sensitivity evaluation of subspace-based damage detection technique saeid allahdadian dr. carlos...

18
Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia, Vancouver, Canada The 5 th Tongji-UBC Symposium on Earthquake Engineering Facing Earthquake Challenges Together” May 4-8 2015, Tongji University, Shanghai, China

Upload: ethan-jacobs

Post on 15-Jan-2016

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Sensitivity Evaluation of Subspace-based Damage Detection Technique

Saeid AllahdadianDr. Carlos Ventura

PhD Student, The University of British Columbia, Vancouver, Canada

The 5th Tongji-UBC Symposium on Earthquake Engineering “Facing Earthquake Challenges Together” May 4-8 2015, Tongji University, Shanghai, China

Page 2: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

2

• Subspace based damage detection

• Damage detection approach

• Investigation of sensitivity

• Case study

• Damage and noise simulation

• Results

• Conclusion

Outline

Page 3: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

• Benefits:• Rapid assessment of whether or not a system has

changed its dynamic behavior.• Upload a new measurement file and get a Yes or

No whether a significant change has happened. Applicable for real time monitoring of structures.

• Limited user interaction once the system is set for testing.

Subspace based damage detection

Page 4: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Yes

- 4

Structure healthy

Measurement in healthy state

Measurement in current state

Statistical comparison concerning

the vibration characteristics

Significant change?

No

Damage(or something else?)

Damage detection approach

Page 5: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

¿1 2

2 3 1

1 2

...

...

q

q

p p p q

RR R

R R R

R R R

H

vec( ),Tn nζ S H 12 ,T

n n ζ ζ Tn n E ζ ζ

Formulation

Page 6: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Sensitivity analysis of the damage detection approach to:

◦Damage location◦Damage ratio◦Noise ratio

6

Page 7: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

S101 bridge structure

Page 8: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Pier settlement as controlled damage

Page 9: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

• Natural frequencies of the bridge structure in undamaged condition obtained from the measured data and finite element model

Data analysis

Page 10: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Damage ratio

Minor damage 20%

Intermediate damage 40%

Severe damage 80%

• The damage is simulated only in one finite element corresponding to the center of the main girder by reducing a ratio of its section dimension around the strong axis

• The imposed noise on the data is created using a random generation algorithm with evenly distributed probability distribution

Nr: noise ratiomi: maximum value measured for each channel

random( ,even)i r iN mR

Damage and noise simulation

Page 11: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Noise superposition

i i i ND D R

Page 12: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Noise characteristics• The effect of the noise in the data is more visible on low

amplitude parts of the signal

High effect

Low effect

zoom

Freq. domain

Page 13: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

• Measuring-points corresponding to sensor locations in ARTeMIS® software and spectral densities

Operational modal analysis

Page 14: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

14

Girder center Deck

Bearings Cap beams

Damage location and ratio effect

Page 15: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

Noise and damage ratio effect

Page 16: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

• The minimum of the value occurs when there is equivalent noise ratio in the measured and reference data

• To have a comparison regardless of this effect the results with equal noise ratio in the reference state and measured data are compared

• The optimum range of noise ratio in this set of data is from 2% to 30% equal in reference and measured data with a peak about 5%

Discussion on results

Page 17: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

• When there is noise in the reference data, there should be more samples and measurements performed in order to acquire a reliable safety threshold.

• Not only the higher noise in the measured data than the reference state can be interpreted as damage but also the lower noise in the measured data considering the reference state noise ratio can also affect the outcome of the damage detection technique

• Results are stable for the region between 2% and 30% noise on the outputs

• In most of the cases, even with very high noise ratios the damage can be identified using the SSDD technique. The reason: evaluation of the changes in the eigen-structure of the model.

Conclusion

Page 18: Sensitivity Evaluation of Subspace-based Damage Detection Technique Saeid Allahdadian Dr. Carlos Ventura PhD Student, The University of British Columbia,

18

Certificate of Structural

Engineering Program

Thanks for your attention!