Wireless Sensor Network Architecture for Structural Health Monitoring

1
Wireless Sensor Network Architecture for
Structural Health Monitoring

• Michael Sirivianos
• April 17, 2003

2
Paper

• Two-tiered wireless sensor network architecture
  for structural health monitoring
• Appeared in SPIEs 10th Annual International
  Symposium on Smart Structures and Materials, San
  Diego, March 2003
• V. A. Kottapalli, A S. Kiremidjian, J.P Lynch, Ed
  Carryer, T. W. Kenny, K. H Law, Ying Lei
• John A. Blume Earthquake Engineering Center,
  Stanford University

3
Structural Health Monitoring

• Recent advances and technologies assist
  structural engineers in their attempts to ensure
  the safety and reliability of structures over
  their life spans through monitoring systems
• Structural Health monitoring an application of
  advanced monitoring systems
• The technology employs smart sensors in a
  configuration that provides materials monitoring
  needed to detect and remotely address any
  compromise in material structural integrity.

4
Traditional Structural Monitoring

• Employs conventional cables to allow sensors
  deployed in a few critical locations of the
  structure to communicate their measurements to a
  central Data Acquisition System module.
• Older systems had analog sensors and A/D
  converters at the DAQ to convert the analog
  vibration signal into a digital format
• Newer systems incorporate digital sensors to
  avoid A/D conversion at the central point to
  enable more reliable communication and relieve
  the central DAQ from the conversion load

5
Traditional Structural Monitoring (2)

• Cons
• Cabled based sensing systems for structures have
  high installation and maintenance costs
• Wires vulnerable to ambient signal noise
  corruption.
• Wired links are prone to breakage and
  environmental wear.
• centralized approach with all system sensors
  sending measurement data to one data server. 
  Such an approach adds latency to the system
  during real-time data processing and represents a
  single point of failure. 

6
Wireless Monitoring Systems

• Research effort has been initiated towards the
  development of a wireless modular monitoring
  system.
• Lower installation and maintenance cost.
• More reliability in the communication of sensor
  measurements

7
Wireless Monitoring Systems (2)

• Areas of innovations
• Use of a wireless communication system for
  inter-sensor communication. Low cost wireless
  technologies have significantly contributed
  towards this direction.
• Bluetooth. Operate in the unlicensed, 2.4 GHz
  radio spectrum. These radios use a spread
  spectrum, frequency hopping, full-duplex signal
  at up to 1600 hops/sec. The signal hops among 79
  frequencies at 1 MHz intervals to give a high
  degree of interference immunity.

    
8
Wireless Monitoring Systems (3)

• The 802.11 standard specifies a single Medium
  Access Control (MAC) sub layer and 2 radio and
  one infrared Physical Layer ( PHY )
  Specifications.
• The standard provides multiple data rates and
  power management (stations can switch off their
  transceivers to conserve power).
• The MAC protocol is Carrier Sense Multiple Access
  with Collision Avoidance (CSMA/CA).
• 2 Physical layer specifications for radio,
  operating in the 2 400 - 2 483.5 MHz band and one
  for infrared. 1 or 2 Mbps data rates
• (1) Frequency Hopping Spread Spectrum Radio
• PHY.
• (2) Direct Sequence Spread Spectrum Radio PHY.
• (3) Infrared PHY.

9
Wireless Monitoring Systems (4)

• Utilization of micro-electro mechanical (MEMs)
  sensing elements and
• Use of advanced microprocessor architectures for
  computationally expensive real-time damage
  detection and assessment methods.
• Wireless sensing units have the flexibility
  to communicate peer to peer (decentralized P2P)
  or in a traditional centralized fashion.

     
10
Problem Statement

• The nature of monitoring systems dictates a
  specialized network architecture that would
  specify
• optimum network topology,
• the best suitable wireless
• technology and
• the appropriate protocol stack.

11
Roadmap

• Study of the features and requirements of the
  structural monitoring application.
• The two-tiered architecture as the answer to
  these issues.
• Network and System Components Architecture
• Description of a communication protocol
• Power saving techniques incorporated in the
  sensor unit architecture
•  

12
RoadMap (Cont.)

• System Analysis
• Conclusions on the trade-offs
• Estimation of the maintenance cycle based on the
  power consumption of the sensor unit
• A basic laboratory implementation of the
  suggested work.

13
Characteristics and Requirements

• Two modes of operation
• Extreme Event Monitoring
• Long Term Periodic Monitoring
• Size of monitored infrastructure
• Generally large, miles long in case of bridges
• Measured parameters
• Acceleration
• Linear and angular displacement
• Environmental variables as temperature and
  humidity

14
Characteristics and Requirements (Cont.)

• Data generation rate
• Depends on the sampling rate
• Vibration Data Synchronization
• Maintenance
• Long maintenance cycles-years
• Long lasting batteries
• Environmental variables as temperature and
  humidity

15
Characteristics and Requirements-Summary

• High throughput for real time performance
• Synchronization of Distributed Sensors Data
• Large transmission range
• Minimum power consumption
• Large transmission range and data rates
  requirement in conflict with power consumption
• Solution two tiered wireless sensor network

16
Two-tiered Wireless Sensor Network

• First subsystem. Sensor Units.
• Low data rate, low transmission range, low
  consumption
• Second subsystem. Large Coordinator Units.
• High data rate, large transmission range, not
  energy constrained

17
Two-tiered Architecture

• Network Architecture and Topology.
• Clustering of distributed sensor units (SUs)
  similar to the structure of a cellular network
• A Local Site Master LSM assigned in each cluster
  to coordinate SUs and collect their data.
• SU clusters form lower tier. LSM network and
  Central Site Master form upper tier

18
Two-tiered Architecture (Cont.)
19
Lower Tier

• R/F within cluster is over the 915 MHz ISM band.
       
• Spread Spectrum Signal is modulated with sequence
  of digits generated by pseudo random number
  generator, so that signal to be transmitted has
  wider bandwidth.
• Frequency Hopping. Signal hops from frequency
  to frequency at fixed intervals. Receiver
  hopping ,in synchronization with transmitte,r
  picks up message.
• 26 MHz available band divided into 250KHz
  channels gt 104 channels that can be divided into
  chunks of 52 frequencies so that adjacent cells
  can use a different chunk.

20
Lower Tier (2)

• Chunking further increases S/I ratio. This
  optimization does not cover the case of three
  clusters that are each others neighbors
• TDMA scheme followed for the communication
  between SUs and LSM each others neighbors
• Minimal Handshaking protocol to maximize power
  savings

21
Upper Tier

• LSM transmits at 915 MHz ISM for communication
  with SUs and at 2.4 GHz ISM for communication
  with adjacent clusters.
• Clock synchronization among LSMs also at 915 MHz
• LSM routes received data to the CSM through the
  other LSMs.
• In the absence of energy constraints IEEE 802.11b
  appears as good choice of wireless network
  standard preferred over expensive custom designed
  ones

22
Sensor Unit

• Components
• SU controller
• 915 MHz radio transceiver
• SRAM Memory
• Low sensitivity, high g accelerometer ( ADLXL120
  ) for extreme event responses
• Sensor module with a high sensitivity low noise
  accelerometer ( 1221 accmtr ) for ambient
  responses
• High resolution low speed A/D converter
• Various other sensors i.e thermometer

23
Sensor Unit (2)

• Operation States
• Sleep. Sensor module and R/F transceiver in sleep
  mode. SU controller and the low sensitivity
  accelerator ( Acc-Low ) intermittently powered.
  Right After Acc-Low startup time elapsed switches
  on the A/D converter and samples the Acc-Low
  output. If the data do not indicate an extreme
  seismic event Acc-Low and A/D are powered off.
• If an event ( 5mg and above ) is detected unit
  enters Awake state. Sampling rate is increased
  accordingly and data are saved in SRAM. Event
  time is noted. Synchronizes with LSM and adjusts
  event time accordingly. Awake state is
  synchronized with TDMA slot.

24
Sensor Unit (3)

• Operation States
• It passes in Semi-Awake state all modules but
  transceiver are on. Active sampling of both
  accelerators output. Semi-Awake periodically
  alternates with Awake state were radio
  transceiver is also on, allowing transmission of
  sensor data to LSM along with active sampling
• After event has passed and all collected data are
  sent SU enters sleep state.
• CSM determines at which instances ambient
  vibration info is to be recorded to enable
  periodic monitoring. SUs usually in sleep states
  are waken up in fixed number of times per day and
  enter the Update State.

25
Sensor Unit (4)

• Operation States
• In the Update state state sensor module is off,
  SU ctlr and transceiver on and Ac-Low cycle
  powered. It synchronizes with LSM and sets a wake
  up timer for the next scheduled monitoring phase.
  After that enters Awake State alternating with
  Semi-Awake transmiting data
• After a predefined time interval it goes back to
  sleep.

26
Local Site Master

• Components
• 915 MHz radio transceiver for upper tier
  communication
• 2.4 GHz radio transceiver for lower tier
  communication
• LSM Controller
• Memory to store the received data to be routed.
•  
• Operates continuously throughout the life of the
• network

27
Communication Protocol

• A special purpose customized TDMA protocol. For
  N sensor units per cluster we have a round of N2
  slots. N for data, 2 for control. A data slot
  assigned per SU. 1st control slot for Synch-Ack
  and second for Global-Synch signal. One packet
  per data slot. Frequency hopping at every slot.
• LSM acknowledges SU packets broadcasting a
  Synch-Ack signal containing ack bits for all
  packets received in the current and the previous
  round. Of course contains the local clock
  information for synchronization of SU clocks with
  the global clock.
• SU notifies LSM before entering the sleep state
  at the end of a monitoring or update state. Uppon
  notification LSM is able to use the SU slot for
  startup packet.
• Synch-acks are appended with the ambient
  monitoring schedules enabling schedule updating
  upon synchronization.

28
Analysis of the proposed monitoring system

• Trade Off conclusions
• Best transmission range, possible with low power
  results, to low transmission rates. Low
  transmission rates support less sensors per
  cluster
• When clusters are in a straight chain then one
  single LSM is the only access point to the CSM
  and therefore receives all the routed data. Worst
  case upper tier data rate requirement is set to
• (2N-n) SU data
• _____________
• link throughput
•  
• where N is total number of SUs and n SUS per
  cluster.

29
Analysis of the proposed monitoring system ( 2 )

• Trade Off conclusions
• CSM receives data generated from all Nodes ( N)
  SU data. Latest 2.4GHz radio transceivers can
  support 10Mbps gt 2200 nodes worst case. But if
  multiple LSMs have access to CSM gt 4400 sensors
• High degree of scalability but it is based on
  very
• optimistic assumptions on the supported bit
  rate.

30
Estimation of the maintenance cycle

• 10 sensors per cluster, extreme event
  duration 3 mins, periodic monitoring phase 3
  mins, and time needed by SU for startup
  synchronization 4 mins.
• Power consumption components on a yearly basis 
• Energy consumed in radio transceiver 275mAh
• Energy consumed in microcontroller unit and Low
  Sensitivity Accelerator 300mAh
• Energy consumed by memory sensor modules and
  others 1000 mAh
• Total of 1600 mAh

31
Implementation

• Network
• Two cluster network
• 80Hz sampling
• frequency
• one channel per SU
• sample resolution 16
• bits
• 5 SUs per cluster
• 0.75 throughput

32
Implementation H/W

• Sensor unit
• An EVK915 module containing a 915 MHz Bluechip
  radio transceiver at 20kbps baud rate. Receiver
  current 12 mA and transceiver 50mA at its peak 50
  db transmit power.
• Interfaced with a lab made sensor board
• PIC Microchip MCU used as the SU controller
• ADXL Accelerometer on the sensor board 
• LSM
• An ProximRangelLAN radio modem operating at 2.4
  GHz with 2 Mbps bit rate. Uses IEEE802.11b
• Interfaced to an EVK915 radio module
• PIC Microchip MCU used as the LSM controller

33
Implementation S/W

• SU and LSM controller were programmed in assembly
  and the code was loaded in the microcontrollers.
• Local clock is a S/W clock based on a 16bit
  timer.
• FHSS patterns obtained from a linear feedback
  shift register implemented in the code as pseudo
  random generator.
• CSM software written in C

34
Limitations

• Applicable only in static structures with regular
  power supplies which are needed for the operation
  of the upper tier.
• LSMs are single points of failure for the SUs in
  its cluster and CSM a single point of failure for
  the whole system!

35
Conclusions

• Significant role of power efficiency on the
  viability of the system.
• System throughput of hundreds or thousands kbps
  for real time performance.
• Required data rate and range require higher power
  consumption.
• The goal of power efficiency is achieved through
  partition in two tiers.
• TDMA communication protocol contributes to
  reduced power consumption.
• Lab and Field testing under realistic vibration
  and environmental conditions

36
Future work

• Self reconfigurable SUs so that they are not
  dependant on a specific LSM. In case of failure
  it can be assigned to the closer LSM in range or
  to a backup one.
• More robust sensor and Site Master Units

37
The end. Questions?