Monitoring of Civil Infrastructure: from Research to Engineering Practice

Transcription

1 Monitoring of Civil Infrastructure: from Research to Engineering Practice Billie F. Spencer, Jr. Nathan M. and Anne M. Newmark Endowed Chair in Civil Engineering Director, Newmark Structural Engineering Laboratory Director, MUST-SIM: Director, Smart Structures Technology Laboratory (SSTL)

2 Where is the University of Illinois (UIUC)?

3 Where is the University of Illinois (UIUC)?

4 Outline Background and Motivation Enabling Wireless Smart Sensor Technology o High-fidelity Hardware o Service-oriented Software Framework Full-scale Implementations o Jindo Bridge, Korea o CN Railroad Bridge, USA Current Activities Concluding Remarks

5 Background and Motivation

6 Computers are Faster and Cheaper. $400,000/MIPS (Cray-I).. $500/MIPS (i860) $1/MIPS

7 Where Computing is Done Number Crunching Data Storage productivity interactive year streaming information to/from physical world

8 Networking Plays an Important Role 1940s The first computer 1960s High performance Computing 1980s Personal Computers 1990s Internet Now IoT/WSN Cyber Physical Systems

9 (2013)

10 Monitor Infrastructure To detect and localize damage To monitor and control the construction process To validate the structural designs and characterize performance To characterize loads in situ To assist with maintenance /repair/replacement strategies To inform the retrofit of infrastructures To assist with emergency response efforts, including building evacuation and traffic control Fatigue (Bay Bridge in CA, 2009) Corrosion Sung-su bridge collapse in Korea (1994)

11 Structural Health Monitoring Limitation of traditional methods Dense arrays of sensor are required to effectively monitor structures Wired monitoring systems are expensive, with much of the cost derived from cabling and installation Centralized data collection is not challenging for monitoring large civil infrastructure Cabling & instrumentation for 400 wired strain sensors Bill Emerson Memorial Bridge SHM system: $1.3M for 86 sensors, ~$15k/sensor (Caicedo et al. 2002; Celebi et al. 2004)

12 Enabling Wireless Smart Sensor Technology * low cost * ease of installation * accurate data

13 Smart Sensors But... what is a smart sensor? Radio Embedded Computing Data Storage Local Power First open source hardware and software platforms developed at Berkeley in the late 1990s as part of DARPA s Smart Dust project (Mica family of smart sensors).

14 Historic Decade of Smart Sensors WINS 1 (1999) SmartDust WeC (1999) BTnode rev3 (2004) U3 (2002) EYES (2003) Intel Imote (2004) Prototype by Prof. Lynch (2002) Crossbow Mica2 (2004) Imote2 (2006)

15 Full-scale Implementations to Bridges (selected) Implementation Purpose Sensor Platform # of Nodes Sensor (channels) Test duration In-network Processing Results Alamosa Canyon (Lynch et al. 2003) Wireless system proof-ofconcept WiMMS (prototype) 7 Accel. (14) Short-term Independent FFT Modal frequencies Ben Franklin (Galbreath et al. 2004) Demonstration of wireless system with remote programming Microstrain SG-Link 10 Strain + Temp. (10) Continuous Medium-term None Streaming realtime strain time histories Guemdang (Lynch et al. 2006) Wireless sensor prototype and embedded computing validation WiMMS (prototype) 14 Accel. (14) Short-term Independent FFT and peak picking Modal frequencies, ODS Golden Gate (Pakzad et al. 2008) Test wireless sensor network requiring multihop communication MicaZ 64 Accel. (128) Temp. (64) Medium-term None Modal analysis after central data collection Gi-Lu (Weng et al. 2008) Test for health monitoring of cable-stayed bridge WiMMS (prototype) 12 Accel. (12) Short-term None Modal analysis and cable tension force New Carquinez (Lynch et al. 2010) Wireless system proof-ofconcept Narada/ Imote2 30 Accel. (14) Long-term SSI Modal Analysis Mode Shapes

16 Full-scale Implementations to Bridges (selected) Implementation Alamosa Canyon (Lynch et al. 2003) Ben Franklin (Galbreath et al. 2004) Guemdang (Lynch et al. 2006) Purpose Wireless system proof-ofconcept Demonstration of wireless system with remote programming Sensor Platform WiMMS (prototype) Microstrain SG-Link # of Nodes Sensor (channels) Test duration 7 Accel. (14) Short-term 10 Strain + Temp. (10) Continuous Medium-term Wireless sensor prototype WiMMS and embedded back computing to the 14 base Accel. (14) station! Short-term (prototype) validation In-network Processing Independent FFT None Independent FFT and peak picking Results Modal frequencies Streaming realtime strain time histories Requires 12 hours to transmit 80 seconds of data Modal frequencies, ODS Golden Gate (Pakzad et al. 2008) Test wireless sensor network requiring multihop communication MicaZ 64 Accel. (128) Temp. (64) Medium-term None Modal analysis after central data collection Gi-Lu (Weng et al. 2008) New Carquinez (Lynch et al. 2010) WiMMS Despite the (prototype) fact that smart sensors have been readily available for over a Test for health monitoring of cable-stayed bridge 12 Accel. (12) Short-term None Wireless system proof-ofconcept decade, full-scale Narada/ SSI Modal 30 Accel. (14) Long-term Imote2 implementations are Analysis still limited. Modal analysis and cable tension force Mode Shapes

17 What s Been Lacking? Hardware o Platform with computational capacity for high-data rate applications and distributed computing o Sensor hardware that to produce high-fidelity data that is appropriate for SHM Software o Middleware services to acquire high-fidelity data o Application software to implement distributed SHM o Flexible software that supports network and application scalability Full-scale implementation considerations o Communication performance evaluation o Autonomous network operation o Power management o Fault tolerance

18 High-fidelity Hardware

19 Imote2 with Sensor Boards Sensor Board External Antenna (Antenova 2.4GHz) imote2 Battery Board w/ 3 AAA Batteries From MEMSIC (2010) MicaZ Telos imote2 Microprocessor Atmel ATmega128L TI MSP430 Intel XScalePXA271 Clock speed (MHz) Active Power (mw) MHz Non-volatile memory (bytes) 128 K (Flash) K (EEPROM) 48K (Flash) 32 M (Flash) Volatile memory (bytes) 4 K 1024K 256 K + 32 M (SDRAM) Dimensions (mm) / Weight (g) 58*32*7 / 18 65*31*6 / *48*9 / 12 Imote2 and Basic Sensor Board LED Basic I/O Advanced I/O 36mm PMIC Basic I/O PXA27 Mini USB Connector CC2420 Advanced I/O Antenna SMA Connector 48mm Imote2: Top and Bottom Basic Sensor Board

20 Aliasing Time (sec) Cannot have significant energy above the Nyquist frequency (f s /2) Anti-aliasing filters must be used in dynamic measurements to preserve signal integrity Consider a 48 Hz signal What if the signal is sampled at 50 Hz? The resulting measured signal has an apparent frequency of 2 Hz? Aliasing f N f s

21 Multifunctional Sensor Boards SHM-A and SHM-H Board Accelerometer ST Microelectronics LIS344ALH Light Sensor TAOS 2561 Humidity & Temp. Sensor SHT11 External Input Connector Basic Connector OP AMP TI OPA 4344 SHM-A Board (top) SHM-A Board (bottom) Programmable Filter w/ ADC Quickfilter QF4A512 Basic Connector SHM-H Board SHM-DAQ Board Data Acquisition using commercial voltage-output sensors Output Voltage: -5 to 5 V 0 to 5 V Interfaced Voltage Anemometer (RM Young, Model 8100)

22 Strain monitoring using SHM-S board Strain sensor board for Imote2 platform High-throughput synchronized strain monitoring High precision auto-balanced Wheatstone bridge Up to 2500 times gain 0.3 µ-strain resolution at 20Hz B/W Stacked use with SHM-A or SHM-DAQ board Both foil-type strain gage and magnetic strain sensor can be used with SHM-S board RemoteCommand SHMSAutoBalance SHM-S board: top, stacked on SHM-DAQ and stacked on SHM-A Tokyo Sokki magnet strain checker

23 High-precision Strain sensor board (SHM-S board) SHM-S vs. Ni-DAQ (foil type) -strain 50 0 a) Time history Wired: Ni-DAQ Wireless: SHMS -strain 2 /Hz 10 0 b) PSD Wired: Ni-DAQ Wireless: SHMS -strain time(sec) c) Detail (high-level strain) Wired: Ni-DAQ time(sec) Wireless: SHMS -strain frequency(hz) d) Detail (low-level strain) Wired: Ni-DAQ Wireless: SHMS time(sec)

24 Imote2 Sensor Enclosure Rechargeable Battery Antenna Cable Solar Panel External 5dBi Antenna Bolt & Spacer Acrylic Jig Uni-directional magnet

25 Service-oriented Software Framework

26 Service-Oriented Architecture (SOA) Previous smart sensor applications: Significant effort to create very specific applications Difficult to modify for other applications, even with extensive CS knowledge Service-Oriented Architecture simplifies SHM software development Applications are comprised of manageable, modular services that exchange data in a common format The middleware framework connects the services by providing communication and coordination SDLV Application Services Numerical Services Foundation Services SHM Application

27 Time Synchronization Each sensor has its own local clock which drifts over time Synchronization errors can be reduced to ~20 s Not the whole story beacon exchange beacon time & adjust clocks

28 Time Synchronization Uncertainty in start of acquisition Independent processors Sensor 1 t 0,1 3 Sensor 2 Resampling middleware service achieves t 0,2 precise synchronized sensing Samples Samples Time (sec) Time (sec) 28

29 Nonlinear Clock Drift Generally, clock speeds are quite constant if temperatures do not change However, the temperature of the Imote2 oscillator could change due to internal and external heating Clock offset (μs) Node 9 Node 82 Node 150 Node 69 Temperature ( o C ) Time (sec) Time (sec)

30 Time Synchronization Broadcast beacon signals (global time) during sensing Acc Acc Acc t t Resampling middleware service achieves precise synchronized sensing Leaf node 1 Leaf node 2 Leaf node 3 t These marked global time stamps are then used to adjust the individual local time stamps to remove the effect of both the uncertainty in starting time and nonlinear clock drift Gateway node Base Station

31 Power Management and Energy Harvesting Effective Power Management with Autonomous Operation SnoozeAlarm: ThresholdSentry: AutoMonitor: Deep Sleep Deep Sleep Deep Sleep Deep Sleep WakeUp and Listen Energy Harvesting Base Station Solar panel (or Micro wind turbine) + Rechargeable Battery PMIC charger manipulates voltage and current for fast and stable charging. PMIC Imote2 Solar Panels Micro Wind Turbine

32 ISHMP Services Toolsuite Fault tolerant WSSN Skipping of unresponsive nodes Data storage in non-volatile memory Monitoring of sensor power Exclusion of low-power sensor nodes Watchdog timer and etc Enhanced operation Autonomous resuming of AutoMonitor ThresholdSentry for multi-hop communication notification of structural response and network anomalies

33 Full-scale Implementation: Jindo Bridge

34 Seoul Daejeon (KAIST) US-Korea-Japan Collaborative Project on SHM Test-bed Using Wireless Smart Sensor Network (September ) Jindo Jindo Bridges Jeju Island US : B.F. Spencer, Jr. & G. Agha (UIUC) Korea : H.J. Jung & C.B. Yun (KAIST), H.K. Kim (SNU) J.W. Seo (Hyundai Institute of Const. Tech.) Japan : Y. Fujino & T. Nagayama (U. of Tokyo) 2 nd Jindo Bridge Haenam (Inland) 2 nd Jindo Bridge Type Spans Girder Design velocity Designed by Constructed by Cable-stayed bridge = 484m Steel box (12.55m width) 70 km/hr Yooshin cooperation (2000, Korea) Hyundai construction (2006, Korea)

35 Deployment in 2010

36 Deployment in 2010 Pylon node Wind turbine Cable node Base station Pylon node 3D ultra-sonic anemometer TeamViewer (for remotely access to basestation) Deck node

37 Deployment in Jindo side (South) Haenam side (North) T East West 118 West East West Deck networks WT H T H H H 151 H H H H W W W Cable networks S 70 East West 32 ST T 101 Jindo side Pylon East Haenam side Pylon : Jindo Deck network (single hop) : Haenam Deck network (single hop) : Jindo Cable network (multi hop) : Haenam Cable network (single hop) W : Anemometer (SHM-DAQ) T : Temperature WT : Wind Turbine power H : High-sensitivity (SHM-H) S : Strain sensor (SHM-S1) ST : Strain + Temp correction (SHM-S2)

38 Deployment in Jindo side (South) Haenam side (North) East West 118 West East West Deck networks with 113 sensor nodes were deployed WT H T H H H 151 H H H H W W W Cable networks T In total, 669 sensing channels S 70 East West 32 ST T 101 Jindo side Pylon East Haenam side Pylon : Jindo Deck network (single hop) : Haenam Deck network (single hop) : Jindo Cable network (multi hop) : Haenam Cable network (single hop) W : Anemometer (SHM-DAQ) T : Temperature WT : Wind Turbine power H : High-sensitivity (SHM-H) S : Strain sensor (SHM-S1) ST : Strain + Temp correction (SHM-S2)

39 Evaluation: Wind Monitoring SHM-DAQ + Anemometer Wind Speed 15-20m/s Haenam (Inland) Wind Direction Wind Speed (m/s) Measured Wind by KMA-Jindo (Mt. Chun-Chal) at Sep. 01, Wind Speed (m/s) Jindo Azimuth Dir. (Deg) Azimuth Dir. (Deg) Elevation (Deg) Time (sec) Anemometer: Jindo-side (at 21:14) Elevation (Deg) Time (sec) Anemometer: Haenam-side (at 21:12)

40 Evaluation: Wind Monitoring SHM-DAQ + Anemometer Wind Speed 15-20m/s Haenam (Inland) Wind Speed (m/s) Wind Direction Ch1: wind speed Measured Wind by KMA-Jindo (Mt. Chun-Chal) Ch2: wind direction at Sep. 01, (horizontal) Ch3: wind direction (vertical) Wind Speed (m/s) Narrow Strait: Increased wind speed - Slight change in direction Jindo Azimuth Dir. (Deg) Azimuth Dir. (Deg) Elevation (Deg) Time (sec) Anemometer: Jindo-side (at 21:14) Elevation (Deg) Time (sec) Anemometer: Haenam-side (at 21:12) 40

41 Evaluation: Acceleration Responses acc(mg) acc(mg) acc(mg) acc(mg) 1 (center span) time(s) time(s) acc(mg) acc(mg) acc(mg) acc(mg) time(s) z-axis time(s) time(s) 8 acc(mg) time(s) time(s) time(s) time(s) freq(hz) time(s) acc(mg) (pylon) time(s) acc(mg) acc(mg) PSD (Haenam side deck) time(s) acc(mg) acc(mg) (girder end) time(s) time(s)

42 Evaluation: System ID NExT & ERA method Identified frequency range: 0~3Hz 1000 seconds data at 25Hz sampling DV1 Mode Name Identified natural frequencies (Hz) Wired System (2007) FE analysis WSSN (During Kompasu) Haenam Jindo Avg. DV DV DV DV DV DV DT DV DV DV DV2 DV3 DT1 DV9

43 Cable Tension Force Estimation Vibration-based Tension Force Estimation Least-square method using high mode frequencies fn T EI n 2 4 a (Shimada 1994, Park et al. 2008) n 4mLeff 4mLeff Practical Formulas are not applicable due to lack of information. Linear regression to find a and b Only two cable properties are required (m and L eff ), 2 while EI can be obtained by regression result. Estimated Tension forces T ml a 4 eff bn 2 Cable Routine inspection In 2007 In 2008 Tension force (tonf) Installed WSSNs in 2010 (Avg.) HC (2.04) * HC (-3.19) * HC (0.90) * HC (3.00) * HC (2.18) * JC (1.30) * JC (-1.09) * JC (2.15) * JC (2.33) * JC (0.14) * Tension (tonf) Design Tension Park et al. (2008) CableTensionEstimation c1 c2 c3 c4 c5 c6 c7 c8 c9 c11 c12 Cable ID [ Cable tension comparisons ]

44 Deployment Achievements Hardware 669 sensors at 113 nodes High-sensitivity accelerometer board Solar panel with rechargeable battery Wind-induced energy harvesting Software Multi-hop reliable data transfer and printing Fault tolerant operation Remote software updates Better user interface Analysis & Monitoring Modal analysis with system synchronization Vibration-based tension estimation Multi-scale structural health monitoring (Modal information + Cable tension force) Wind-vibration correlation analysis Collaboration International collaboration for test-bed Web-based data repository and sharing

45 Deployment Achievements Hardware 669 sensors at 113 nodes High-sensitivity accelerometer board Solar panel with rechargeable battery Wind-induced energy harvesting Software Multi-hop reliable data transfer and printing Fault tolerant operation Remote software updates Better user interface World s largest deployment to date of wireless sensors for civil infrastructures monitoring. Analysis & Monitoring World s first long-term deployment. Collaboration Modal analysis with system synchronization Vibration-based tension estimation Multi-scale structural health monitoring (Modal information + Cable tension force) Wind-vibration correlation analysis International collaboration for test-bed Web-based data repository and sharing

46 Full-scale Implementation: CN Railroad Bridge

47 Campaign Monitoring of Railroad Bridges With approximately 100,000 railroad bridges in the United States, the priority of the railroad bridge engineers is to cost-effectively maintain their bridge network and railroad operations

48 Slide 48 Importance of Railroads Today in Freight Transportation in North America

49 Target bridge: CN 1 and CN 2 CN Railroad Bridge Description CN has identified a four track steel truss bridge over the Little Calumet River at MP 16.9 on the south side of Chicago for instrumentation and testing. North

50 21SW: (1) conventional strain gage 22SW, 23SW: (2) magnetic strain gages Experiment Sensors Layout 01SW, 02SW, 03SW 01SW, 02SW, 03SW: (3) conventional strain gages installed at CN1 rail 001SW, (inside bridge) 002SW 21SW, 22SW, 23SW ACC 3 ACC 2 Power source ACC 1 001SW, 002SW: (2) strain gages installed at CN1 rail (outside bridge) ACC 4 Location for base stations does not fault the track

51 Matlab FEM Model of the Structure Typical element built-in section Section D-D 754 elements total 25 sections from CN shop drawings Very rigid floor system All connections between elements are assumed rigid A.R.E.A. and I.C. Specifications Materials A36 ASTM Rivets A502 ASTM High strength bolts No welding

52 Modal Response Update

53 Rail Strain Data (North Approach)

54 Wireless strain measurements/load estimations CN Little Calumet wheel loading and rail strains with NS freight train Wheel weight, w (kips) NS train consist did not include car load #10 (skipped from 9 to 11) Based on similar strain spacing and total number of cars from strains, car load #10 estimated as #5 Last loaded car (#31) shows higher strain (higher impact) Wheel weight Wheel weight envelope Rail strain Train consist shows a longer car (# 43), strain record captured tri-axial car Rail strain, s (ue) 0 0 Train slowed down while crossing the bridge: car consist information is drawn to match the timing of strain record = 2 engines followed by 49 cars Time, t (sec)

55 Axial Strain in U5-L4 North L4 U5 Axial Strain Node 25 shear strain at Calumet M ic r o s t r a in, u e MPH 25 MPH 40 MPH 50 MPH Scaled time to 50 MPH train, sec

56 FE Model Validation 80 Strain Comparison Strain, s ( ) Field FE Model Time, t (sec) The two strain signatures as the work train passes over the bridge are very similar. These results demonstrate the predictive power of the FE model. Slide 56

57 Strain Estimation under Known Loads Compression Tension This model provides a good tool for understanding the behavior of the bridge under in-service loads. Slide 57

58 Consequence-based Management Framework for Railroad Bridge Infrastructure Track Class Probability MPH Bridge Condition Fragility Curves 60 MPH 40 MPH 25 MPH 10 MPH OUTAGE Displacement, d (mm) Hazard Definition & Assessment Inventory Fragility Slow Order Risk Update with campaign monitoring Decision Support Estimated Expenses Monitoring + Analyses Expenses 0.5 Maintenance, Repair, Replacement (MRR) Prioritization Expected Expenses per Measured Displacement 1.5 Track Outage Permanent Slow Order (PSO) Temporary Slow Order (TSO) 1 Inspection Total Cost, TC ($M) Network Level Service Limit State Bridge Transverse Displacement, d(in) Slide 58

59 Current Activities

60 xnode : Next Generation Wireless Smart Sensor Sensor Board Radio /Power Board 1 Processor Board

61 xnode : Hardware Performance Enhancements Platform Processing Speeds up to 200 MHz Expandable Data Storage Radio Range Over 1km On-board ADC Energy Harvesting On-board Strain Circuit MICAz WaspMote Microstrain N/A - - imote Martlet xnode Other Key Enhancements Over Previous Generation WSS The maximum sampling rate is not limited by the size of network due to burst transmission of large buffers of data with reliable protocols Intelligent power management for long-term monitoring 24 bit ADC with up to 16 khz sampling rate for high-fidelity data acquisition 3-ch on-board accelerometers and 3-ch on-board strain circuit for multi-metric measurement

62 VISION-BASED BRIDGE DEFLECTION MEASUREMENT USING UAVS Raw Video Camera Calibration Intrinsic Matrix Video Frame Natural Feature Detection Foreground Features Structure Motion Tracking Structure Rel. Motion Background Features Egomotion Estimation Camera Motion Egomotion Compensation Abs. Disp. of Structure

63 VALIDATION TEST Video Result for moving 6-Story Building Structure Motion Tracking

64 USACE SMART Gate The problem: Most locks are nearing/exceeding end of design life $6 Billion (46%) projected shortfall in maintenance funding through 2020 $2.8 Million/day in economic loss due to unscheduled lock closure The solution: Structural Health Monitoring program SMART Gate Project goals: Characterize and evaluate causes of miter gate damage Determine gate features that are sensitive to damage Determine gate instrumentation needed for damage detection Develop damage detection algorithm for one of the damage scenarios

65 Cyberinfrastructure for Smart Structures Bridges Tunnels SHM Dams Structure Monitoring Hurricane & Earthquake Monitoring Traffic Monitoring Emergency Response Cloud Services Deflection Monitoring Leakage Detection Fire Detection Traffic Monitoring Emergency Response SHM The Other Infrastructure SHM Pressure Monitoring Leakage Detection Deflection Monitoring Emergency Response Structure Monitoring Hurricane & Earthquake Monitoring Load Rating Emergency Response SHM SHM applications for structural monitoring Repository of structure data and monitoring algorithms Repository of models

66 Asia-Pacific-Europe Summer School on Smart Structure Technology Goals of APSS Program To enhance students' understanding of the cross-disciplinary technological developments on the emerging subjects of smart structure technologies To develop the cross-cultural human-network and understanding for the future cooperation in their professional career development. 3 weeks program for 6 years among US/Korea//China/Japan/India supported by NSF, NRF, JSPS, & NSFC. Coordinators Korea : C-B. Yun (KAIST) US : B.F. Spencer (UIUC) Japan : Y. Fujino (U. of Tokyo) China : L. Sun (Tongji U.) Europe: K. Soga (Cambridge U.)

67 APSS 16 Cambridge, 20 June 8 July

68 Concluding Remarks

69 Vision of the Future relying on and leveraging real-time access to living databases, sensors, diagnostic tools, and other advanced technologies to ensure informed decisions are made

70 Vision of the Future Wireless Smart Sensors will act as the fundamental building block to realize this vision of the future Low costs allow for dense deployment as needed Modularity provides inherent flexibility for use in both permanent and temporary applications

71 Conclusions Recent advances and field validation of a state-of-the-art WSSN framework developed at the University of Illinois at Urbana-Champaign have been discussed The open source ISHMP Services Toolsuite a wide variety of services and fault-tolerant features Tremendous potential and a level of maturity of WSSN for SHM has been demonstrated through full-scale deployment on the Jindo Bridge in Korea and the CN Bridge in Illinois Education of the next generation of engineers in smart structures technology is of high importance

72 Thank You!