- Email: [email protected]
Real-time monitoring system for local storage and data transmission by remote control
Advances in Engineering Software 112 (2017) 46–53
Contents lists available at ScienceDirect
Advances in Engineering Software journal homepage: www.elsevier.com/locate/advengsoft
Research paper
Real-time monitoring system for local storage and data transmission by remote control Sara Casciati a, Michele Vece b,∗ a b
DICAr, School of Architecture, University of Catania at Siracusa, P.za Federico di Svevia, Siracusa 96100, Italy DICAr, University of Pavia, Via Ferrata 3, Pavia 27100, Italy
a r t i c l e
i n f o
Article history: Received 5 October 2016 Revised 22 January 2017 Accepted 15 June 2017
Keywords: Displacement GNSS Real-time Structural health monitoring Wireless connection
a b s t r a c t The real time estimation of the displacements of civil structures is quite sensitive to the availability of wired links between the sensors and the remote control room. Many kinds of wireless displacement sensors or indirect measurements of them have been proposed. However, most of them suffer of large time delays and accuracy issues. In this paper, the authors adopt a Kalman-filter-based data fusion to make a precise measurement of the displacements in for civil structures and infrastructures. The required accuracy can be reached exploiting the real-time satellite corrections provided by a single reference station and combining them with the acceleration signals coming from three axial accelerometers. A wireless communication transfers the information coming from the coupling of GNSS receivers and three axial accelerometers. The proposed system is validated by control mechanism and laboratory tests. The ultimate goal is a reliable scheme for a real-time structural health monitoring managed in remote control. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction Real-time structural health monitoring (SHM) is regarded as a hot spot in the area of civil engineering. It addresses the acquisition of structural responses and environmental factors in order to assess the structural health in view of maintenance actions. The first step is to identify the types of physical quantities to be monitored: key issues are generally accuracy, user-friendliness, and availability of the measurements. Acceleration and displacement are the most widely measured responses in SHM to determine the dynamic characteristics, the stability and the soundness of large structures. Two kinds of techniques for measuring the displacements are currently available: (i) direct measurements, and (ii) indirect estimation approaches. The direct methods include linear variable differential transformers (LVDT), LASER-based transducers [1–2], and vision based system [3–6]. Their main drawbacks are the high costs and the synchronization issues when multiple points are required. Furthermore, the data acquisition systems have to be placed between a measurement point and a stationary reference, with challenging issues resulting from the need of wired connections. The indirect estimation approach converts other structural responses (i.e., acceleration and strain) into displacements on the basis of their physical relationships. However, it suffers from low ∗
Corresponding author. E-mail address: [email protected] (M. Vece).
http://dx.doi.org/10.1016/j.advengsoft.2017.06.010 0965-9978/© 2017 Elsevier Ltd. All rights reserved.
accuracy due to numerical errors and imperfect physical relationships [7]. Accuracy enhancements are achieved by exploiting multiple responses, via a data fusion model, to reduce the drawbacks of each single acquisition [8–10]. Indeed, the conversion of the accelerations can be improved by the simultaneous collection of displacements measured by a high precision GNSS (Global Navigation Satellite System) device. This method has been actively studied in the GPS (Global Positioning System) navigation field, but it needs to make a post-data fusion between accelerations and displacements as well [11–16]. The data acquisition is also a key issue for a successfully SHM (Structural Health Monitoring). The need of wireless connections has emerged in recent years, since the wired monitoring systems suffer for installation problems, its invasive effect, the vulnerability to mechanical damage, and the high costs for the maintenance of the wires as well. Many wireless sensor schemes have been recently proposed, but some of them suffer of large time delays and accuracy issues [16–22]. In this paper, a Kalman filter-based data fusion is adopted to make a precise measurement of the displacements induced on civil structures. The needed accuracy can be reached exploiting the realtime satellite corrections provided by a single reference station and making a combination between the resulting data and the accelerations coming from three axial accelerometers [23–25]. After that, a recent wireless communication technology is used to develop an updated solution that achieves the expected perfor-
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
47
mance without additional costs and significant changes with respect to the existing analog cables. For this purpose, the ZigBee technology and the IEEE802.15.4 compliant transceiver CC2530 resulted to be the most suitable components for applications in wireless SHM systems. Once the message has been locally acquired, the information to be sent to a remote center. In this way, it remote center operators can analyze the data and detect the effectiveness of the structure under study. The communication is unable to use internet, when an environmental crisis interferes with the existing terrestrial network. Therefore, a VPN (Virtual Private Network) over satellite is adopted, ensuring the continuity of the monitoring operations. The features of the wireless transceivers and the tasks of a reliable satellite communication are discussed. The hardware and the software architectures are described, and an experimental validation of the connections is performed.
The state-space model can be written using the definition of acceleration:
2. Governing relations
where w ∼ (0, Q), Q = [
The European Union FP7 project named SPARTACUS (Satellite Based Asset Tracking for Supporting Emergency Management in Crisis Operations) funded the software development and the entire experimental campaign that is reported in this paper. A reliable and efficient sensor network needs low-cost and low power to become successful in the SHM field. For this reason, the expenses for local data collection are minimized using two affordable devices assembled in a system able to produce the same performance of more expensive systems.
2.2. Data fusion based on Kalman filter The formulation of the displacement estimation based on the Kalman filter, as provided by the Ellipse, is summarized in this section in order to clarify the sequence of steps, which have to be implemented in the software. The aim of the proposed method is to extract the needed information to reach the best accuracy from two different measurements. The formulation to fuse acceleration and displacement is also adopted in several recent contributions [30–33].
z = xm = 1 0
x x˙
x 0 0 + x¨ + x˙ 1 m 1
ηa (1)
+ ηd
where y are the state variables, grouping x¨m and xm that are the measured acceleration and displacement, ηa and ηd are the associated measurement noises of acceleration and displacement. It is assumed that ηa and ηd are white noise Gaussian processes with covariance q and r, respectively. Eq. (1) can be written in matrix form as:
y˙ = A y + B u + w
(2)
z=H y+υ 0 0
0 ], υ ∼ (0, R ), and R = r. Moreover, let q
0 the initial state is y (0 ) ∼ ([ ], p0 ), and all the uncorrelatedness as0 sumptions hold. Eq. (2) rigorously represents the relationships between the states, the measurements and the associated measurement noises. Since all measured signals are discrete, a state-space model in the discrete time domain is desired. The discrete version of Eq. (1) is [31]:
yd ( k + 1 ) =
x1 ( k + 1 ) x2 ( k + 1 )
2.1. Specific hardware The devices responsible for sending the acquired message are the ZigBee units. They consist of a CC2530 transceiver, equipped with a Wi-Fi antenna (dual band 2.4/5 GHz), and ended with a USB connector. Further details are provided in Section 3. The task of running them is assigned to a PCB (Printed Circuit Board), which has the size of a credit card and allows one to collect information as an usual personal computer. The adopted electronic board is named Raspberry Pi and is prized in the range from US$20 to US$35. Secure Digital (SD) cards with different capacities may be used to store the operating system and program memory. In the tests carried out, the authors adopted the device EllipseN, which is the miniature sensor produced by SBG [26]. It includes a three axial accelerometer. This system also embeds an industrial GNSS receiver, and runs an on-board Kalman filter able to make a data fusion of both measures. The limitation is that the model only supports a proprietary protocol. Indeed, the standard NMEA provides only the ultimate positioning (after the data fusion). It is worth noting that several devices that can be detached from the own manufacture protocol are available on the market, and can replace the compact SBG sensor [27–29]. In this way, one can exploit all the features of their each single component. One uses the raw data to check each kind of measures acquired by the sensors, or to run other estimation algorithms.
x˙ x˙ 0 1 y˙ = 1 = = x˙ 2 0 0 x¨
+
2
dt /2 dt
=
1
dt
0
1
u (k ) +
2
x1 ( k ) x2 ( k )
dt /2 dt
(3)
ηa ( k )
x1 ( k ) z (k ) = 1 0 + ηd ( k ) x2 ( k )
where dt is the sampling time. Eq. (43) written in compact form becomes:
yd ( k + 1 ) = Ad yd ( k ) + Bd u ( k ) + w ( k ) z ( k ) = H yd ( k ) + υ ( k )
(4)
1 dt dt 2 /2 ], and Bd = [ ]. The noise processes have the 0 1 dt covariance matrices defined as follow [31]: where Ad = [
Qd = q
dt 3 /3 dt 2 /2 dt 2 /2
dt
(5)
Rd = r/dt The state-space model in Eq. (4) is then utilized to develop a discrete-time Kalman filter for displacement and velocity estimation. Let the initial values put to be:
yd − ( k = 1 ) =
0 0
(6)
P − ( k = 1 ) = p0 z − ( k = 1 ) = xm ( k = 1 )
One computes the quantities referred to as measurements updates at k by introducing the Kalman gain K(k). Subsequently, the Kalman gain is computed by the Eq. (7):
K (k ) = P − (k )H T H P − (k )H T + Rd
−1
(7)
48
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
Fig. 1. ZigBee network type for the real-time SHM.
The measurement update is given by:
yˆd (k ) = yˆd− (k ) + K (k ) z− (k ) − H yˆd− (k ) P (k ) = [I − K (k )H ]P − (k )
(8)
zˆ(k ) = H yˆd (k ) Such a measurement update is then followed by the time update:
ZigBee network can be implemented by the system-on-chip (SoC) transceiver CC2530 and its corresponding protocol stack offered by Texas Instrument [34]. It integrates a radio frequency (RF) power amplifier CC2591, which allows adapting to short and medium range communication. The hardware used for each device consists of a ZigBee to USB (Virtual RS232) adapter with a CC2530 module, a rubber antenna, an UART to USB bridge, an USB connector and a JTAG interface for downloading and debugging the firmware, as shown in Fig. 2. The CC2530 is a true SoC solution for IEEE 802.15.4 and other ZigBee applications as well. It enables to combine the excellent performance of a leading RF transceiver with an industry-standard enhanced 8051 MCU, with in-system programmable flash memory of 8KB RAM. The firmware code for this kind of device is the 8051 microcontroller core integrated in the CC2530 transceiver, which is compiled using an Integrated Development Environment (IDE) software named IAR Embedded Workbench 8051. Finally, the ZigBee protocol stack (Z-StackTM ) which is needed is Open Source and provided by Texas Instrument.
yˆd− (k + 1 ) = Ad yˆd (k ) + Bd u(k ) P − (k + 1 ) = Ad P (k )Ad T + Qd
(9)
z ( k + 1 ) = xm ( k + 1 )
3.2. Physical architecture
−
where yˆd− (k + 1 ) and yˆd (k ) are the prior and the updated state estimates, respectively, while P − (k + 1) and P(k) are the prior and the updated error covariance matrices. It has to be underlined that u(k ) = x¨m (k ). The procedure allows one to achieve an accurate displacement estimation in all frequency regions. As stated above, the software implementing such a procedure runs on board the inertial sensor produced by SBG. In the “proprietary free” approach pursued in this manuscript, the previous equations have to be arranged in a software to be executed at the local collection node. Its result is passed via satellite to a remote center, thus achieving a lightening of the data transmission. 3. Real-time displacement monitoring system The methodology of the real-time displacement monitoring system consists of four sections: (i) communication framework, (ii) physical architecture, (iii) functional procedure, and (iv) software implementation (see Appendix). It is designed through a set of practical steps for the local storage and the data transmission from the prototype to a remote center via satellite. 3.1. Communication framework This study is targeted to the real-time data collection and to the long-range communication (when the distance is more than 100 m) between the Sensor Nodes and the Router and from the Router to the Coordinator Node. The process from the Sensor to the Coordinator is expected to be operational only in case of emergencies (i.e., blackout cases or malfunction of the current structural health monitoring systems) and, for this reason, can be carried out by the ZigBee technology. ZigBee technology offers a great trade-off in terms of power, range, data rate, and security. It organizes multi-hop network, which imposes that two devices exchange data relying on intermediate forwarding devices. Furthermore, the nature of this system requires that each device performs a specific networking function that determines its role within the network, as shown in Fig. 1. The developed network consists of a coordinator, a router and several sensor nodes, but the type of device does not restrict its application because it is usually determined at compile-time. The
The proposed system architecture consists of a wireless network to store locally the information, and to transmit the measurements of each sensor to the coordinator node. After that, the information may be sent via satellite to a remote centre for monitoring and maintenance purposes (Fig. 3). The wireless network is made of GPS devices able to receive the corrections provided by a base station located near the structure. This technique is named RTK (Real Time Kinematic) and is used to enhance the precision of the positioning systems. It measures the phase of the carrier wave of the signal, and it relies on a single reference station to provide a centimeter-level accuracy. The system is based on the open-access software environment of a credit card-size computer, which is used as data collector and named (Raspberry Pi 2). It is run as Raspbian, a Debian-based computer operating system, and the operations are coded in Python [35]. The core activities are: - store sensor data inside the memory of the data collector in real-time, - create and transmit data frame to the coordinator through ZigBee, - transfer the information with a remote center via satellite, and - process the data with the displacement estimation method. Each measurement point is made by a sensor node (ZigBee) and a GPS receiver coupled with three axial accelerometers for realtime data collection and transmission. This network covers a small area, and the wireless communication represents a good solution for collecting and transmitting the monitoring data. Furthermore, the coordinator node may communicate with the remote center via satellite in case of the terrestrial network failed. The satellite antenna system for the field-testing was also proposed in the ESA 3InSat project [36]. The system is from Hughes and is identified by the model 9450 [37]. It is composed of a satellite terminal (User Terminal, UT) and of an auto-pointing antenna. To overcome the constraints of GSM/3G/4G connections, it exploits a VPN on site and connects it over the satellite link to a remote server. The information is stored by the central server into a database, and it is post-processed. In this way, the monitoring of a structure can be updated in real-time and with high precision.
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
49
Fig. 2. Top view (A) and bottom view (B) of the ZigBee device.
Fig. 3. Architecture of the real-time monitoring system.
3.3. Operational principle At idle time, the sensor node remains in sleeping state until the monitoring system designed for the civil infrastructure works properly. When it is cut-off, and the communication can not be executed by the existing networks, the power generated by portable and rechargeable batteries is provided to the prototyped system. To clarify its operational principle, a functional flowchart of the designed prototype is shown in Fig. 4. The main premise is the availability of a device incorporating GPS positioning together with a standard inertial sensor. As soon as the battery power is provided to the system, the sensing device produces, at each time step, NMEA-0183 strings, and comes with its wire ending (USB interface). The bottleneck is exactly the power one has to provide to the data collector. Thus, low consumption surrogates of the computers are welcome and modules of the Raspberry PI class are used. To minimize the power consumption, the conceived architecture creates a byte array (i.e., a sequence of numbers) and send the sentence by a ZigBee modulus to the Coordinator Node. Once the buffer is full, the software stores the message inside the memory of the Coordinator Node, which can be equipped with
higher power availability. From here, one can rely on internet or, to bypass any liability problem, on satellite phones, to transmit the information to the remote center. 4. Test results and discussion The results of two different experimental campaigns are presented in this section. The first part addresses the tests carried out in the laboratory to validate the performance of the wireless data acquisition system. The second sub-section reports on experiments executed in a train maintenance yard. Focus is here on the satellite transmission of positioning data for a railway convoy. 4.1. Wireless data acquisition testing Some laboratory tests are analyzed from the point of view of the wireless link quality. Two aspects are worth being discussed: data packet loss, in term of packet reception rate, and energy consumption. The data loss issue is addressed by a reliable communication method with a character checking mechanism. This type of ap-
50
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
Fig. 5. A three-dimensional graph of the NMEA string transmitted (blue) and received (orange) during the tests.
Fig. 4. Functional flow chart of the proposed system.
proach is adopted to avoid data loss and sentence duplication as well. After acquiring a NMEA string from the GPS sensor, the RasPi_I will send each single character of the GGA message to the RasPi_II within the time window of a second. If the message is checked to be correct by the character control, it is regarded as having been successfully transmitted. Otherwise, the Sensor Node will retransmit the sentence to the Coordinator Node. Since the RasPi_I continues to acquire a new string every second from the GPS sensor, this procedure may cause a delay in the final collection of the transmitted data. To avoid this problem, the “corrupt” strings are sent with a period (0.1 s) less than the sampling of the system (t = 1s). In this way, the next operation can be performed without changing the software behavior. Each sentence is also marked by a sequence number, which reports the date and the acquisition time. After the development of the proposed data acquisition system, some tests were carried out to estimate the extent of the data loss. During the tests, some parameters was changed such as the time length when the distance between the two devices was 10 m (T01, T02, and T03), 50 m (T04, T05, and T06), and 100 m (T07, T08, T09, and T10). The results are summarized in Table 1. The PRR (Packet Reception Rate) is defined as the number of successfully received data packets divided by the number of actually transmitted data packets. In Table 1, its value is expressed in percentage. The three-dimensional graph of Fig. 5 compares the percentage of NMEA strings transmitted (blue) and those received (orange) by the Coordinator Node via ZigBee.
It is worth noting that the antenna setup greatly affects the radio communication. The PRR of the wireless data acquisition system is almost 100% because the antenna of the Sensor Node was elevated to the same height of the Coordinator, as listed in Table 1. The wireless data acquisition system adopts omni-directional antennas with a high gain of 5.4dBi, but the communication range can be extended. However, a higher gain will lead to a narrower beam width and a higher directivity, since the radiation power is concentrated on the main lobe. Therefore, all the device antennas have to be set up in such as a way that their main lobes are the same plane in order to minimize the path loss of the radio signal channel. The RasPi is powered at 4.94 V and the power consumption, when receiver and wireless device are powered on, is almost 400 mA. Therefore, an external battery of 10 Ah (ROMOSS Sailing 5), which delivers 5 V and 20 0 0 mA of current output, is used to provide an autonomy of 24 h [38].
4.2. Satellite transmission at a specialty test-bed The system assessment, in terms of robustness and reliability, of the communication links (wireless and satellite), was the purpose of the tests carried out, April 2016, at the railway site of Barrow Hill. This locomotive maintenance yard is located near Chesterfield in the United Kingdom [39], and the case study was designed with the collaboration of the Centre for Railway Research (NewRail). The experimental investigations were carried out in an area of 8.5 ha with continuous track of 1200 m. The site is divided in three areas according to the maximum speed allowed for each path: (i) Yard Lines at 2.2 m/s, (ii) Network Rail Access at 4.4 m/s, and (iii) Branch Line at 8.9 m/s, as shown in Fig. 6. The convoy is made by the locomotive, three wagons and one container. Each wagon is equipped with a SPARTACUS Tracking Unit (STrackU), consisting of an inertial sensor (Ellipse-N), a wireless device (ZigBee), and a data collector (RasPi). The container is carried on board the last wagon. The STrackUs power is supplied by portable batteries. They are set into a protective box and fixed on the window of the vehicles with a double-sided tape, as shown in Fig. 7(b).
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
51
Table 1 The PRR values for the laboratory tests. Id
UTC start (hh:mm:ss)
UTC end (hh:mm:ss)
Length (mm:ss)
NMEA strings (n)
PPR (%)
T01 T02 T03 T04 T05 T06 T07 T08 T09 T10
12:01:09 12:27:44 12:35:05 12:43:57 12:47:00 12:55:53 13:05:23 13:15:15 13:20:17 13:29:24
12:09:27 12:34:14 12:43:22 12:46:35 12:55:17 13:04:10 13:13:40 13:19:54 13:28:34 13:33:48
08:16 06:58 08:17 02:38 08:17 08:17 08:17 04:39 08:17 04:24
499 392 499 159 499 499 499 281 499 266
99,80 99,74 99,80 99,37 99,80 99,80 99,80 99,64 99,80 99,62
Fig. 6. Layout and railway convoy of the Barrow Hill site.
Fig. 7. Systems on board the locomotive (a) and wagons (b).
The desired information is recorded in NMEA format [40]. The sentences are acquired by the developed software of each STrackU, and sent via wireless to the SComU (SPARTACUS Communication Unit), which is placed inside the locomotive (Fig. 7(a)). The system on board the locomotive (SColU) is conceived: (i) to collect the positioning data of each STrackU, and (ii) to transmit via satellite the information to a remote center. For this set of tests, the position accuracy of each vehicle is the aim for which
the transmission system was developed. The results for this case study are reported by estimates of the PPR of the two types of communication for a freight train in the railway site. First, the ZigBee behavior from a wagon or a container to the head train, and then the transmission, based on a VPN over satellite, of the NMEA messages from the locomotive to a remote center. The results are summarized in Table 2.
52
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53 Table 2 PRR values achieved during the field tests. Id
UTC start (hh:mm:ss)
UTC end (hh:mm:ss)
Length (mm:ss)
NMEA strings (n)
PPR (%)
L1 W1 W2 W3 C1
09:15:01 09:37:49 09:37:47 09:38:01 09:37:48
09:51:05 10:37:18 10:37:16 10:37:21 10:37:26
36:04 59:29 59:29 59:20 59:38
2732 372 373 364 366
100,0 99,46 99,73 98,90 99,45
The first test (L1) is addressed to check the data transmission via satellite of the SComU, while the results labelled W1, W2, W3 and C1 are referred to the wireless communication of the message from wagons/container to the locomotive. In particular, the above listed results were elaborated from the data stored by the software in the last day of measurements. The outcome of this analysis allows one to conclude that the system behavior is fully performing also when the message has to be forwarded from the data collector (RasPi) to the SComU, enabling the satellite broadcast. 5. Conclusions An inertial sensor, which also includes a GPS receiver (Ellipse), is coupled with a data collector (RasPi) and wireless technologies (ZigBee) to develop a low-cost data acquisition system in the SHM area. For this purpose, a Kalman filter-based approach is exploited in view of an accurate estimation of the displacement induced on civil structures by external actions. It implements a data fusion of the measured responses. The technology also uses the ability to receive corrections from a reference station, providing a centimeterlevel accuracy. The ultimate goal is to ensure the communications even if standard telecommunication tools fail due to catastrophic values of the external excitation intensity. Details on the hardware components and their software implementation are provided. A reliable wireless method of communication between nodes is adopted to avoid data loss issues. It is based on a suitable charcter checking mechanism. At last, the test results of an experimental campaign are discussed to demonstrate that the entire system features a good balance between the three considered aspects: power consumption, communication range, and link quality.
required to manage the counters, (iii) serial, which drives the serial communication, and (iv) time, used to access the system clock. Some constants (baud-rate and buffer-size) are then assigned to fix a file whose name comes with date and time. port = "/dev/ttyUSB0" baud = 38,400 length = 82 control = "^" ser = Serial(port, baud) The next step is referred to the RasPi_II and sees as input the file name. It also starts the device and collects the received information in an array of suitable size: def getFileName(name): def increment_filename(fn): fn, extension = os.path.splitext(name) n =1 yield fn + extension for n in itertools.count(start= 1, step= 1): yield ’%s%d%s’ % (fn, n, extension) for filename in increment_filename(name): if not os.path.exists(filename): return filename The array elements are sent one by one to the ZigBee with a control character ending the transfer: def check(): while 1: if(ser.read(1) == control): read() This tool appends what it reads to a new string through the “save” function, but only if such a string does not exceed the buffer’s length: def read(): string = "" while 1: if(len(string) > length): check() break elif (len(string) == length): save(string) string += ser.read (1)
Acknowledgments The research activity summarized in this paper has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n° [313002]. It was also supported by Athenaeum grants from the University of Pavia.
At last, the software opens a file and appends a string to it. The file is eventually closed.
Appendix According to Fig. 4, the GPS device is set to provide a GGA string per sec in the NMEA-0183 format. It is connected to the first data collector (RasPi_I) by a USB link. A second USB interface hosts the ZigBee Sensor that transfer the data to the ZigBee Coordinator and from here to the second data collector (RasPi_II), where the strings are saved in a text file. All the steps listed above requires a specific software. As RasPi_I is switched on, the following commands are executed: import serial, itertools, os from time import strftime, gmtime, sleep The software imports first the libraries of the Python language. They are: (i) os, necessary to access the File System, (ii) itertools,
def save(string): file = open(filename, "ab") file.write(string) file.close() filename = getFileName() check() References [1] Nassif HH, Gindy M, Davis J. Comparison of laser Doppler vibrometer with contact sensors for monitoring bridge deflection and vibration. NDT&E Int 2005;38(3):213–18. [2] Bai R, Ostachowics W, Radzienski M, Cao M. Vibrational damage detection using fractal surface singularities with noncontact laser measurement. J Vib Control 2014;22(11):2569–81. [3] Balkaya C, Casciati F, Casciati S, Faravelli L, Vece M. Real-time identification of disaster areas by an open-access vision-based tool. Adv Eng Softw 2015;88(1):83–90.
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53 [4] Wahbeh AM, Caffrey JP, Masri SF. A vision-based approach for the direct measurement of displacements in vibrating systems. Smart Mater Struct 2003;12(5):785–94. [5] Lee JJ, Shinozuka M. A vision-based system for remote sensing of bridge displacement. NDT&E Int 2006;39(5):425–31. [6] Wu L, Casciati F. Local positioning systems versus structural monitoring: a review. Struct Control Health Monit 2014;21(9):1209–21. [7] Park KT, Kim SH, Park HS, Lee KW. The determination of bridge displacement using measured acceleration. Eng Struct 2005;27(3):371–8. [8] Sim SH, Spencer BF Jr, Nagayama T. Multimetric sensing for structural damage detection. J Eng Mech 2011;137(1):22–30. [9] Kim J, Kim K, Sohn H. Autonomous dynamic displacement estimation from data fusion of acceleration and intermittent displacement measurements. Mech Syst Signal Process 2014;42(1-2):194–205. [10] Sung SH, Park JW, Nagayama T, Jung HJ. A multi-scale sensing and diagnosis system combining accelerometers and gyroscopes for bridge health monitoring. Smart Mater Struct 2014;23(1):1–14. [11] Roberts GW, Meng X, Dodson AH. Integrating a global positioning system and accelerometers to monitor the deflection of bridges. J Surv Eng 2004;130(2):65–72. [12] Park JW, Sim SH, Jung HJ. Displacement estimation using multimetric data fusion. IEEE/ASME Trans Mechatron 2013;18(6):1675–82. [13] Park JW, Sim SH, Jung HJ. Wireless displacement sensing system for bridges using multi-sensor fusion. Smart Mater Struct 2014;23(4):1–12. [14] Cho S, Sim SH, Park JW, Lee J. Extension of indirect displacement estimation method using acceleration and strain to various types of beam structures. Smart Struct Syst 2014;14(4):699–718. [15] Cho S, Yun CB, Sim SH. Displacement estimation of bridge structures using data fusion of acceleration and strain measurement incorporating finite element model. Smart Struct Syst 2015;15(3-4):645–63. [16] Sim S. Estimation of flexibility matrix of beam structures using multisensor fusion. J Struct Integr Maint 2016;1(2):60–4. [17] Lynch JP, Loh KJ. A summary review of wireless sensors and sensor networks for structural health monitoring. Shock Vib Digest 2006;38(2):91–128. [18] Wang Y, Lynch JP, Law KH. Design of a low-power wireless structural monitoring system for collaborative computational algorithms. In: Proceedings of SPIE 10th annual international symposium on NDE for health monitoring and diagnostics, 5768; 2005. p. 106–17. [19] Nagayama T, Rice JA, Spencer BF Jr. Efficacy of Intel’s Imote2 wireless sensor platform for structural health monitoring applications. In: Proceedings of AsiaPacific workshop on structural health monitoring, Yokohama, Japan; 2006. [20] Ni YQ, Xia Y, Liao WY, Ko JM. Technology innovation in developing the structural health monitoring system for Guangzhou new TV tower. Struct Control Health Monit 2008;16(1):73–98. [21] Casciati S, Faravelli L, Chen ZC. Energy harvesting and power management of wireless sensors for structural control applications in civil engineering. Smart Struct Syst 2012;10(3):299–312.
53
[22] Casciati S, Chen ZC. An active mass damper system for structural control using real-time wireless sensors. Struct Control Health Monit 2012;19(8):758–67. [23] Casciati F, Casciati S. Structural health monitoring by Lyapunov exponents of non-linear time series. Struct Control Health Monit 2006;13(1):132–46. [24] Chen ZC, Casciati S, Faravelli L. In-situ validation of a wireless data acquisition system by monitoring a pedestrian bridge. Adv Struct Eng 2015;18(1):97–106. [25] Casciati S, Chen ZC. A multi-channel wireless connection system for structural health monitoring applications. Struct Control Health Monit 2010;18(5):588–600. [26] Guinamard A. Ellipse AHRS & INS - high performance, miniature inertial sensors user manual. France: SBG Systems; 2014. [27] CXL-LF Series - High Sensitivity Accelerometers (2016). Document Part: 60200 0 02-01, USA. Crossbow Technology Inc. [28] Trimble 147A High Resolution Accelerometer (2016). Trimble Navigation Limited, USA. [29] Trimble R10 GNSS System (2016). Trimble Navigation Limited, USA. [30] Caron F, Duflos E, Pomorski D, Vanheeghe P. GPS/IMU data fusion using multisensor Kalman filtering: introduction of contextual aspects. Inf Fusion 2006;7(2):221–30. [31] Smyth A, Wu M. Multi-rate Kalman filtering for the data fusion of displacement and acceleration response measurements in dynamic system monitoring. Mech Syst Signal Process 2007;21(2):706–23. [32] Zhou Z. GPS navigation data fusion using composed correcting Kalman filtering method. J Comput Inf Syst 2011;7(6):2006–12. [33] Chatzi EN, Fuggini C. Structural identification of a super-tall tower by GPS and accelerometer data fusion using a multi-rate Kalman filter. In: Proceedings of the 3rd international symposium on life-cycle civil engineering, 2012; 2012. p. 144–51. [34] Texas Instruments (2009). CC2530 software examples user’s guide, Texas Instruments Incorporated, Dallas, Texas, USA. [35] Bradbury A, Everard B. Learning Python with Raspberry Pi, Chichester, United Kingdom: John Wiley & Son Ltd; 2014. ISBN 978-1-118-71705-9. [36] 3InSat, Train Integrated Safety Satellite System (2014). ESA Project: Contract 40 0 0105788/12NL/NR. [37] Hughes 9450-C11 BGAN mobile satellite terminal (2010). Hughes Network Systems, LLC. [38] Casciati S, Chen ZC, Faravelli L, Vece M. Synergy of monitoring and security. Smart Struct Syst 2016;17(5):743–51. [39] Casciati F, Casciati S, Faravelli L, Vece M. Implementation of communication, collection and tracking units in crisis operation. In: Proceedings of EWSHM 2016, 8th European workshop on structural health monitoring, Bilbao, Spain; 2016. [40] Trimble. NMEA-0183 messages guide for AgGPS receivers. Trimble Navigation Limited; 2004.
Contents lists available at ScienceDirect
Advances in Engineering Software journal homepage: www.elsevier.com/locate/advengsoft
Research paper
Real-time monitoring system for local storage and data transmission by remote control Sara Casciati a, Michele Vece b,∗ a b
DICAr, School of Architecture, University of Catania at Siracusa, P.za Federico di Svevia, Siracusa 96100, Italy DICAr, University of Pavia, Via Ferrata 3, Pavia 27100, Italy
a r t i c l e
i n f o
Article history: Received 5 October 2016 Revised 22 January 2017 Accepted 15 June 2017
Keywords: Displacement GNSS Real-time Structural health monitoring Wireless connection
a b s t r a c t The real time estimation of the displacements of civil structures is quite sensitive to the availability of wired links between the sensors and the remote control room. Many kinds of wireless displacement sensors or indirect measurements of them have been proposed. However, most of them suffer of large time delays and accuracy issues. In this paper, the authors adopt a Kalman-filter-based data fusion to make a precise measurement of the displacements in for civil structures and infrastructures. The required accuracy can be reached exploiting the real-time satellite corrections provided by a single reference station and combining them with the acceleration signals coming from three axial accelerometers. A wireless communication transfers the information coming from the coupling of GNSS receivers and three axial accelerometers. The proposed system is validated by control mechanism and laboratory tests. The ultimate goal is a reliable scheme for a real-time structural health monitoring managed in remote control. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction Real-time structural health monitoring (SHM) is regarded as a hot spot in the area of civil engineering. It addresses the acquisition of structural responses and environmental factors in order to assess the structural health in view of maintenance actions. The first step is to identify the types of physical quantities to be monitored: key issues are generally accuracy, user-friendliness, and availability of the measurements. Acceleration and displacement are the most widely measured responses in SHM to determine the dynamic characteristics, the stability and the soundness of large structures. Two kinds of techniques for measuring the displacements are currently available: (i) direct measurements, and (ii) indirect estimation approaches. The direct methods include linear variable differential transformers (LVDT), LASER-based transducers [1–2], and vision based system [3–6]. Their main drawbacks are the high costs and the synchronization issues when multiple points are required. Furthermore, the data acquisition systems have to be placed between a measurement point and a stationary reference, with challenging issues resulting from the need of wired connections. The indirect estimation approach converts other structural responses (i.e., acceleration and strain) into displacements on the basis of their physical relationships. However, it suffers from low ∗
Corresponding author. E-mail address: [email protected] (M. Vece).
http://dx.doi.org/10.1016/j.advengsoft.2017.06.010 0965-9978/© 2017 Elsevier Ltd. All rights reserved.
accuracy due to numerical errors and imperfect physical relationships [7]. Accuracy enhancements are achieved by exploiting multiple responses, via a data fusion model, to reduce the drawbacks of each single acquisition [8–10]. Indeed, the conversion of the accelerations can be improved by the simultaneous collection of displacements measured by a high precision GNSS (Global Navigation Satellite System) device. This method has been actively studied in the GPS (Global Positioning System) navigation field, but it needs to make a post-data fusion between accelerations and displacements as well [11–16]. The data acquisition is also a key issue for a successfully SHM (Structural Health Monitoring). The need of wireless connections has emerged in recent years, since the wired monitoring systems suffer for installation problems, its invasive effect, the vulnerability to mechanical damage, and the high costs for the maintenance of the wires as well. Many wireless sensor schemes have been recently proposed, but some of them suffer of large time delays and accuracy issues [16–22]. In this paper, a Kalman filter-based data fusion is adopted to make a precise measurement of the displacements induced on civil structures. The needed accuracy can be reached exploiting the realtime satellite corrections provided by a single reference station and making a combination between the resulting data and the accelerations coming from three axial accelerometers [23–25]. After that, a recent wireless communication technology is used to develop an updated solution that achieves the expected perfor-
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
47
mance without additional costs and significant changes with respect to the existing analog cables. For this purpose, the ZigBee technology and the IEEE802.15.4 compliant transceiver CC2530 resulted to be the most suitable components for applications in wireless SHM systems. Once the message has been locally acquired, the information to be sent to a remote center. In this way, it remote center operators can analyze the data and detect the effectiveness of the structure under study. The communication is unable to use internet, when an environmental crisis interferes with the existing terrestrial network. Therefore, a VPN (Virtual Private Network) over satellite is adopted, ensuring the continuity of the monitoring operations. The features of the wireless transceivers and the tasks of a reliable satellite communication are discussed. The hardware and the software architectures are described, and an experimental validation of the connections is performed.
The state-space model can be written using the definition of acceleration:
2. Governing relations
where w ∼ (0, Q), Q = [
The European Union FP7 project named SPARTACUS (Satellite Based Asset Tracking for Supporting Emergency Management in Crisis Operations) funded the software development and the entire experimental campaign that is reported in this paper. A reliable and efficient sensor network needs low-cost and low power to become successful in the SHM field. For this reason, the expenses for local data collection are minimized using two affordable devices assembled in a system able to produce the same performance of more expensive systems.
2.2. Data fusion based on Kalman filter The formulation of the displacement estimation based on the Kalman filter, as provided by the Ellipse, is summarized in this section in order to clarify the sequence of steps, which have to be implemented in the software. The aim of the proposed method is to extract the needed information to reach the best accuracy from two different measurements. The formulation to fuse acceleration and displacement is also adopted in several recent contributions [30–33].
z = xm = 1 0
x x˙
x 0 0 + x¨ + x˙ 1 m 1
ηa (1)
+ ηd
where y are the state variables, grouping x¨m and xm that are the measured acceleration and displacement, ηa and ηd are the associated measurement noises of acceleration and displacement. It is assumed that ηa and ηd are white noise Gaussian processes with covariance q and r, respectively. Eq. (1) can be written in matrix form as:
y˙ = A y + B u + w
(2)
z=H y+υ 0 0
0 ], υ ∼ (0, R ), and R = r. Moreover, let q
0 the initial state is y (0 ) ∼ ([ ], p0 ), and all the uncorrelatedness as0 sumptions hold. Eq. (2) rigorously represents the relationships between the states, the measurements and the associated measurement noises. Since all measured signals are discrete, a state-space model in the discrete time domain is desired. The discrete version of Eq. (1) is [31]:
yd ( k + 1 ) =
x1 ( k + 1 ) x2 ( k + 1 )
2.1. Specific hardware The devices responsible for sending the acquired message are the ZigBee units. They consist of a CC2530 transceiver, equipped with a Wi-Fi antenna (dual band 2.4/5 GHz), and ended with a USB connector. Further details are provided in Section 3. The task of running them is assigned to a PCB (Printed Circuit Board), which has the size of a credit card and allows one to collect information as an usual personal computer. The adopted electronic board is named Raspberry Pi and is prized in the range from US$20 to US$35. Secure Digital (SD) cards with different capacities may be used to store the operating system and program memory. In the tests carried out, the authors adopted the device EllipseN, which is the miniature sensor produced by SBG [26]. It includes a three axial accelerometer. This system also embeds an industrial GNSS receiver, and runs an on-board Kalman filter able to make a data fusion of both measures. The limitation is that the model only supports a proprietary protocol. Indeed, the standard NMEA provides only the ultimate positioning (after the data fusion). It is worth noting that several devices that can be detached from the own manufacture protocol are available on the market, and can replace the compact SBG sensor [27–29]. In this way, one can exploit all the features of their each single component. One uses the raw data to check each kind of measures acquired by the sensors, or to run other estimation algorithms.
x˙ x˙ 0 1 y˙ = 1 = = x˙ 2 0 0 x¨
+
2
dt /2 dt
=
1
dt
0
1
u (k ) +
2
x1 ( k ) x2 ( k )
dt /2 dt
(3)
ηa ( k )
x1 ( k ) z (k ) = 1 0 + ηd ( k ) x2 ( k )
where dt is the sampling time. Eq. (43) written in compact form becomes:
yd ( k + 1 ) = Ad yd ( k ) + Bd u ( k ) + w ( k ) z ( k ) = H yd ( k ) + υ ( k )
(4)
1 dt dt 2 /2 ], and Bd = [ ]. The noise processes have the 0 1 dt covariance matrices defined as follow [31]: where Ad = [
Qd = q
dt 3 /3 dt 2 /2 dt 2 /2
dt
(5)
Rd = r/dt The state-space model in Eq. (4) is then utilized to develop a discrete-time Kalman filter for displacement and velocity estimation. Let the initial values put to be:
yd − ( k = 1 ) =
0 0
(6)
P − ( k = 1 ) = p0 z − ( k = 1 ) = xm ( k = 1 )
One computes the quantities referred to as measurements updates at k by introducing the Kalman gain K(k). Subsequently, the Kalman gain is computed by the Eq. (7):
K (k ) = P − (k )H T H P − (k )H T + Rd
−1
(7)
48
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
Fig. 1. ZigBee network type for the real-time SHM.
The measurement update is given by:
yˆd (k ) = yˆd− (k ) + K (k ) z− (k ) − H yˆd− (k ) P (k ) = [I − K (k )H ]P − (k )
(8)
zˆ(k ) = H yˆd (k ) Such a measurement update is then followed by the time update:
ZigBee network can be implemented by the system-on-chip (SoC) transceiver CC2530 and its corresponding protocol stack offered by Texas Instrument [34]. It integrates a radio frequency (RF) power amplifier CC2591, which allows adapting to short and medium range communication. The hardware used for each device consists of a ZigBee to USB (Virtual RS232) adapter with a CC2530 module, a rubber antenna, an UART to USB bridge, an USB connector and a JTAG interface for downloading and debugging the firmware, as shown in Fig. 2. The CC2530 is a true SoC solution for IEEE 802.15.4 and other ZigBee applications as well. It enables to combine the excellent performance of a leading RF transceiver with an industry-standard enhanced 8051 MCU, with in-system programmable flash memory of 8KB RAM. The firmware code for this kind of device is the 8051 microcontroller core integrated in the CC2530 transceiver, which is compiled using an Integrated Development Environment (IDE) software named IAR Embedded Workbench 8051. Finally, the ZigBee protocol stack (Z-StackTM ) which is needed is Open Source and provided by Texas Instrument.
yˆd− (k + 1 ) = Ad yˆd (k ) + Bd u(k ) P − (k + 1 ) = Ad P (k )Ad T + Qd
(9)
z ( k + 1 ) = xm ( k + 1 )
3.2. Physical architecture
−
where yˆd− (k + 1 ) and yˆd (k ) are the prior and the updated state estimates, respectively, while P − (k + 1) and P(k) are the prior and the updated error covariance matrices. It has to be underlined that u(k ) = x¨m (k ). The procedure allows one to achieve an accurate displacement estimation in all frequency regions. As stated above, the software implementing such a procedure runs on board the inertial sensor produced by SBG. In the “proprietary free” approach pursued in this manuscript, the previous equations have to be arranged in a software to be executed at the local collection node. Its result is passed via satellite to a remote center, thus achieving a lightening of the data transmission. 3. Real-time displacement monitoring system The methodology of the real-time displacement monitoring system consists of four sections: (i) communication framework, (ii) physical architecture, (iii) functional procedure, and (iv) software implementation (see Appendix). It is designed through a set of practical steps for the local storage and the data transmission from the prototype to a remote center via satellite. 3.1. Communication framework This study is targeted to the real-time data collection and to the long-range communication (when the distance is more than 100 m) between the Sensor Nodes and the Router and from the Router to the Coordinator Node. The process from the Sensor to the Coordinator is expected to be operational only in case of emergencies (i.e., blackout cases or malfunction of the current structural health monitoring systems) and, for this reason, can be carried out by the ZigBee technology. ZigBee technology offers a great trade-off in terms of power, range, data rate, and security. It organizes multi-hop network, which imposes that two devices exchange data relying on intermediate forwarding devices. Furthermore, the nature of this system requires that each device performs a specific networking function that determines its role within the network, as shown in Fig. 1. The developed network consists of a coordinator, a router and several sensor nodes, but the type of device does not restrict its application because it is usually determined at compile-time. The
The proposed system architecture consists of a wireless network to store locally the information, and to transmit the measurements of each sensor to the coordinator node. After that, the information may be sent via satellite to a remote centre for monitoring and maintenance purposes (Fig. 3). The wireless network is made of GPS devices able to receive the corrections provided by a base station located near the structure. This technique is named RTK (Real Time Kinematic) and is used to enhance the precision of the positioning systems. It measures the phase of the carrier wave of the signal, and it relies on a single reference station to provide a centimeter-level accuracy. The system is based on the open-access software environment of a credit card-size computer, which is used as data collector and named (Raspberry Pi 2). It is run as Raspbian, a Debian-based computer operating system, and the operations are coded in Python [35]. The core activities are: - store sensor data inside the memory of the data collector in real-time, - create and transmit data frame to the coordinator through ZigBee, - transfer the information with a remote center via satellite, and - process the data with the displacement estimation method. Each measurement point is made by a sensor node (ZigBee) and a GPS receiver coupled with three axial accelerometers for realtime data collection and transmission. This network covers a small area, and the wireless communication represents a good solution for collecting and transmitting the monitoring data. Furthermore, the coordinator node may communicate with the remote center via satellite in case of the terrestrial network failed. The satellite antenna system for the field-testing was also proposed in the ESA 3InSat project [36]. The system is from Hughes and is identified by the model 9450 [37]. It is composed of a satellite terminal (User Terminal, UT) and of an auto-pointing antenna. To overcome the constraints of GSM/3G/4G connections, it exploits a VPN on site and connects it over the satellite link to a remote server. The information is stored by the central server into a database, and it is post-processed. In this way, the monitoring of a structure can be updated in real-time and with high precision.
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
49
Fig. 2. Top view (A) and bottom view (B) of the ZigBee device.
Fig. 3. Architecture of the real-time monitoring system.
3.3. Operational principle At idle time, the sensor node remains in sleeping state until the monitoring system designed for the civil infrastructure works properly. When it is cut-off, and the communication can not be executed by the existing networks, the power generated by portable and rechargeable batteries is provided to the prototyped system. To clarify its operational principle, a functional flowchart of the designed prototype is shown in Fig. 4. The main premise is the availability of a device incorporating GPS positioning together with a standard inertial sensor. As soon as the battery power is provided to the system, the sensing device produces, at each time step, NMEA-0183 strings, and comes with its wire ending (USB interface). The bottleneck is exactly the power one has to provide to the data collector. Thus, low consumption surrogates of the computers are welcome and modules of the Raspberry PI class are used. To minimize the power consumption, the conceived architecture creates a byte array (i.e., a sequence of numbers) and send the sentence by a ZigBee modulus to the Coordinator Node. Once the buffer is full, the software stores the message inside the memory of the Coordinator Node, which can be equipped with
higher power availability. From here, one can rely on internet or, to bypass any liability problem, on satellite phones, to transmit the information to the remote center. 4. Test results and discussion The results of two different experimental campaigns are presented in this section. The first part addresses the tests carried out in the laboratory to validate the performance of the wireless data acquisition system. The second sub-section reports on experiments executed in a train maintenance yard. Focus is here on the satellite transmission of positioning data for a railway convoy. 4.1. Wireless data acquisition testing Some laboratory tests are analyzed from the point of view of the wireless link quality. Two aspects are worth being discussed: data packet loss, in term of packet reception rate, and energy consumption. The data loss issue is addressed by a reliable communication method with a character checking mechanism. This type of ap-
50
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
Fig. 5. A three-dimensional graph of the NMEA string transmitted (blue) and received (orange) during the tests.
Fig. 4. Functional flow chart of the proposed system.
proach is adopted to avoid data loss and sentence duplication as well. After acquiring a NMEA string from the GPS sensor, the RasPi_I will send each single character of the GGA message to the RasPi_II within the time window of a second. If the message is checked to be correct by the character control, it is regarded as having been successfully transmitted. Otherwise, the Sensor Node will retransmit the sentence to the Coordinator Node. Since the RasPi_I continues to acquire a new string every second from the GPS sensor, this procedure may cause a delay in the final collection of the transmitted data. To avoid this problem, the “corrupt” strings are sent with a period (0.1 s) less than the sampling of the system (t = 1s). In this way, the next operation can be performed without changing the software behavior. Each sentence is also marked by a sequence number, which reports the date and the acquisition time. After the development of the proposed data acquisition system, some tests were carried out to estimate the extent of the data loss. During the tests, some parameters was changed such as the time length when the distance between the two devices was 10 m (T01, T02, and T03), 50 m (T04, T05, and T06), and 100 m (T07, T08, T09, and T10). The results are summarized in Table 1. The PRR (Packet Reception Rate) is defined as the number of successfully received data packets divided by the number of actually transmitted data packets. In Table 1, its value is expressed in percentage. The three-dimensional graph of Fig. 5 compares the percentage of NMEA strings transmitted (blue) and those received (orange) by the Coordinator Node via ZigBee.
It is worth noting that the antenna setup greatly affects the radio communication. The PRR of the wireless data acquisition system is almost 100% because the antenna of the Sensor Node was elevated to the same height of the Coordinator, as listed in Table 1. The wireless data acquisition system adopts omni-directional antennas with a high gain of 5.4dBi, but the communication range can be extended. However, a higher gain will lead to a narrower beam width and a higher directivity, since the radiation power is concentrated on the main lobe. Therefore, all the device antennas have to be set up in such as a way that their main lobes are the same plane in order to minimize the path loss of the radio signal channel. The RasPi is powered at 4.94 V and the power consumption, when receiver and wireless device are powered on, is almost 400 mA. Therefore, an external battery of 10 Ah (ROMOSS Sailing 5), which delivers 5 V and 20 0 0 mA of current output, is used to provide an autonomy of 24 h [38].
4.2. Satellite transmission at a specialty test-bed The system assessment, in terms of robustness and reliability, of the communication links (wireless and satellite), was the purpose of the tests carried out, April 2016, at the railway site of Barrow Hill. This locomotive maintenance yard is located near Chesterfield in the United Kingdom [39], and the case study was designed with the collaboration of the Centre for Railway Research (NewRail). The experimental investigations were carried out in an area of 8.5 ha with continuous track of 1200 m. The site is divided in three areas according to the maximum speed allowed for each path: (i) Yard Lines at 2.2 m/s, (ii) Network Rail Access at 4.4 m/s, and (iii) Branch Line at 8.9 m/s, as shown in Fig. 6. The convoy is made by the locomotive, three wagons and one container. Each wagon is equipped with a SPARTACUS Tracking Unit (STrackU), consisting of an inertial sensor (Ellipse-N), a wireless device (ZigBee), and a data collector (RasPi). The container is carried on board the last wagon. The STrackUs power is supplied by portable batteries. They are set into a protective box and fixed on the window of the vehicles with a double-sided tape, as shown in Fig. 7(b).
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53
51
Table 1 The PRR values for the laboratory tests. Id
UTC start (hh:mm:ss)
UTC end (hh:mm:ss)
Length (mm:ss)
NMEA strings (n)
PPR (%)
T01 T02 T03 T04 T05 T06 T07 T08 T09 T10
12:01:09 12:27:44 12:35:05 12:43:57 12:47:00 12:55:53 13:05:23 13:15:15 13:20:17 13:29:24
12:09:27 12:34:14 12:43:22 12:46:35 12:55:17 13:04:10 13:13:40 13:19:54 13:28:34 13:33:48
08:16 06:58 08:17 02:38 08:17 08:17 08:17 04:39 08:17 04:24
499 392 499 159 499 499 499 281 499 266
99,80 99,74 99,80 99,37 99,80 99,80 99,80 99,64 99,80 99,62
Fig. 6. Layout and railway convoy of the Barrow Hill site.
Fig. 7. Systems on board the locomotive (a) and wagons (b).
The desired information is recorded in NMEA format [40]. The sentences are acquired by the developed software of each STrackU, and sent via wireless to the SComU (SPARTACUS Communication Unit), which is placed inside the locomotive (Fig. 7(a)). The system on board the locomotive (SColU) is conceived: (i) to collect the positioning data of each STrackU, and (ii) to transmit via satellite the information to a remote center. For this set of tests, the position accuracy of each vehicle is the aim for which
the transmission system was developed. The results for this case study are reported by estimates of the PPR of the two types of communication for a freight train in the railway site. First, the ZigBee behavior from a wagon or a container to the head train, and then the transmission, based on a VPN over satellite, of the NMEA messages from the locomotive to a remote center. The results are summarized in Table 2.
52
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53 Table 2 PRR values achieved during the field tests. Id
UTC start (hh:mm:ss)
UTC end (hh:mm:ss)
Length (mm:ss)
NMEA strings (n)
PPR (%)
L1 W1 W2 W3 C1
09:15:01 09:37:49 09:37:47 09:38:01 09:37:48
09:51:05 10:37:18 10:37:16 10:37:21 10:37:26
36:04 59:29 59:29 59:20 59:38
2732 372 373 364 366
100,0 99,46 99,73 98,90 99,45
The first test (L1) is addressed to check the data transmission via satellite of the SComU, while the results labelled W1, W2, W3 and C1 are referred to the wireless communication of the message from wagons/container to the locomotive. In particular, the above listed results were elaborated from the data stored by the software in the last day of measurements. The outcome of this analysis allows one to conclude that the system behavior is fully performing also when the message has to be forwarded from the data collector (RasPi) to the SComU, enabling the satellite broadcast. 5. Conclusions An inertial sensor, which also includes a GPS receiver (Ellipse), is coupled with a data collector (RasPi) and wireless technologies (ZigBee) to develop a low-cost data acquisition system in the SHM area. For this purpose, a Kalman filter-based approach is exploited in view of an accurate estimation of the displacement induced on civil structures by external actions. It implements a data fusion of the measured responses. The technology also uses the ability to receive corrections from a reference station, providing a centimeterlevel accuracy. The ultimate goal is to ensure the communications even if standard telecommunication tools fail due to catastrophic values of the external excitation intensity. Details on the hardware components and their software implementation are provided. A reliable wireless method of communication between nodes is adopted to avoid data loss issues. It is based on a suitable charcter checking mechanism. At last, the test results of an experimental campaign are discussed to demonstrate that the entire system features a good balance between the three considered aspects: power consumption, communication range, and link quality.
required to manage the counters, (iii) serial, which drives the serial communication, and (iv) time, used to access the system clock. Some constants (baud-rate and buffer-size) are then assigned to fix a file whose name comes with date and time. port = "/dev/ttyUSB0" baud = 38,400 length = 82 control = "^" ser = Serial(port, baud) The next step is referred to the RasPi_II and sees as input the file name. It also starts the device and collects the received information in an array of suitable size: def getFileName(name): def increment_filename(fn): fn, extension = os.path.splitext(name) n =1 yield fn + extension for n in itertools.count(start= 1, step= 1): yield ’%s%d%s’ % (fn, n, extension) for filename in increment_filename(name): if not os.path.exists(filename): return filename The array elements are sent one by one to the ZigBee with a control character ending the transfer: def check(): while 1: if(ser.read(1) == control): read() This tool appends what it reads to a new string through the “save” function, but only if such a string does not exceed the buffer’s length: def read(): string = "" while 1: if(len(string) > length): check() break elif (len(string) == length): save(string) string += ser.read (1)
Acknowledgments The research activity summarized in this paper has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n° [313002]. It was also supported by Athenaeum grants from the University of Pavia.
At last, the software opens a file and appends a string to it. The file is eventually closed.
Appendix According to Fig. 4, the GPS device is set to provide a GGA string per sec in the NMEA-0183 format. It is connected to the first data collector (RasPi_I) by a USB link. A second USB interface hosts the ZigBee Sensor that transfer the data to the ZigBee Coordinator and from here to the second data collector (RasPi_II), where the strings are saved in a text file. All the steps listed above requires a specific software. As RasPi_I is switched on, the following commands are executed: import serial, itertools, os from time import strftime, gmtime, sleep The software imports first the libraries of the Python language. They are: (i) os, necessary to access the File System, (ii) itertools,
def save(string): file = open(filename, "ab") file.write(string) file.close() filename = getFileName() check() References [1] Nassif HH, Gindy M, Davis J. Comparison of laser Doppler vibrometer with contact sensors for monitoring bridge deflection and vibration. NDT&E Int 2005;38(3):213–18. [2] Bai R, Ostachowics W, Radzienski M, Cao M. Vibrational damage detection using fractal surface singularities with noncontact laser measurement. J Vib Control 2014;22(11):2569–81. [3] Balkaya C, Casciati F, Casciati S, Faravelli L, Vece M. Real-time identification of disaster areas by an open-access vision-based tool. Adv Eng Softw 2015;88(1):83–90.
S. Casciati, M. Vece / Advances in Engineering Software 112 (2017) 46–53 [4] Wahbeh AM, Caffrey JP, Masri SF. A vision-based approach for the direct measurement of displacements in vibrating systems. Smart Mater Struct 2003;12(5):785–94. [5] Lee JJ, Shinozuka M. A vision-based system for remote sensing of bridge displacement. NDT&E Int 2006;39(5):425–31. [6] Wu L, Casciati F. Local positioning systems versus structural monitoring: a review. Struct Control Health Monit 2014;21(9):1209–21. [7] Park KT, Kim SH, Park HS, Lee KW. The determination of bridge displacement using measured acceleration. Eng Struct 2005;27(3):371–8. [8] Sim SH, Spencer BF Jr, Nagayama T. Multimetric sensing for structural damage detection. J Eng Mech 2011;137(1):22–30. [9] Kim J, Kim K, Sohn H. Autonomous dynamic displacement estimation from data fusion of acceleration and intermittent displacement measurements. Mech Syst Signal Process 2014;42(1-2):194–205. [10] Sung SH, Park JW, Nagayama T, Jung HJ. A multi-scale sensing and diagnosis system combining accelerometers and gyroscopes for bridge health monitoring. Smart Mater Struct 2014;23(1):1–14. [11] Roberts GW, Meng X, Dodson AH. Integrating a global positioning system and accelerometers to monitor the deflection of bridges. J Surv Eng 2004;130(2):65–72. [12] Park JW, Sim SH, Jung HJ. Displacement estimation using multimetric data fusion. IEEE/ASME Trans Mechatron 2013;18(6):1675–82. [13] Park JW, Sim SH, Jung HJ. Wireless displacement sensing system for bridges using multi-sensor fusion. Smart Mater Struct 2014;23(4):1–12. [14] Cho S, Sim SH, Park JW, Lee J. Extension of indirect displacement estimation method using acceleration and strain to various types of beam structures. Smart Struct Syst 2014;14(4):699–718. [15] Cho S, Yun CB, Sim SH. Displacement estimation of bridge structures using data fusion of acceleration and strain measurement incorporating finite element model. Smart Struct Syst 2015;15(3-4):645–63. [16] Sim S. Estimation of flexibility matrix of beam structures using multisensor fusion. J Struct Integr Maint 2016;1(2):60–4. [17] Lynch JP, Loh KJ. A summary review of wireless sensors and sensor networks for structural health monitoring. Shock Vib Digest 2006;38(2):91–128. [18] Wang Y, Lynch JP, Law KH. Design of a low-power wireless structural monitoring system for collaborative computational algorithms. In: Proceedings of SPIE 10th annual international symposium on NDE for health monitoring and diagnostics, 5768; 2005. p. 106–17. [19] Nagayama T, Rice JA, Spencer BF Jr. Efficacy of Intel’s Imote2 wireless sensor platform for structural health monitoring applications. In: Proceedings of AsiaPacific workshop on structural health monitoring, Yokohama, Japan; 2006. [20] Ni YQ, Xia Y, Liao WY, Ko JM. Technology innovation in developing the structural health monitoring system for Guangzhou new TV tower. Struct Control Health Monit 2008;16(1):73–98. [21] Casciati S, Faravelli L, Chen ZC. Energy harvesting and power management of wireless sensors for structural control applications in civil engineering. Smart Struct Syst 2012;10(3):299–312.
53
[22] Casciati S, Chen ZC. An active mass damper system for structural control using real-time wireless sensors. Struct Control Health Monit 2012;19(8):758–67. [23] Casciati F, Casciati S. Structural health monitoring by Lyapunov exponents of non-linear time series. Struct Control Health Monit 2006;13(1):132–46. [24] Chen ZC, Casciati S, Faravelli L. In-situ validation of a wireless data acquisition system by monitoring a pedestrian bridge. Adv Struct Eng 2015;18(1):97–106. [25] Casciati S, Chen ZC. A multi-channel wireless connection system for structural health monitoring applications. Struct Control Health Monit 2010;18(5):588–600. [26] Guinamard A. Ellipse AHRS & INS - high performance, miniature inertial sensors user manual. France: SBG Systems; 2014. [27] CXL-LF Series - High Sensitivity Accelerometers (2016). Document Part: 60200 0 02-01, USA. Crossbow Technology Inc. [28] Trimble 147A High Resolution Accelerometer (2016). Trimble Navigation Limited, USA. [29] Trimble R10 GNSS System (2016). Trimble Navigation Limited, USA. [30] Caron F, Duflos E, Pomorski D, Vanheeghe P. GPS/IMU data fusion using multisensor Kalman filtering: introduction of contextual aspects. Inf Fusion 2006;7(2):221–30. [31] Smyth A, Wu M. Multi-rate Kalman filtering for the data fusion of displacement and acceleration response measurements in dynamic system monitoring. Mech Syst Signal Process 2007;21(2):706–23. [32] Zhou Z. GPS navigation data fusion using composed correcting Kalman filtering method. J Comput Inf Syst 2011;7(6):2006–12. [33] Chatzi EN, Fuggini C. Structural identification of a super-tall tower by GPS and accelerometer data fusion using a multi-rate Kalman filter. In: Proceedings of the 3rd international symposium on life-cycle civil engineering, 2012; 2012. p. 144–51. [34] Texas Instruments (2009). CC2530 software examples user’s guide, Texas Instruments Incorporated, Dallas, Texas, USA. [35] Bradbury A, Everard B. Learning Python with Raspberry Pi, Chichester, United Kingdom: John Wiley & Son Ltd; 2014. ISBN 978-1-118-71705-9. [36] 3InSat, Train Integrated Safety Satellite System (2014). ESA Project: Contract 40 0 0105788/12NL/NR. [37] Hughes 9450-C11 BGAN mobile satellite terminal (2010). Hughes Network Systems, LLC. [38] Casciati S, Chen ZC, Faravelli L, Vece M. Synergy of monitoring and security. Smart Struct Syst 2016;17(5):743–51. [39] Casciati F, Casciati S, Faravelli L, Vece M. Implementation of communication, collection and tracking units in crisis operation. In: Proceedings of EWSHM 2016, 8th European workshop on structural health monitoring, Bilbao, Spain; 2016. [40] Trimble. NMEA-0183 messages guide for AgGPS receivers. Trimble Navigation Limited; 2004.