STRUCTURAL HEALTH MONITORING FOR CIVIL …web.stanford.edu/group/sacl/workshop/documents/Keynote...

STRUCTURAL HEALTH MONITORING FOR CIVIL INFRASTRUCTURE –INFRASTRUCTURE 

FROM INSTRUMENTATION TO DECISION SUPPORT

Anne S. KiremidjianDept. of Civil and Environmental Engineeringp g g

Stanford University

IWSHM 2011S t b 13 15 2011September 13‐15, 2011

This research is supported by the NSF CMMI Research Grant No. 0800932, the John A. Blume Fellowship, and the Samsung Scholarship.

Diversity of Civil InfrastructureDiversity of Civil Infrastructure

Why Monitor Civil Infrastructure?Gradual Long‐Term Deterioration

(from Dr. Hae Young Noh)

3/17

Why Monitor Civil Infrastructure? ‐Extreme Event DamageExtreme Event Damage

OutlineOutline

• Wireless/Wired Monitoring System DesignWireless/Wired Monitoring System Design

• Algorithmic Development

• Example Applications

• Decision Support System

• Conclusion

5

Structural Monitoring System Sensors &Structural Monitoring SystemControl Center

Sensors & 

Network

Data management and archiving 

Data StorageManager onManager on 

site

Base Station

Decision Support & Emergency Response 

S t

Data Analysis and Post‐processing System

System

Synchronization

Structure ‐level l i

Signal Processing

Spectral Analysis

Feature Extraction &analysis

Decision making

Feature Extraction & Damage Classification

The Smart Structural Monitoring System

Sensors Decision Support

• structure specific

System

specific• Damage specific

• Data

•model updating• decision making• warning

id Data collection/storage

• guidance• active control

DataAnalysis

&Sensor N t k&

Archiving

• physical modeling 

Network

•Wired•Wireless

• signal processing• statistical analysis

7/17

•Wireless• Combined• Communication

Damage‐Specific Sensorsg p

Currently Available Still neededWired sensors Wireless/Wired CrackWired sensors Wireless/Wired 

sensorsCrack

Fiber optic sensors Accelerometers Corrosion

High definition digital cameras

Strain gages Displacement

Laser GPS Materials specificLaser interferometers

GPS Materials specific

Laser scanner Tilt metersDeep penetration radar

Temperature 

HumidityyCorrosionAnemometers

The Smart Structural Monitoring System

Sensors Decision Support

• structure specific

System

specific• Damage specific

• Data

•model updating• decision making• warning

id Data collection/storage

• guidance• active control

DataAnalysis

&Sensor N t k&

Archiving

• physical modeling 

Network

•Wired•Wireless

• signal processing• statistical analysis

9/17

•Wireless• Combined• Communication

Sensor NetworksSensor Networks

• Wireless vs. wiredWireless vs. wired • Advantages of wireless systems

– ScalableScalable – Ease of installation– PortabilityPortability– Lower cost

• ChallengesChallenges– Potential signal loss– Communication barriersCommunication barriers

The Smart Structural Monitoring System

Multiple Sensor

Decision Support

• structure specific

System

specific• Damage specific

• Data

•model updating• decision making• warning

id Data collection/storage

• guidance• active control

Analysis & Data Sensor

N t kArchiving

• physical modeling 

Network

•Wired•Wireless

• signal processing• statistical analysis

11/17

•Wireless• Combined• Communication

Types of Algorithmyp g

• Device control• Damage diagnosis– statistical pattern recognition methods

– Wake up – Database structure

pattern recognition methods– AR/ARMA/ARX

• Hypothesis testing

– Baseline collection• By time

• Gaussian Mixture Modeling‐ GMM

– Wavelet Based – Haar and Morelet wavelets

• By season• By temperature• By humidity

• Comparison of wavelet energies at high scales

• Gaussian Mixture Modeling– Synchronization

• Hardware – internal clock

– Rotation ‐ to drift ‐to damage

• Software – pre ‐& post synchronization

Damage Detection Using Statistical g gSignal Processing

• Main approach

– Use single sensor pre‐ and post‐ damage measurements

– Combine information from multiple sensors

– Computationally efficient ‐ local micro‐processingComputationally efficient  local micro processing

– Independent of the sensor – can be used with acceleration, strain, etc.,

– Scalable with increased sensor density

– Reduces amount of transmitted data – power saving– Reduces amount of transmitted data – power saving

Steps in Damage Diagnosis

Statistics BasedARX

Collect data ARGMM

Extractfeatures

WaveletChange point detection

features

Classify damage

14/17

ExamplesExamples

• Using stationary vibration signals –Using stationary vibration signals – AR with information criteria testing

AR with Gaussian Mixture Model– AR with Gaussian Mixture Model

• Using non‐stationary vibration signals – e.g. th k tiearthquake motions

NEES – UNR Project‐¼ Scale Bridge TestAR & G i Mi t Al ithAR & Gaussian Mixture Algorithm 

(Nair & Kiremidjian, 2006)

Decision Support SystemDecision Support System

Test Schedule – 4‐span Bridge Test at UNR

Baseline Signalsg

Minor cracks at base of 

Major cracks at base of l & lli

column

Major cracks at base of l & lli

Concrete spalling and exposure or rebar

col. & spallingcol. & spalling

Rebar buckling/ breaking; concrete 

i t fpouring out of core

Final Test – White Noise

Test Damage MeasureReno - Setup Day DC1 and DC2 baselineMild Shaking Day 1, White Noise 21.05Mild Shaking Day 2, White Noise 21 36.79Mild Shaking Day 3, White Noise 41 56.97Final Test Day White Noise Run 51 59 80Final Test Day, White Noise Run 51 59.80

C l iColumn Device

DM

Test Damage Measure

Reno - Setup Day DC1 and DC2

baseline

Mild Shaking Day 1 21 05Mild Shaking Day 1, White Noise

21.05

Mild Shaking Day 2, White Noise 21

36.79

Mild Shaking Day 3, White Noise 41

56.97

Final Test Day, White Noise Run 51

59.80

Example of Wavelet Damage Diagnosis(Noh et al., 2011)                

• Feature fromFeature from wavelet energies of signal

• Used with non‐stationary signals –e.g. earthquake response motions

• Develop fragilities for rapid damage i di iindication

NEES 4 Story Steel FrameWavelet Based  Algorithm g

• Scaled structural system tests Stanford/SUNYtests – Stanford/SUNY Buffalo

• Development of fragilityDevelopment of fragility functions in terms of structural response 

t bt i blparameters obtainable from real time measurements – wavelet based fragilities 

(Noh, Lignos, Nair and Kiremidjian 

)2011)

The Smart Structural Monitoring System

• structure specific

SystemMultiple Sensor

Decision Support specific

• Damage specific

• Data

•model updating• decision making• warning

id Data collection/storage

• guidance• active control

Analysis & Data Sensor

N t kArchiving

• physical modeling 

Network

•Wired•Wireless

23/17

• signal processing• statistical analysis

•Wireless• Combined• Communication

Decision Support SystemDecision Support System

• Visual representation of the• Visual representation of the structure

• Visual representation locations pof the wireless system

• Interface to wireless network

• System command and control center  

• Display results of monitoringDisplay results of monitoring analyses

• Issue alerts

Components of a Decision Support System

• Monitor the sensor• Monitor the sensor 

communications system• Serve as the 

communications i t b tenvironment between 

manager and system– Initial set‐up– Modifications of system– Modifications of system 

parameters– Modifications of 

monitoring settings• Serve as the information 

delivery environment– Periodic queries

Following a major event– Following a major event

Decision Support System• Provide support for decision making for follow‐on actions

• Enable web services for – wide distribution of alerts and other information

– remote access by operators and other users.

C l iConclusion• Wireless monitoring systems – inevitable part of 

h fthe future• Few applications in Europe, Asia and the US • Must combine structural with other monitoring systems, e.g.

B ildi i t l/ /li hti / it– Building environmental/energy/lighting/security monitoring

– Bridge, highway, tunnel, pipeline, transmission line, g , g y, , p p , ,etc. management systems 

• Key to success is providing information and d ddecision support, not just data

27

Need for Full‐scale and Field Testing

• ObjectivesObjectives– Systematic damage at different levelsdifferent levels

– Different damage patternspatterns

– Different damage sequencesq

NIED – E‐Defense Test 5‐Story Full Scale RC Building, y g,August, 2011

AcknowledgementAcknowledgement• National Science Founcdation

• National Institute of Standards and Technologygy

• Students

Dr. E. Straser ‘98A Kotapalli ‘99

Dr. K. Nair ‘07Allen CheungA.Kotapalli 99

N. Mastroleon ‘00C Ch i i ‘03

Allen CheungDr. H‐ Y Noh ‘11

C. Charistis ‘03

THANK YOU!THANK YOU!