Docstoc

Cyber-Physical Codesign of Distributed Structural Health Monitoring With Wireless Sensor Networks - Washington University in St

Document Sample
Cyber-Physical Codesign of Distributed Structural Health Monitoring With Wireless Sensor Networks - Washington University in St Powered By Docstoc
					Cyber-Physical Codesign of Distributed
Structural Health Monitoring With
Wireless Sensor Networks

    Gregory Hackmann∗, Weijun Guo∗, Guirong Yan†,
    Chenyang Lu∗, Shirley Dyke†
    ∗ Department of Computer Science and Engineering,

      Washington University in St. Louis
    † School of Mechanical Engineering, Purdue University
Structural Health Monitoring (SHM)
Ø    Problem: detect and localize
     damage to a structure
Ø    Wireless sensor networks (WSNs)
     monitor at unprecedented
     temporal and spatial granularities

Ø    Key Challenges:
    q Long-term monitoring
    q Rapid on-demand analysis
    q Resource and energy constraints




                                          2
Related Work
Ø Wisden [Xu, SenSys 04]
   q Services for reliable transmission of raw data
Ø Golden Gate Bridge [Kim, IPSN 07]
   q 46-hop network deployed along Golden Gate Bridge
Ø BriMon [Chebrolu, MobiSys 08]
   q Trains as data mules
Ø Torre Aquila [Ceriotti, IPSN 09]
   q Heterogeneous sensors, most with low data rate
Ø Key limitation: separate the designs of
   q Cyber (data transport) components
   q Physical (damage detection) components
                                                        3
Cyber-Physical Codesign Approach
Ø Cyber-physical codesign: jointly develop WSN architecture
  and SHM approach into a hierarchical architecture
Ø Key ideas specific to this work:
   q Embedding processing into hierarchical architecture
   q Multi-resolution damage localization that exploits local nature of
     SHM approach
Ø Example of general approach of integrating hierarchical
  sensing and control




                                                                          4
Prior Work: DLAC
Ø Proof-of-concept system pushed portions of Damage
  Location Assurance Criterion (DLAC) algorithm into
  network [Castaneda, ASEM 08], [Hackmann, RTSS 08]

Ø Careful system design highlights potential gains for cyber-
  physical codesign approach
   q 66% lower latency and 71% lower energy consumption than
     centralized scheme
Ø But a limited architecture with no collaboration among
  sensors => significant limitations in SHM capabilities

                                                                5
Flexibility-Based Methods
Ø Structures flex slightly when a force is applied

Ø Structural weakening => decreased stiffness
Ø Flexibility acts as a “signature” of the structure’s health

Ø Two flexibility-based methods of interest for our work
   q Beam-like structures: Angles-Between-String-and-Horizon
     flexibility method (ASHFM) [Duan, J. Structural Engineering and
                              θ
     Mechanics 09]
   q Truss-like structures: Axial Strain flexibility method (ASFM) [Yan, J.
     Smart Structures and Systems 09]
                                                                          6
Embedding Processing Into CPS
Architecture
Ø Sensors form physically-colocated groups
Ø Group members collect raw vibration data and process into
  power spectrum data
Ø Group leaders collect corresponding power spectrum data
  from children, correlating into modal parameters (natural
  frequencies + mode shapes)




                                                          7
Embedding Processing Into CPS
Architecture
                 Ø Base station collects modal
                   parameters from group
                   leaders, completes
                   processing into structural
                   flexibility
                 Ø Output is compared against
                   “baseline” collected when
                   structure was known to be
                   healthy
                 Ø Differences in flexibility can
                   be used to detect and
                   localize damage
                                                    8
 Standard Data Flow
Group Member                 Group Leader                    Base Station


    Sensing
                             Cross Spectral                    Flexibility
2xD
                                Density
ints
         FFT                 D
                          matrices                       D: # of samples
   D
floats                                                   P: # of natural freq.
                             Singular Value              (D » P)
Power Spectrum               Decomposition

                    D
                 floats                       P natural frequencies +
                                                  mode shapes
                                                                             9
 Enhanced Distributed Data Flow
Group Member                 Group Leader                    Base Station

    Sensing
2xD                          Cross Spectral                    Flexibility
ints                            Density
         FFT
                             P
   D                      matrices
floats                                                   D: # of samples
                                                         P: # of natural freq.
Power Spectrum               Singular Value              (D » P)
                             Decomposition
   D
floats
  Peak Picking      P                         P natural frequencies +
                 floats                           mode shapes
                                                                             10
Multi-Resolution Damage
Localization
Ø Under ASHFM and ASFM, only a handful of sensors are
  needed to detect damage
Ø As more sensors are added, localization gets more fine-
  grained
Ø Significant energy savings by exploiting localized nature of
  flexibility-based approach




                                                             11
Implementation
Ø Hardware platform: Intel/Crossbow
  Imote2 + ITS400 sensorboard
   q   13 – 416 MHz PXA271 XScale CPU
   q   32 MB ROM, 32 MB SDRAM
   q   CC2420 802.15.4-compliant radio
   q   3-axis accelerometer on sensor board


Ø Software platform
   q TinyOS 1.1 operating system
   q UIUC’s ISHM toolsuite used for sensing,
     reliable communication, and time sync

                                               12
Evaluation: Cantilever Beam
Ø 2.75 m long steel cantilever beam
  fixed to the ground
Ø Eight motes deployed directly on
  beam
Ø Damage introduced by attaching a
  steel plate between two sensors

Ø Result: damage correctly localized
  between sensors 3 and 4


                                       13
Evaluation: Truss
Ø Simulation of 5.6 m, 14-member steel truss structure at
  UIUC

Ø Simulated sensor data generated in MATLAB and injected
  into live application using “fake” sensor driver
   q Intact data set: no damages
   q Damaged data set: three members reduced on left side of truss,
     four on right side




                                                                      14
Evaluation: Truss
Ø Level 1: nine sensors at uniform points along truss’s length

                                                Damage identified
                                                  on right half




            Damage identified
               on left half




                                                                    15
Evaluation: Truss
Ø Level 2: move all nine sensors to respective halves (higher
  density)




          Damage localized correctly to all seven members




                                                                16
Evaluation: Truss
Ø Codesigned architecture reduces           Cluster Member
  communication latency from           Synchronization       12.1 J
  estimated 87 s to 0.21 s             Sensing               23.0 J
                                       Computation           9.28 J
Ø 78.9% of energy attributable to      Communication         0.08 J
  synchronization and sensing
                                              Cluster Head
                                       Synchronization       16.2 J
Ø Compare to theoretical energy        Sensing               21.2 J
  supply of 20,250 J (3x 1.5 V, 1250   Computation           8.52 J
  mAh AAA batteries)                   Communication         0.76 J

                                                                 17
Conclusion
Ø Our system integrates flexibility-based SHM methods with
  an efficient distributed computing architecture
Ø Multi-level search strategy only activates sensors in area of
  interest; many sensors remain asleep
Ø Shown to localize damage to real beam and simulated
  truss structures, including multiple-damage case on latter

Ø Long-term goal: a general cyber-physical codesign
  approach to integrated sensing and control

             http://www.cse.wustl.edu/~lu/shm/
                                                             18
19
Evaluation: Cantilever Beam
Ø Level 1 localization: six sensors activated
  uniformly across beam




   Damage localized between
       sensors 2 and 5




                                                20
Evaluation: Cantilever Beam
Ø Level 2 localization: 2 additional
  sensors activated at higher density
  around damage




    Damage localized to plate
    between sensors 3 and 4




                                        21

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:1
posted:9/25/2013
language:
pages:21