A Wireless Sensor Network For Monitoring The Structural .
A Wireless Sensor Network for Monitoring theStructural Health of a Football StadiumDeepa Phanish , Paul Garver, Ghaith Matalkah, Tal Landes, Fu Shen, Jesse Dumond, Randal Abler,Dapeng Zhu, Xinjun Dong, Yang Wang, Edward J. Coyle†College of Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332, USA Email: deepa.phanish@gatech.edu, † Email: ejc@gatech.eduAbstract—We discuss the design, development, and deploymentof an inexpensive, power-efficient, clustered, and scalable wirelesssensor network (WSN) testbed. The testbed operates in a harshenvironment in which neither GPS nor Internet connectivityare available. We use this testbed to collect real-time dataduring football games and other major events at Bobby Doddstadium at Georgia Tech. The sensing devices in the testbedare synchronized without GPS or beacons, yet achieve sufficientaccuracy to support modal analysis and detect if the stands areexperiencing torsion. We have also developed a cognitive radiobackhaul link to establish communication between the WSN inthe stadium and a server in our lab. We present in detail thearchitecture, hardware components, and embedded software ofthe structural health monitoring platform. We also provide datacollected during recent football games to verify the accuracy ofthe new synchronization algorithm and demonstrate that crowdbehavior, such as rhythmic stomping, can be detected during agame.I. I NTRODUCTIONWe have developed a Wireless Sensor Network (WSN)testbed as part of the eStadium project of the VerticallyIntegrated Projects (VIP) Program [1] at Georgia Tech. Thegoals of the project include enhancing the game-day experience and safety of football fans. This is accomplished byserving innovative infotainment and venue-related informationto their mobile devices. Driven by these goals, we havedeveloped a low-power WSN and deployed it in Bobby Doddstadium. It facilitates unique applications that support crowdtailored in-stadium content, interaction among fans, crowdsafety and security, etc. Potential applications include measuring the popularity of a play by the level of cheering andbooing that follows it, estimating waiting times for concessionsand restrooms, detecting bio-chemical hazards, and especiallyStructural Health Monitoring (SHM) of the stadium.SHM systems have been widely explored for measuringthe response of large-scale civil structures. Various types ofsensors, such as accelerometers, strain gauges, displacementand velocity transducers can be used for monitoring structuralbehavior. In order to overcome the high costs associatedwith cable installation, wireless monitoring systems havebeen developed. To date, a number of prototypes have beenproposed and tested in the field. For example, Lynch et al.validated the performance of a prototype wireless sensor onthe Alamosa Canyon Bridge in southern New Mexico [2].The wireless SHM platform designed by Wang [3] has beenvalidated on a number of bridge structures. In general, these978-1-5090-0366-2/15/ 31.00 2015 IEEEefforts cannot detect twisting in a structure because of thelack of accurate synchronization of measurements. They havea comparatively reduced lifetime due to higher standby powerconsumption. Also, they do not support operation of thenetwork from a remote server. Some recent work has also beencarried out to determine the structural response of stadiumsto crowd behavior [4] [5]. These latter efforts are typicallybased on measurements from one position in the stands orunsynchronized measurements at different positions in thestands over a short period of time. The primary purposesof such endeavors is to determine if the dynamic behaviorof the stands exceeds thresholds at which people becomeuncomfortable or to determine the spectral content of thevibrations at individual positions in the stands.In this paper, we are interested in studying the structuralbehavior of the North stands of Bobby Dodd Stadium atGeorgia Tech. The stands are cantilevered over a plaza, as seenin Fig. 1. The physical response of the stands is particularlyinteresting when fans jump to their feet during an excitingplay, bounce with music during cheers, and when they allstart moving at half-time. The stands’ physical responses ofinterest is therefore correlated with major events in the game.These events may excite resonant modes of the stands in the.5 to 5 Hz range that result in twisting of the concrete deck.Detection of these potentially damaging modes requires highlysynchronized measurements of acceleration at many points inthe stands. These measurements are collected at a 100 Hzsampling rate over a wireless network and forwarded to ourserver for analysis. The vibration sensing SHM applicationdiscussed here demonstrates the design and functionality of theWSN tesbed. The pure embedded systems approach devoid ofabstraction layers in our design allows for a better definitionof the applications compared to the existing embedded OSplatforms [6] [7] [8].Our main results: (1) A WSN designed to operate over along period of time; i.e, for one or two football seasons. It iswell-suited for rare, high user-density events since the networkcan be remotely operated. (2) A reliable GPS- and beaconfree synchronization algorithm that yields synchronization towithin 300 µsec. (3) Wireless backhaul of the data from thestadium via a TV whitespace link. (4) Deployment of the firstcluster of the testbed in the stadium. These results are achievedwhile maintaining the underlying simplicity of the low-costinfrastructure.
Fig. 1. North stands of Bobby Dodd Stadium at Georgia Tech. Location of the sensor network currently deployed is marked in red on top-right.MnSimpliciTI/IEEE 802.15.4S0SkM0-MnS0-SkAccSCBCMCHUSRPSensor MotesSensorsAccelerometerSignal Conditioning BoardCoordinator MoteCluster HeadUniversal Software ServerCMAccSerial/USBCHUSRPAccess PointUSRPGT Network500 MHz TV-White SpaceBackhaul LinkFig. 2. The WSN architecture. Note the communication path from the sensors of a single-hop cluster to the server.A. TerminologyEnd Device (ED) Sensor network node consisting of a sensor,processing unit and transceiver. Alternatively called a sensornode or sensor mote.Coordinator Mote (CM) Master node governing a cluster andresponsible for collecting data from the nodes in its cluster.Cluster-Head (CH) Computational unit to aggregate andprocess the information gathered by the coordinator.Access Point (AP) CM and CH coupled as one unit.Backhaul Communication link between the access point andthe remote server.II. A RCHITECTUREThe WSN is designed to have a clustered, hierarchicalarchitecture. The general layout of the single-hop two-levelWSN that we have developed and deployed is shown in Fig.2. This can be easily extended to include additional clustersand levels, thus making it scalable. Each cluster of the WSNconsists of eight to ten battery-powered end devices that arewirelessly connected to one access point. The access pointsare connected via the wireless backhaul to the remote serverwhich acts as the sink.The sensor nodes gather data from local digital and analogsensors for various applications. The sensed data is packetizedand sent wirelessly using the SimpliciTI [9] protocol to thecoordinator. The coordinator node then appends a customheader to the packets and forwards them over UniversalAsynchronous Receiver/Transmitter (UART) and USB to thecluster-head (CH). The CH aggregates data within the clusterwhen applicable and generates appropriate queries to the sink.The CH also controls the behavior of the cluster by issuingcommand packets downstream. Command packets transmittedby the CH are either initiated by the CH or forwarded onbehalf of the eStadium SensorNets server. Example scenarioswhere command packets are applicable include: triggering datacollection, setting sensor reporting time, and specifying sleepduration. The coordinator and cluster-head together form theaccess point, which is the gateway to the backhaul network.The CH communicates with the remote server through aTCP/IP connection. A cognitive-radio-enabled TV whitespacebi-directional link is used wherever a wired connection is notavailable. All of the data collected by the sensor network isstored in a MYSQL database on the server for analysis andfor end-user applications. The server also acts as the level-2cluster-head, thus issuing commands to control functionalityof the level-1 cluster-heads. The hardware and software components used to build this network are listed in Table I.A. Clustered Sensor Network1) End Device: Each sensor mote is a power-efficientsystem consisting of the MSP-EXP430F5438 microcontrollerand a CC2520 (IEEE 802.15.4) radio from Texas Instruments(TI). In addition to several onboard sensors, it has I/O portextensions that allow for interfacing with external sensors. Inorder to achieve high-resolution acceleration measurement inthe vibration sensing project, a low-cost integrated accelerom-
TABLE IN ETWORK C OMPONENTSNetwork unitAccelerometerSensor moteCluster CommunicationCo-ordinator moteCluster-headSDRBackhaul linkeStadium SensorNets ServerZ7.6A14Y7.5A13X7.4A12Acc I/OSCB TBuffer 0Timer BHardwareLT microelectronics LIS344ALHTI MSP430F5438A Experimenter boardTI CC2520EMTI MSP430F5438A Experimenter boardAdvantech PCM-9363D 3.5” Single Board ComputerUSRP B100 with an Intel NUC PC500 MHz Yagi AntennasDell 2950, 2 Xeon quad core processorsXINTDMABuffer nX16 x 2SMclkINTCPUYY16 x 2TT1x4packetBuffer 31Fig. 3. Automated vibration sampling and parallel data processing on anED. The timer operates in up mode with reset/set output, ADC in repeatsequence-of-channels mode, DMA channels 0 and 1 in single transfer mode,and DMA channel 2 in block transfer mode. The sampling in the ADC iscontrolled by a pulse-width modulated signal generated by the timer. Oncea new sample is ready in the internal memory register, the ADC modulegenerates an interrupt signal. The new sample needs to be read immediatelyand moved to the data buffer to avoid being overwritten by the next sample.This task must be performed without interrupting the ongoing processingof the samples already residing in the buffer. To accomplish this, the ADCinterrupt is set to trigger the Direct Memory Access (DMA) module. TheCPU is notified by the DMA module only after the data set for one packet isready in the data buffer. Multiple data buffers are managed in a round-robinfashion, so that the DMA continues to process subsequent samples withoutwaiting for the CPU to service the current interrupt subroutine. Note thesample-timestamping performed by the DMA for synchronization purposes.eter package has been developed [10]. The package consists ofa MEMS accelerometer (LIS344ALH by STMicroelectronics)and a signal conditioner capable of providing triaxial measurements.The range of the accelerometer can be selected as 2 g or 6 g. The noise density of the measurementis 25 µg/ Hz along the x and y axes, and 50 µg/ Hz along thez axis. The cut-off frequency and gain can be programmedon the fly through an I 2 C interface. The power consumptionof the integrated accelerometer board is about 12 mA underworking conditions and 1 µA while asleep.Vibration data is acquired during games, concerts, and majorweather events for structural health monitoring. For such applications, time-domain vibration data is important. Therefore,SoftwareSimliciTI-CCS-1.1.1.exe(Rev. A)SimpliciTI RF protocolSimliciTI-CCS-1.1.1.exe(Rev. A)ch-embedded OS and MYSQL clientGNURadio and gr-macRHEL 6.6 OS with MYSQL Databasesensed vibration data is sent as raw data to the CH and then tothe server. The vibration signals of interest have frequenciesbetween 0 and 25 Hz. Hence, a sampling rate of 100 Hzis adequate. Since the vibration measurements of structuresas large as a stadium stands are usually in the order ofmm/sec2 , small fluctuations in amplitude are of significanceto the measurements. To capture such small variations duringsampling, a precision of 12 bits/sample is used. The interactionbetween different modules for an automated processing of thedata on the ED is shown in Fig. 3. The processed data ispacketized and transmitted to the coordinator, as explained inthe next section. The ED enters a low power mode sleep stateand links to the CM at set intervals. The CM controls thesleep cycle of each ED depending on the game time knownat the server. The current consumption by the ED is 37 mAin the active state and 20 µA in the sleep state. Hence, thetotal estimated energy used to collect data during games ofone football season is approximately 2 Ahr.2) Communication Protocol: Each sensor node has a radiomodule that consists of a CC2520EM daughter board and anantenna. The radio module is interfaced with the MSP430micro-controller using a Serial Peripheral Interface (SPI) forbi-directional communication of data and radio commands.The CC2520 is a 2.4 GHz transceiver that is compliantwith IEEE 802.15.4, which is the standard protocol intendedfor low-power, low-rate Personal Area Networks (PAN). TheIEEE 802.15.4 protocol supports only single-hop networksand comprises only two layers: a physical and a MediumAccess Control (MAC) layer. Most IEEE 802.15.4 PANs areconfigured in a star topology where the central node acts as acoordinator for the rest nodes (i.e., similar to configurationof the network in Fig. 2). SimpliciTI [9] builds on theIEEE 802.15.4/Zigbee protocol and defines two more layers,the network and application layers. This allows for moreadvanced features to be implemented in the network, such asmulti-hop communication and advanced network management.SimplicTI code runs on the main microcontroller while IEEE802.15.4 lower layers are implemented in the radio module.The SimpliciTI stack includes an intermediate sub-layer calledthe Minimal Radio Frequency Interface that conceals thehardware differences.The vibration data is inserted into the application payloadof SimpliciTI packets and sent to the CM at 250 kbps onchannel 25 or 26 of 802.15.4. The wireless channel uses the
FieldRSSISource Addressbytes14PartitionSequence numberCM Received TSLengthApplication ID1411First sample TSPrevious sent TS44CM headerED headerData64ED PayloadFig. 4. Structure of the packet passed to the CH from the CM. The payload has 16 vibration samples from the two axes of measurement.CSMA/CA random access with a uniform random backoffscheme. Packing the redundant zero bits in the data resultsin a further 25% reduction of the wireless traffic and theassociated power usage. The ED also appends a header withboth an ID to identify the application and timestamps requiredfor synchronization. On reception of the packet from the ED,the application on the CM extracts the required informationfrom the headers of lower layers. This is appended as theCM header to the payload and passed to the CH over theserial connection at a baud rate of 230400. The structure ofthe packet delivered to the CH is shown in Fig. 4.3) Access Point: The coordinator mote has the same hardware configuration as that of the sensor mote. It is alsoan MSP-EXP430F5438 experimenter board equipped with aCC2520 radio module. The coordinator constantly monitorsthe SimpliciTI channels for packets and passes them to theattached CH in application specific formats.The Clusterhead (CH) is comprised of an Advantech PCM6363D 3.5” single board computer (SBC) equipped with anIntel Atom D2525 Dual Core 1.8 GHz processor, Gigabitethernet, and up to 4 GB of RAM. The role of the CH is togather the sensor information from the Coordinator over USB,parse it, and update the appropriate MYSQL database via thebackhaul network. It is designed to be lightweight, reliable,and efficient. Therefore, a custom minimal but highly efficientLinux distribution, ch-embedded, is developed for the CH. Theentire distribution is 30 MB. It consists of the Linux kerneland few selected programs required for operation, as shown inTable II. The kernel was extracted from Ubuntu 11.04. There isno persistent file system, only an initial ramdisk (initrd) imageis used. The disadvantage of this read-only system is the lackof local writable storage. However, boot time is reduced andthe system is more robust against sudden power cycles.The software architecture of the multi-threaded user-spaceprogram that reads data sent over the USB/Serial connectionand performs action based on the application type is shownin Fig. 5. Frame synchronization is performed on the serialdata stream in the main thread by using invariant headerbits. Once this is achieved, the payload is extracted and theapplication id field is read. Each id is mapped to a thread via aconfiguration file. Multiple ids may map to a single thread. Themain thread passes the payload data to the processing thread,which performs application-specific processing. For example,the audio and vibration data is uploaded to a MYSQL server.Such an architecture provides abstraction, extensibility, androbustness against failures.TABLE IIC LUSTERHEAD S OFTWARE C OMPONENTSSoftwareLinux KernelBusyboxDropbearntpmysqlVersion2.6.381.21.9 (Stable)2013.594.2.6p56.0.2-linux-x64-64Main ThreadPurposeOperating SystemBasic Linux UtilitiesSSH ServerTiming SynchronizationMYSQL Client LibraryDefault ThreadRead SerialDiscard DataFrame SyncAudioThreadExtractPayloadVibration ThreadType ?MYSQL updateMYSQL updateFig. 5. Serial Monitor Platform to hand the applications off to right threads.B. Cognitive Radio Backhaul1) Whitespace Software-Defined Radio: If a wired networkis unavailable, a software-defined radio (SDR) can bridge theCHs deployed in the football stadium and the main serverinfrastructure. Each node consists of an Intel Next Unit ofComputing (NUC) Ivy Bridge general purpose computer andan Ettus Research B100 USRP RF digitizer with WBX RFDaughterboard. These SDRs operate in the TV whitespacespectrum (470-690 MHz). The particular operating channelis dictated by the FCC allocation database.Each NUC uses Ubuntu 14.04 as an operating system andGNURadio [11] for the software radio processing platform.John Malsbury’s gr-mac [12] module for GNURadio is usedfor the PHY/MAC layer implementation but with a modification to use the tap/tun interface. With this change, the bridgebetween networks is transparent and can be used by multipleclients on each side without any issue. The default modulationscheme in the gr-mac module is Gaussian Minimum Shift Key(GMSK) with a sample rate of 1 Megasample per second andfour samples per symbol. The normalized filter bandwidth,BT 0.35 is set as the default.C. Network load vs CapacityThe bit rates of data generated from the EDs and CHs onthe sensor network with m clusters and p nodes per cluster aregiven in Table III. This includes the sensor generated data andthe overhead due to the SimpliciTI, ED, CM and CH headers.For the target deployment of 50 nodes (m 5, p 10),the load on the SimpliciTI channel is 33.4 kbps in a cluster
0.015Touchdownby MiamiHalf-timebreak0.01Acceleration in g"MakesomeNoise"Advertisement on thebig screen0.0050-0.005Introducing 1990national championteam half time show-0.01-0.0150"Make someNoise"1000IntroducingGT fans20003000Touchdown by GTTouchdown by Miami4000500060007000Time in secondsFig. 6. Structural vibrations indicating events during a football game.and the load on the backhaul network is 175 kbps. Raw datadelivered to the server database in a four hour game is 49.2Mb per node, for a total of 37 MB per node in a 6-home-gameseason.TABLE IIIDATA RATE AND N ETWORK CAPACITYNetworkSimpliciTI channelWhitespace backhaulSourceEDCHData Rate (kbps)3.34 p3.5 m pCapacity (kbps)250250III. D EPLOYMENT AND G AME DATAA sample of the data we have collected is shown in Fig.6. It is evident from the plot that crowd behavior and othermajor events have an influence on the structural excitation. Acrowd stomping in unison could excite a resonance i
while maintaining the underlying simplicity of the low-cost . Co-ordinator mote TI MSP430F5438A Experimenter board SimliciTI-CCS-1.1.1.exe(Rev. A) Cluster-head Advantech PCM-9363D 3.5” Single Board Computer ch-embedded OS and MYSQL client SDR USRP B100 with an Intel NUC PC
Bruksanvisning för bilstereo . Bruksanvisning for bilstereo . Instrukcja obsługi samochodowego odtwarzacza stereo . Operating Instructions for Car Stereo . 610-104 . SV . Bruksanvisning i original
ZigBee, Z-Wave, Wi -SUN : . temperature sensor, humidity sensor, rain sensor, water level sensor, bath water level sensor, bath heating status sensor, water leak sensor, water overflow sensor, fire sensor, cigarette smoke sensor, CO2 sensor, gas s
a low-range wireless network which covers an area of only a few dozen metres wireless sensor network WSN self-organizing, multi-hop networks of wireless sensor nodes used to monitor and control physical phenomena wireless wide area network WWAN wireless network that provides communication ser
10 tips och tricks för att lyckas med ert sap-projekt 20 SAPSANYTT 2/2015 De flesta projektledare känner säkert till Cobb’s paradox. Martin Cobb verkade som CIO för sekretariatet för Treasury Board of Canada 1995 då han ställde frågan
service i Norge och Finland drivs inom ramen för ett enskilt företag (NRK. 1 och Yleisradio), fin ns det i Sverige tre: Ett för tv (Sveriges Television , SVT ), ett för radio (Sveriges Radio , SR ) och ett för utbildnings program (Sveriges Utbildningsradio, UR, vilket till följd av sin begränsade storlek inte återfinns bland de 25 största
Hotell För hotell anges de tre klasserna A/B, C och D. Det betyder att den "normala" standarden C är acceptabel men att motiven för en högre standard är starka. Ljudklass C motsvarar de tidigare normkraven för hotell, ljudklass A/B motsvarar kraven för moderna hotell med hög standard och ljudklass D kan användas vid
LÄS NOGGRANT FÖLJANDE VILLKOR FÖR APPLE DEVELOPER PROGRAM LICENCE . Apple Developer Program License Agreement Syfte Du vill använda Apple-mjukvara (enligt definitionen nedan) för att utveckla en eller flera Applikationer (enligt definitionen nedan) för Apple-märkta produkter. . Applikationer som utvecklas för iOS-produkter, Apple .
Lastly, a wireless sensor network must be adaptable once it is set up. Due to the fact that most wireless sensor networks are completely autonomous, a WSN may have to reconfigure itself to keep the network optimized in case of a loss of node(s), a change in the environment, or even a change of the application for a wireless sensor network.
