Structural Health Monitoring

1
Structural Health Monitoring

• Sukun Kim, David Culler
• James Demmel, Gregory Fenves, Steve Glaser,
  Shamim Pakzad
• UC Berkeley

CENTS Retreat
2
Structure Monitoring
Data Acquisition
Data Collection
Data Processing Feedback
3
Accelerometer Board

• Two accelerometers for two axis
• Thermometer
• 16bit ADC

4

• Hardware
• Board
• Accelerometers Comparison
• Noise Floor (Vault Test)
• Tilting Calibration
• Temperature Calibration
• Mote
• Antenna
• Options and Our Choice
• Power
• Power Consumption Profile
• Power Source Options and Choice
• Package
• Software Architecture
• Overall Structure
• High Frequency Sampling
• Jitter Test
• Jitter Analysis

Mint Route (Alec Woo, et al)
The Flooding Time Synchronization
Protocol (Miklos Maroti, et al)
Drip (Gilman Tolle)
5
Why temperature calibration?

• Accelerometer is sensitive to temperature change
• In bridge environment, there exists significant
  variation in temperature (up to 45F, 40 mG)
• We are looking at very subtle signal (down to 0.5
  mG)
• Signal to noise ratio becomes small under
  temperature change (down to 1)

6
Temperature Calibration Test
F
C
27.3
81.1
19.5
67.1
11.7
53.0
Temperature
3.9
39.0
mG
27.5
0
Thanks to Crossbow
-27.5
Acceleration
7
Temperature Calibration

• Suggestion to estimate instantaneous regression
  parameters by windowing the signals
• a window of length 199 samples is considered and
  a linear regression model of the form
• accel.count a btemp.count e
• is fit to that windowed segment of the data
• The following graphs are the estimated
  parameters. The parameters include a and b (as
  defined above) and the standard deviation of e,
  the error term in the regression model

8
mG
2.75
0.92
Temperature change will not be as dramatic as in
the test (with insulation) However, concern for
temperature hysteresis drives us to second run
9
Bridge Dimension
2
500 ft
1125 ft
2100 ft
4
8
246 ft
1
Blue number presents rough ratio
Need to cover large bridge, so directional
antenna is needed
10
Types of Antenna
Dish
Yagi
Horn
Patch
Bigger Size Longer Range
Smaller Size Shorter Range
11
How much gain do we need?

• Golden gate bridge is 6,450 ft including main
  span and side spans
• Assuming 20 hops along the span, each hop is 340
  ft
• 433MHz Mica2 reached 100ft (84 success)
• 916MHz Mica2 will reach 50ft
• 916MHz with maximum radio power will reach 225ft
  (4.5X increase 10dBm to -3dBm)
• 1.5X increase in range is needed
• 3.3 dBi antenna is needed

12
Antenna Candidates
4.8
Bi-directional patches (2.4GHz) From Superpass
4.4
11.5
4.5
5.5 dBi
9 dBi
13
Performance of antenna
5.5dBi antenna with Telos Maximum output power
(0dBm) 0.5ft above the ground in front of Soda
Hall
3ft above the ground, success rate was close to
100 9dBi antenna showed very similar behavior
(0 above 0.5ft, gt99 above 3ft) The distance to
the ground is more important than the gain of the
antenna
14
Power Consumption Data

• Board only 26.7
• Idle 39.8
• One led on 42.6
• Erasing flash 77.5 (with one led on)
• Sampling 42.6 (with one led on)
• Transferring data 46.0 (with one led on)

unit is mA at 9V
15
Node Deployment Plan
10 nodes
30 nodes
10 nodes
Nodes on both sides of span
16
Power Consumption

• The most burdened node transfers data one third
  of time
• Tadiran 5930 3.6V, 19Ah, 17, D size
• Usual 9V alkaline battery has 625mAh (12X)
• Usual 1.5V D battery has 18Ah (2.5X)
• 3 of Tadiran 5930 costs 51, and lasts 23 days
• 3 weeks is good enough

17
Power Consumption (cont)

• With optimal sleeping, 30 days
• Board itself consumes significant amount of
  energy

Power source
Power source
Switch
Switch
Sensor
Mote
Mote
ADC
Sensor
ADC
18
Reliable Data Collection
19
Footbridge Location of Nodes
Half
Quarter
1/8
Quarter
260ft
2
4
6
7
16ft
Marina
Berkeley
1
3
5
8
Base Station
20
Data is from single-hop version
21
We will collect data for multi-hop version
22
Data Analysis

• Characteristic vibration modes were observed

23
Conclusion

• Practical problems to solve
• Interesting challenges and research topics (more
  efficient reliable transfer)
• Future Work
• Handling temperature hysteresis
• Antenna
• More efficient data collection (pipelining?)

24
Questions
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
Examples
Gain, angle, size, weight, (price), range
compared to wire whip antenna
31
Comment on Availability

• Availability of horn antenna is limited
• Antenna for 433MHz can be found, but with some
  more work (probably from abroad)

32
Suggested Choice Yagi, Patch

• Dish is overkill
• Yagi without shell
• Enough gain
• Long (15 for 916MHz) and narrow
• Robust to strong wind
• Horn has limited availability
• Patch
• Enough gain
• With large surface (8.5X8.5 for 916MHz) and thin
• Plate part can be a problem under strong wind

33
Bi-directional Patches
34
References (Antenna)

• Antenna vender
• http//www.hyperlinktech.com
• http//www.antennafactory.com
• http//www.antennafactor.com
• http//www.antennasystems.com
• Horn antenna
• http//www.phazar.com
• Antenna in general
• https//ewhdbks.mugu.navy.mil/ANTENNAS.HTM
• dBi, dBd
• http//www.tmeg.com/tutorials/antennas/antennas.ht
  m
• Golden Gate Bridge Facts
• http//www.thoma.com/thoma/ggbfactsX.html

35
dBi, dBd

• dBi (decibels-isotropic) a unit of measuring
  how much better the antenna is compared to an
  isotropic radiator
• dBd (decibels-dipole) compared to a dipole
  antenna
• Dipole antenna typically has a 2.4dBi gain
• Wire whip antenna (used in mica2) would have
  1.5dBi gain

36
(No Transcript)
37
(No Transcript)
38
(No Transcript)
39
(No Transcript)
40
One Base Station w/o Pipelining

• If only one base station is located at kth place
  from the right, total transfer time is
• 2 (k-1)k/2 (165-k)(165-k1)/2 4(16-k)
  6 (single one-hop transfer time)
• Minimum 248 when k12 or 13
• When flash is full (6min data at 200Hz, 5
  channels 10min xfer), and with 800bytes/s
  bandwidth ? 2.1 days of data collection

Every data in this file is based on Mica2
Minimum positions
41
One Base Station w/ Pipelining

• Assuming communication can occur 3 hops away,
  lower bound is 3 50 times of single one-hop
  transfer (10min) ? 25 hours
• Bottleneck is speed of data arriving at the base
  station with space among them
• With N base stations, time will become 25/N hours

Space preventing interference
Base Station
42
More thought on pipelining

• As the network gets bigger, we can get more
  benefit (N versus N2)
• However, in small network, path is not long
  enough for multiple transfers to happen
• With 4 sinks, assuming perfect pipelining (can be
  unrealistic), it takes 4.2 hours (20min 50 / 4)

Base Station
43
(No Transcript)
44
Preliminary BOM
45
(No Transcript)
46

• 1. Radio with Mica2
• 19.2Kbps 2400bytes/s 66.7pkts/s
• Media access control reduces this to 52pkts/s.
• Single mote can achieve 42pkts/s. (probably due
  to processing overhead?)
• 2. UART
• 57.6Kbps 7200bytes/s 200pkts/s
• 3. TOSBase
• In TOSBase with Mica2,
• 34pkts/s max in theory.
• 23pkts/s is a reliable upper bound.

47
Overall Process
1. Trigger Sampling
2. Transfer Metadata
3. Transfer Data
PC has most of intelligence. Motes are almost
stateless.