Presentation is loading. Please wait.

Presentation is loading. Please wait.

Embedded Networks Laboratory 1 Embedded Sensing of Structures : A Reality Check Krishna Kant Chintalapudi, Jeongyeup Paek, Nupur Kothari, Sumit Rangwala,

Similar presentations


Presentation on theme: "Embedded Networks Laboratory 1 Embedded Sensing of Structures : A Reality Check Krishna Kant Chintalapudi, Jeongyeup Paek, Nupur Kothari, Sumit Rangwala,"— Presentation transcript:

1 Embedded Networks Laboratory 1 Embedded Sensing of Structures : A Reality Check Krishna Kant Chintalapudi, Jeongyeup Paek, Nupur Kothari, Sumit Rangwala, Ramesh Govindan, Erik Johnson Jeongyeup Paek

2 Embedded Networks Laboratory 2 Goals of the Talk Original vision Where are we today? Where are we heading? –Is the original vision still meaningful? “ Millions of tiny sensors embedded in concrete detect damages in buildings and bridges ”

3 Embedded Networks Laboratory 3 Agenda Introduction to Structural Health Monitoring Requirements of SHM Applications WISDEN - a wireless sensor network data acquisition system NetSHM – a programmable sensor network for SHM applications Speculations about the future

4 Embedded Networks Laboratory 4 What Is Structural Health Monitoring (SHM)? Structural integrity assessment for buildings, bridges, offshore oil rigs, aerospace structures etc. Goals of SHM: –Detection “is there damage?” –Localization “where is the damage?” –Quantification “how severe?” –Prognosis “future prediction” Why SHM?

5 Embedded Networks Laboratory 5 SHM Basics Measure and analyze structural vibrations induced due to heavy winds or earthquakes, etc Principles behind structural algorithms can be illustrated by strings –Structural response is composed of several harmonics - modes –Mode = Damages alter the structural properties and hence the modes Structural response is measured by using sensors (accelerometers, strain gauges) at several locations in the structure

6 Embedded Networks Laboratory 6 Sensor Networks for SHM Current SHM –Bi-annual visual inspections (most common) Limitations of human accessibility and error Catastrophic failure between inspections –Expensive wired data acquisition systems Extremely high installation, cabling, and maintenance cost Wireless Sensor Network based SHM system –Flexible, fast and low cost deployments –No cabling cost!!

7 Embedded Networks Laboratory 7 Agenda Introduction to Structural Health Monitoring Requirements of SHM Applications WISDEN - a wireless sensor network data acquisition system NetSHM – a programmable sensor network for SHM applications Speculations about the future

8 Embedded Networks Laboratory 8 Existing SHM Techniques Damage DetectionDamage Localization Time Series Modal Frequency Mode Shape Neural Networks Time Domain Frequency Domain Changes in ARMA coefficients Changes in modal frequencies Changes in mode shape Train neural networks with data Reconstruct a structural model from data Reconstruct structural model using mode shapes

9 Embedded Networks Laboratory 9 Basic Requirements for SHM Applications Reliable Delivery –SHM applications are loss-intolerant, sensors need to transmit data reliably Time Synchronization –Data from various sensors should be time-synchronized to within 100 micro-sec for damage localization. High Data Rates –A hundred tri-axial sensors sampling at 500Hz can generate a data rate of 5Mbps. Dense Sensing –The larger the number of sensors the better the performance

10 Embedded Networks Laboratory 10 Importance of In-Network Processing Sensor networks are expected to last for several months or even a year without human intervention With high data rate radio communication and sensing, nodes will typically not last more than few days. In-network processing can lead to long lived SHM systems by reducing communication overhead Most SHM techniques can leverage local computation at node to minimize radio communication –ARMA coefficient for time series based damage detection –FFT for modal frequency shift based damage detection

11 Embedded Networks Laboratory 11 Agenda Introduction to Structural Health Monitoring Requirements of SHM Applications WISDEN - a wireless sensor network data acquisition system NetSHM – a programmable sensor network for SHM applications Speculations about the future

12 Embedded Networks Laboratory 12 Wisden First step –Replace the existing wired data acquisition system Wisden –Wireless sensor network based data acquisition system –Allows continuous sampling and reliable logging of time-synchronized structural response data Advantages –Flexibility Nodes self-organize into a multi-hop network. Nodes can be inserted in and out of the network dynamically –Low time and cost of installation

13 Embedded Networks Laboratory 13 Wisden overview Three components of Wisden –Reliability Over multiple hops with end-to-end and hop-by-hop recovery –Time-synchronization Novel low-overhead residence time based approach –Data compression Necessary at the source nodes to relieve bandwidth bottleneck and reduce communication overhead. Onset detection – transmit only relevant data “A Wireless Sensor Network for Structural Monitoring”, Ning Xu, Sumit Rangwala, Krishna Chintalapudi, Deepak Ganesan, Alan Broad, Ramesh Govindan, Deborah Estrin, In Proceedings of the ACM Conference on Embedded Networked Sensor Systems, Nov.2004

14 Embedded Networks Laboratory 14 Onset Detection Why transmit data when nothing is happening? Detect onset of events at the sensor and transmit only when something is happening “A Wireless Sensor Network for Structural Health Monitoring: Performance and Experience”, Jeongyeup Paek, Krishna Chintalapudi, John Caffrey, Ramesh Govindan, Sami Masri, In Proceedings of the IEEE Workshop on Embedded Networked Sensors, May.2005 Data not transmitted during quiescent period

15 Embedded Networks Laboratory 15 Deployment of WISDEN

16 Embedded Networks Laboratory 16 Deployment Experiences (1) Seismic Structure –Structural vibrations are highly damped, last less than a second Higher sampling rates are needed to collect enough samples for analysis (>200Hz) –Platform limitations (such as EEPROM access latencies) proved to be the obstacles for high sampling rates –After the development of onset-detection and careful re-engineering, Wisden was able to achieve 200Hz

17 Embedded Networks Laboratory 17 Deployment Experiences (2) Four Seasons Building –Communication environment was very lossy Avg. delivery rate 81% and worst case of 30% The path lengths were often 2-3 hops and sometimes even higher Frequent route changes occurred due to the variability of the wireless links –Rate control and hop-by-hop retransmissions were required –Does not scale As number of nodes grows, the bandwidth bottleneck becomes significant Leads to our next step hierarchical system: NetSHM

18 Embedded Networks Laboratory 18 Agenda Introduction to Structural Health Monitoring Requirements of SHM Applications WISDEN - a wireless sensor network data acquisition system NetSHM – a programmable sensor network for SHM applications Speculations about the future

19 Embedded Networks Laboratory 19 NetSHM NetSHM is the next step to WISDEN. A sensor network system that Structural engineers can program in higher level language such as Matlab/C An SHM engineer should be able to write and test variety of algorithms without having to understand the underlying sensor network details The system should be evolvable – we should not need to rewrite applications when the technology evolves

20 Embedded Networks Laboratory 20 Architecture of NetSHM Two-level Hierarchy –For scalability, a higher more endowed layer is required to manage the aggregate data rates generated by the motes. Isolate application code from wireless sensor network details –Wireless sensor network provides a generic task interface getSamples(startTime, noSamples, sampFreq, axis) getFFTSamples(startTime,noSam ples,sampFreq,axis,fftSize)

21 Embedded Networks Laboratory 21 What does code isolation buy us? Reusability –Application programmers can use the generic task interface and write many different SHM applications. –Basic SHM library functions can be provided on motes: FFT, auto-correlation, ARMA coefficient estimation, spectral estimation etc. Evolvability –If a new mote comes along with greater processing power, just add new functionality, no need to rewrite application.

22 Embedded Networks Laboratory 22 Application on NetSHM function shifts = getModalShiftsFromBuilding() % create a group for sensors gidSensors = NetSHMCreateGroup([1,2,3,4]); %create a group for actuators gidActuators = NetSHMCreateGroup([5]); %actuate after 22 seconds NetSHMCmdActuate(gidActuators,22); %collect structural response starting 20 seconds from now, % 4000 samples at 200Hz,along x-axis only, samples = NetSHMCmdGetSamples(gidSensors,20,200,1,4); %find modal frequencies modes = findModes(samples); %read original modes load OriginalModes; shifts = findModalFreqShifts(modes,OriginalModes);

23 Embedded Networks Laboratory 23 The Stacks Gateway node stack Mote-class node stack

24 Embedded Networks Laboratory 24 Deployment Building Details –4–48 inches high, 4 floors, 60 lbs –F–Floors –1/2 x 12 x 18 aluminum plates –s–steel 1/2 x 1/8 inch steel columns –5–5.5 lb/inch spring braces –4–4 actuators on the top floor –8–8 motes, 2/floor –d–dual axis, 200Hz, 2 starGates 4 Test Cases –b–braces from floor 4 removed –b–braces from floor 3 removed –b–braces from floor 2 removed –b–braces from floor 2 and 4 removed Actuators Sensors Motes

25 Embedded Networks Laboratory 25 Damage Detection and Localization on scaled model

26 Embedded Networks Laboratory 26 Agenda Introduction to Structural Health Monitoring Requirements of SHM Applications WISDEN - a wireless sensor network data acquisition system NetSHM – a programmable sensor network for SHM applications Speculations about the future

27 Embedded Networks Laboratory 27 Limitations of Wireless Sensor Network based SHM today Hundreds of nodes per structure Limited lifetime –Couple of days with continuous sampling –Up to couple of months with scheduled monitoring Limited in-network processing –Platform limitations (eg. mica2, micaz) Memory (FFT, ARMA, etc) Processing (floating point, etc)

28 Embedded Networks Laboratory 28 What next? Vision of millions of embedded sensors in concrete seems a bit too farfetched –Energy, form factor, communication, etc Within the next few years, NetSHM like systems will encourage SHM engineers to migrate to sensor network systems Most of the data processing will migrate into the sensors within the next five years with the advent of improved sensor platforms We believe that the wired sensing will be almost entirely replaced by wireless networks within the next ten years

29 Embedded Networks Laboratory 29


Download ppt "Embedded Networks Laboratory 1 Embedded Sensing of Structures : A Reality Check Krishna Kant Chintalapudi, Jeongyeup Paek, Nupur Kothari, Sumit Rangwala,"

Similar presentations


Ads by Google