Deploying Wireless Sensor Networks with Fault Tolerance for Structural

Health Monitoring

The 8th IEEE International Conference on Distributed Computing in Sensor Systems (DCOSS '12), Hangzhou, China, May 16-18, 2012.

Email: zakirulalam@gmail.com

Md Zakirul Alam Bhuiyan

(Joint work with Jiannong Cao and Guojun Wang)

IEEE DCOSS’12 2012-05-17

Outline

Structural Health Monitoring (SHM)SHM and other WSN ApplicationsFault-Tolerant WSN deployment

Sensor Placement Backup Sensor Placement

Simulation and ImplementationConclusions and Limitations

2

IEEE DCOSS’12 2012-05-17

What is a SHM System?

A type of system that provides information about any damage occurring in the structure.

- e. g. buildings, bridges, aircrafts, nuclear plants

3

Damage- is a significant changes in the structure

IEEE DCOSS’12 2012-05-17

A road bridge in Fenghuang, China collapsed at 2007, killing 64 people and injuring 22

A road bridge in Zhuzhou, China, collapsed at 2009, killing 9 people and injuring 16

I-35 Bridge in MN, USA collapsed at 2007, killing 13 people and injuring 145

4

USAF F-16 aircraft at an airshow at Mountain Home Air Force Base, Idaho, on Sept. 14, 2003

American Airlines 587, Airbus A300-600, Nov. 12, 2001.

1Structural Failure (damage)

is about 25%

1http://www.planecrashinfo.com/cause.htm

Why SHM System?

IEEE DCOSS’12 2012-05-17

In designing a SHM system,

we need to know: The structural phenomena to be monitored Sensors Time strategies Damage detection algorithms Data transfer and storage mechanism

5

Designing a SHM System

IEEE DCOSS’12 2012-05-17

Smart sensor nodes Sensors CPU Wireless transceivers

Wireless sensor networks

WSN-Based SHMA New Paradigm of SHM Systems

6

IEEE DCOSS’12 2012-05-17

SHM vs. Other WSN Applications

7

Other applications SHM applicationData types Light, temperature Vibration, (image, acoustics)

Detection model

Sensor nodes collect the energy emitted by the target/event and make corresponding conclusion

(1) The raw data from multiple sensor nodes are collected

(2) Some vibration characteristics are identified

(3) The changes of characteristics damage occurrence?

Detection algorithm

Simple: comparison, majority voting Complicated and centralized:SVD, Eigen-system realization,….

Sampling frequency

X times per second, minute or day X00 times per second

Data volume Received energy level: X bytes Raw measured at each sensor node: X000~X0000 bytes

IEEE DCOSS’12 2012-05-17

The first and fundamental problem in SHM

8

Sensor Placement on Civil Structure

SHM vs. Other WSN Applications

IEEE DCOSS’12 2012-05-17

Placement in general WSN application Random, uniform, grids/trees, rectangular, circular

Are they practically meaningful to the respective application demands? e.g., SHM

Anywhere/Anytime? Constraint:

Energy, communication, connectivity, data delivery, fault-tolerance

Difficulty:Monitoring @ some specific

locations where the damage sensitivity is high

9

SHM vs. Other WSN Applications

IEEE DCOSS’12 2012-05-17

Engineering-driven optimal deployment: Sensor are placed at optimal locations in order to

achieve the best estimates of physical properties of a civil structure. EFI (EFfective Independence) method Genetic algorithm, etc. 2,3

Sensor Placement on Civil Structures

10

2. B. Li, D. Wang, F. Wang, and Y. Q. Ni, “High quality sensor placement for SHM systems: Refocusing on application demands,” INFOCOM, 2010.3. D. Kammer, “Sensor Placement for on-orbit Modal Identification and Correlation of Large Space Structures,” Journal of Guidance, 14: 251-259, 1991.

IEEE DCOSS’12 2012-05-17

EFI Method: It is used to maximize both the spatial independence

and sensor signal strength of the N targeted location by maximizing the determinant of the associated Fisher information matrix (FIM) 3,4

EFI gives EFI values as the location quality Sensors are placed at locations with high EFI

values.

11

Sensor Placement on Civil Structures

3. D. Kammer, “Sensor Placement for on-orbit Modal Identification and Correlation of Large Space Structures,” Journal of Guidance, 14: 251-259, 1991.4. M. Meo and G. Zumpano, “On the Optimal Sensor Placement Techniques for a Bridge Structure,” Engineering Structures, 27 (2005): 1488- 1497, 2005.

IEEE DCOSS’12 2012-05-17

There are M locationse.g., 100 candidate locations

Given N sensors

e.g., 20 sensors

We need to find the optimal locations which are with high EFI values

Location EFI value

1 0.92

2 0.35

3 0.71

4 0.95

5 0.13

… …

99 0.69

100 0.85

Sensor Placement on Civil Structures

12

IEEE DCOSS’12 2012-05-17

Problem Definition (i)

Our problem is to get N sensors that are placed by finding candidate locations out of M using EFI values.

Constraints Connectivity Data delivery Heavy communication load

Objectives: Maximize network lifetime → long-term SHM

13

IEEE DCOSS’12 2012-05-17

Our problem is to get N sensors that are placed by finding candidate locations out of M using EFI values.

Constraints Connectivity Data delivery Heavy communication load

Objectives: Maximize network lifetime → long-term SHM

14

(weak/strong?)

(congestion)

(high rate data collection)

?Can we guarantee that this deployed WSN is able to monitor a civil infrastructure smoothly?

(data packet-losses)Energy constraint!Faults/Failure

Challenges

IEEE DCOSS’12 2012-05-17

32

Challenges

15

the points, where the sensors or communication link failures are highly possible

Fault/ Failure Points

Isolated Pointsu doesn't have a path or communication to

v, or may receive broken messages

Critical Middle points

A longest transmission distance

(duv) and the link between sensor u and v is vulnerable

Separable PointsA connecting point of

other several sensors, whose

removal results in a disconnected network

321

Repairing Points (RPs)

1

IEEE DCOSS’12 2012-05-17

Challenges

16

Backup Sensor Placement

When one or more sensors fail that placed at the optimal locations How to continue obtaining adequate information of

structural behavior or changes in the structure? How to guarantee fault tolerance?

IEEE DCOSS’12 2012-05-17

Idea: Placing a small set of backup sensors by finding remaining/unused near optimal locations using EFI

17

Backup Sensor Placement (BSP) (1/2)

Objectives Mitigating communication failure Mitigating single-point failure:

Backup sensor can take the role when a primary sensor fails Mitigating Packet losses Prolonging network lifetime

IEEE DCOSS’12 2012-05-17

Backup Sensor Placement (BSP) (2/2)

Where to place? @ Remaining /unused locations

There may have a lot of unused

locations available in a structure. M-N

How many backup sensors

are there available?

R (<N), N-R? N+R? How to find the locations for

the backup sensors? at near/the RPs

Location EFI values

1 0.92

2 0.35

3 0.71

4 0.95

5 0.13

… …

99 0.69

100 0.85

18

IEEE DCOSS’12 2012-05-17

Problem Definition (ii)

The problem is to place a set B of R backup sensors into a network with N sensors by optimally finding locations out of M-N remaining/unused locations near the RPs such that: (i) the network is guaranteed to be fault tolerant to the

presence of up to k-1 sensors fault and data packet-loss (ii) T is prolonged

under constraints of long distance communication, k -1 connectivity, data delivery.

19

IEEE DCOSS’12 2012-05-17

Repairing the Network through BSP

Repairing points: BSP1-Separable Points (RPi)

BSP1-Critical Middle Points (RPi)

BSP1- Isolated Points (RPi)

20

IEEE DCOSS’12 2012-05-17

Algorithm: BSP

21

IEEE DCOSS’12 2012-05-17

Algorithm: Search-and-Place

22

IEEE DCOSS’12 2012-05-17

Simulations A total of 100 ((N=80)+ (R=20)) sensor nodes First place the primary sensors and then backup sensors

fr → the percentage of faults/ failures, k-connectivity,

Metric: Communication cost 5,6, T→ Network lifetime

Fault/Failure Injection Communication failure (invalidating

communication module at different points of time Providing limited power

Performance Evaluation (1/4)

23

5. S. Olariu and I. Stojmenovic, “Design Guidelines for Maximizing Lifetime and Avoiding Energy Holes in Sensor Networks with Uniform Distribution and Uniform Reporting,” in IEEE INFOCOM, 2006.6. W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-efficient Communication Protocol for Wireless Micro-sensor Networks,” in IEEEHICSS, 2000.

IEEE DCOSS’12 2012-05-17

The network lifetime (T) vs. the number of sensors (m)

[online]: http://www.cse.polyu.edu.hk/benchmark/

Performance Evaluation (2/4)

24

The communication cost in the network under sensor faults

FTSHM SPEM5

RELAY6

7. Structural Health Monitoring for Guangzhou New TV Tower Using Sensor Networks. [online]: http://www.cse.polyu.edu.hk/benchmark/.

8. K. Xu et al. “Relay Node Deployment Strategies in Heterogeneous Wireless Sensor Networks, ”IEEE Trans. on Mobile Computing,9(2): 145-159, 2010

IEEE DCOSS’12 2012-05-17

Real Implementation

Performance Evaluation (3/4)

Sensor deployment on LSK building

15 floors At least 220 locations (M) 22 sensors (m) [N=16, R=6] Ambient vibration

Sensor fault injection: Sensors attached on the 5rd and 7th

floors are given minimum power at around 2nd hour, we remove

one sensor from the 11th floor,

Observation: Impact of monitoring performance Communication cost Network lifetime

25

IEEE DCOSS’12 2012-05-17

Results

Performance Evaluation (4/4)

26

The total communication cost in the network achieved in different days of

deploymentThe identified affected mode

shape under sensor fault (on day 2)

IEEE DCOSS’12 2012-05-17

A new way to incorporate both WSN and SHM requirements and make use of traditional method used by civil, structural, or mechanical engineering domains for resource-constrained WSNs.

Besides SHM, the idea of backup sensor placement can also be used in generic WSN applications ← repairing points.

Conclusions and Limitations

Limitations: Detailed theoretical analysis and the cost of backup sensor placement

algorithms. Network performance analysis and fault tolerance under physical

damage injection.

27

IEEE DCOSS’12 2012-05-17

Q&A

Contact Info: Alam, Email: zakirulalam@gmail.com

28

IEEE DCOSS’12 2012-05-17

Works @ HK PolyU

Xuefeng Liu, Jiannong Cao, Md Zakirul Alam Bhuiyan, Steven Lai, Hejun Wu, and Guojun Wang, “Fault Tolerant WSN-Based Structural Health Monitoring," DSN 2011

X. Liu, J. Cao, S. Lai, C. Yang, H. Wu, and Y. Xu, “Energy Efficient Clustering for WSN-based Structural Health Monitoring”, in IEEE INFOCOM, 2011.

B. Li, D. Wang, F. Wang, and Y. Q. Ni, “High quality Sensor Placement for SHM Systems: Refocusing on Application Demands,” in IEEE INFOCOM, 2010.

29 May 17, 2012

IEEE DCOSS’12 2012-05-17

Length: 122m U Michigan & Stanford (2005):

Specially designed wireless sensor node prototypes 14 sensor nodes, sampling at 200 Hz, constitute a centralized single-hop network

2 Lynch, J., et al. Validation of a large-scale wireless structural monitoring system on the Geumdang Bridge. 2005: Citesee3 Lynch, J., et al., Performance monitoring of the Geumdang Bridge using a dense network of high-resolution wireless sensors. Smart Materials and Structures, 2006. 15: p. 1561.

Related WorksGeumdang Bridge (Icheon, Korea)

30

IEEE DCOSS’12 2012-05-1731

Wired vs. WSN based SHM Systems

SHM requirements Wired System Wireless System

Low cost

Equipment Expensive Low-cost

Cabling Long cables No cables

Deployment time Months ~ years Hours ~ days

High spatial density X0~X00 X00~X000

Sampling

Fast on command/event

triggered

Delay <μs Seconds ~ minutes(due to the wireless link)

High frequency and synchronized

Frequency >10KHz Sync error <1μs

Frequency < 10KHzLarge sync error

Fast and reliable data delivery

100% data delivery, instant delivery

Data can get lost, single hop bandwidth < 100kbps

Reliable and accurate damage detection

Benefit from centralized algorithms, but constraint by low density & inflexibity

Constraint by limited computation power, but benefits from high density and flexible

IEEE DCOSS’12 2012-05-17

Length: 1280m UC Berkeley (2007):

64 MicaZ motes, with addition sensor board and antenna, sampling at 1kHz rate Sampled data is collected over a 46-hop network

4 Kim, S., et al. Health monitoring of civil infrastructures using wireless sensor networks. 2007 IPSN’07

Golden Gate Bridge (San Francisco, USA)

32

IEEE DCOSS’12 2012-05-17

Guangzhou New TV Tower, China (GNTVT)

A height of 610m, with a 156m antenna

The main tower reached to a full height of 454m.

Wired sensor networks are currently running on the tower

A WSN is also varied on the tower5

33

5. B. Li, D. Wang, F. Wang, and Y. Q. Ni, “High quality sensor placement for SHM systems: Refocusing on application demands,” INFOCOM, 2010.

IEEE DCOSS’12 2012-05-17

Implementation

Performance Evaluation (2/3)

Test Structure: 12 floors, vibrating under hammer strike

Faulty nodes: Sensors attached on the 3rd

and 7th floors are released Structural damage:

Releasing a support ring on the 9th floor

34