

5

6

7

8

9Some S8 Symmlets at Various Scales and Locations

(3, 2)

(3, 5)

(4, 8)

(5,13)

(6,21)

(6,32)

(6,43)

(7,95)

Little waves, Area under curve is 0

Are convolved with a signal (like Fourier Analysis)

Are stretched (scale) and translated (time)

Benefits of Wavelets:I Can detect local events in a signalI Provides multi-scale resolution (multi-scale smoothing)I Turns a 2 dimensional signal into a 3 dimensional surface (enhanced

resolution)


Wavelet Transform Example

500 501 502 503 504 505−0.4

−0.2

0

0.2

0.4

0.6

Accele

ration (

m/s

2)

Time (sec)

healthy Fault 1

500 501 502 503 504 5050

50

100−0.01

0

0.01

Time (sec)Scale

Illustration of the enhanced delineation of two similar signals (top) by thedifferential wavelet transform (bottom) which accentuates the minute differencesbetween the two signals


Response Magnified

510 512 514 516 518 520

−0.2

0

0.2

Accele

ration (

m/s

2)

Time (sec)

510 512 514 516 518 520−2

0

2

Accele

ration (

m/s

2)

Time (sec)

healthy, wind 1 fault 1, wind 1

healthy, wind 2 fault 1, wind 2

Acceleration of the tower at location 7 (top) and the blade at location 7(bottom) after the brake maneuver is performed


0 1 2 3 4 5 6 7 8 9 10 11 12−40

−20

0

20

40

PS

D (

dB

/Hz)

0 1 2 3 4 5 6 7 8 9 10 11 12−40

−20

0

20

40

PS

D (

dB

/Hz)

0 1 2 3 4 5 6 7 8 9 10 11 12−40

−20

0

20

40

Frequency (Hz)

PS

D (

dB

/Hz)

Healthy Mode 3 Healthy Mode 4

Healthy Blade, Wind 1

Healthy Blade, Wind 2

Damaged Blade, Wind 1

Distinct difference found in the third and fourth mode


Definition of a Change Signature in words

510 515 52024

4872

−0.01

0

0.01

Time (sec)Scale

Gauss WT of tower acceleration at location 7 after the shutdown maneuver isperformed

WTs of the baseline signal is compared to a different signal

Locations where the ratio of the two WTs is above a dominancefactor are flagged

This results in a change signature


Change Signature ExampleS

cale

510 515 520

24

48

72

Scale

Time (sec)

510 515 520

24

48

72

Tower Change Signatures

Top: Healthy to Baseline

Bottom: Faulty to Baseline

Sca

le

0 1 2 3 4 5 6 7 8 9 10 11 12

24

48

72

Frequency (Hz)

Sca

le

0 1 2 3 4 5 6 7 8 9 10 11 12

24

48

72

Healthy Mode 3 Healthy Mode 4

Blade Change Signatures




Windowing of Change Signatures

Tower Change Signatures



Blade Change Signatures




Tower Signature Distances

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Output 1

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Output 2

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Output 3

1 2 3 4 50

20

40

60

80Im

ag

e D

ista

nce

Output 4

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Output 5

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Output 6

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Wind Profile

Output 7

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Wind Profile

Output 8

1 2 3 4 50

20

40

60

80

Ima

ge

Dis

tan

ce

Wind Profile

Output 9

Healthy Fault 1 Fault 2 Fault 3

Fault 4 Fault 5 Fault 6

Image distances of the acceleration time histories at 9 output locations of thetower obtained for the healthy tower and a 5% damaged tower with five differentwind profiles having a mean speed of 12 m/s


Blade Signature Distances

1 2 3 4 50

5

10

Ima

ge

Dis

t. Output 1

1 2 3 4 50

5

10

Ima

ge

Dis

t. Output 2

1 2 3 4 50

5

10

Ima

ge

Dis

t. Output 3

1 2 3 4 50

5

10

Ima

ge

Dis

t. Output 4

1 2 3 4 50

5

10

Ima

ge

Dis

t. Output 5

1 2 3 4 50

5

10

Ima

ge

Dis

t. Output 6

1 2 3 4 50

5

10

Imag

e D

ist.

Wind Profile

Output 7

1 2 3 4 50

5

10Im

ag

e D

ist.

Wind Profile

Output 8

1 2 3 4 50

5

10

Imag

e D

ist.

Wind Profile

Output 9

Healthy Fault 1 Fault 2 Fault 3

Image distances of the change signatures in the proximity of the third mode forthe healthy and 30% damaged blade by each of the nine blade output locations


Conclusion

Damage is detectable using these methods

The tower damage method is not sensitive to the output location

Blade damage is sensitive to the output location

Future workI Develop a method for determining the location of damageI Validate the method using test dataI Choose an optimal number and location of sensors


THANK YOU

Thank you for attention!

Contact: [email protected]


Documents

Damage Detection of Operating Wind Turbine Blades and ... · PDF fileDamage Detection of Operating Wind Turbine Blades and Towers Using Signature Distances ... NREL 5MW reference o