IOMAC'15 DEVELOPMENT AND APPLICATON OF A LONG AND SHORT TERM MONITIORING/SOFTWARE SYSTEM WITH A SMART SENSOR NETWORK

Transcription

1 IOMAC'15 6 th International Operational Modal Analysis Conference 2015 May12-14 Gijón - Spain DEVELOPMENT AND APPLICATON OF A LONG AND SHORT TERM MONITIORING/SOFTWARE SYSTEM WITH A SMART SENSOR NETWORK Thomas Schmidt, Arne Büttner, D. Herfert, M. Gollnick 1 Prof. Dr.-Ing Th.Schmidt, University of Applied Sciences Magdeburg-Stendal, Thomas.schmidt@hs-magdeburg.de 2 Dipl-Ing (FH), M.Eng. A.Büttner,, University of Applied Sciences Magdeburg-Stendal, Arne.buettner@hs-magdeburg.de. 3 Dipl-Inf. D. Herfert. Gesellschaft zur Förderung angewandter Informatik e.v., Berlin, herfert@gfai.de 4 M.Sc. M. Gollnick, Gesellschaft zur Förderung angewandter Informatik e.v., Berlin, gollnick@gfai.de ABSTRACT The development of appropriate hard- and software tools to serve in the long- and short term condition monitoring of civil engineering structures is one of the major goals of the ongoing work of the authors. When using the term condition we speak about the structural reliability, endurance and the durability of a building. Furthermore the authors try to get more insight in the long term condition development of buildings in order to reflect this insight in better designed structures. To reach this goal three tools has been developed: first a wireless sensor network, second, the development of analysis (FEM/EMA/OMA) tools to determine the condition and thirdly, the development of a platform that combines the first and second tools in one software package. That means planning and evaluation of experimental investigations of structures as well as the analysis of structural models shall be combined. Measurement and analysis results shall be evaluated simultaneously. Four partners work together in one project to reach the goal. The sensor network is developed by Esys-Gmbh + Dr.Wolf Gmbh, the EMA/OMA-Tools are developed by GfaI and the CAD/FEM/EMA/OMA platform is developed by the authors, and the development of a special shaker including data acquisition software is done by Baudynamik Heiland & Mistler GmbH Consulting Engineers. The paper presents the current state of the project. This is done by a case study of a 12 story reinforced concrete building using in parallel a cable based system and a wireless sensor network system the results obtained from shaker induced and ambient excitations measurements are compared and discussed. Keywords: Wireless Sensor Network, Experimental Modal Analysis, Operational, Modal, Analysis, CAD/FEM/EMA/OMA platform, shaker 1. INTRODUCTION The analysis and condition monitoring of civil engineering structures are getting more and more important since more and more structures (especially bridges) show severe damages but there is no

2 money available for retrofit or restauration. The evaluation of the present structural static and dynamic resistance capacity is of major interest. Also quality insurance and maintenance of new build structures gets more and more into the focus. Appropriate hard- and software tools as sophisticated aid in this context is a major goal the authors. The paper presents the development and application of a wireless smart sensor network system in combination with a specially developed software platform to maintain the above described demands. Within the research project BaSnet a wireless smart sensor network platform is developed by one of the three project partners (Esys-Gmbh [1]). A special transmitter was developed by a sub-partner (Dr. Wolf Gmbh [2]) that guaranties a much better network range compared to most of the system available on the market. The wireless system is presented in chapter 2. The other two project partners are developing a 3D-CAD/FEM software package (University Magdeburg) combined with a software package to perform EMA and OMA analysis with state-of-the-art algorithms (GfaI [3]). Within in the 3D-CAD/FEM-Software the structure can be pre-analyzed as well as the measurement setups and results can be evaluated by the integrated EMA/OMA package. The goal to apply EMA and OMA is to investigate how to combine these methods for the sophisticated long- and/or short term monitoring of civil engineering structures. The software packages in combination with the sensor network are developed for long- and short term monitoring tasks. Chapter 3 presents the software tool Good Vibrations developed by our partner GfaI. Chapter 4 presents the shaker developed by our partner Baudynamik Heiland & Mistler GmbH Consulting Engineers [4]. In the following chapter the CAD/FEM tool developed by the authors is presented briefly. In chapter 6 the application of all parts of the whole monitoring system is presented the analysis of a 12 story reinforced office building the so called Thyssen/Krupp tower in Bochum Germany. The example will verify the different parts of the project, show the current state and further work to be done. 2. WIRELESS SENSOR NETWORK 2.1. Sensor Hardware The first steps made by the authors in cooperation with [1] was the development of a wireless sensor network presented in [4]. The sensor node developed in the project presented here is significantly improved. The radio module was completely redesigned. The new radio module was developed by [2] and shows a significant improvement of range (range in buildings up to 70 m, range in free field 500 m) and data transmission reliability. Especially in conditions were electrical disturbances are an issue (e.g. by railroads) the newly developed radio modules show excellent behavior. Unfortunately the measurements presented in the case study section could not be performed by the improved sensor nodes with the new radio modules because of difficulties during chip production. Therefore only a sensor network with ordinary WLAN radio modules were used. A more detailed description of the new radio modules can be found here [2]. A sensor node and its software interface are shown in Figure 1. Figure 1. Sensor node and configuration software.

3 The developed sensor nodes can work in two modes. First, they can be used as data loggers with a UMTS module for remote access and an SD-Card slot to store the measured data on board. Up to 4 sensors (voltage, strain, temperature) can be attached. The voltage input can be chosen from 1 mv to 25 V. The sampling rate can be chosen from 1 Hz to 1000 Hz. Second, the modules can be used as a sensor node within a sensor network. At the moment the network can be setup with up to 10 sensor nodes (that means 4*10 = 40 sensors). At the end of the project up to 80 sensor nodes should be possible. The next generation of these sensor nodes will have online FFT, special filters, and other signal conditioning procedures directly on board. The nodes are synchronized via GPS. Each sensor node has its own GPS-Module. Each sensor can be attached to a computer via USB or via WLAN. Power supply is realized by two 3.6 V Lithium batteries or directly by power supply plug. A node can be run up to one week with the two batteries depending on the sample rate and surrounding conditions. In order to configure a sensor node a special GUI based software was developed. The configuration of each sensor node is yet not be possible from within the CAD/FEM/OMA/EMA platform. We hope to be able to realize this within the project runtime. For each node a special measurement schedule can be defined in the software. 3. EMA/OMA ANALYSIS TOOL Within the preceding project to the current one an OMA-Tool was developed by [3]. The software works similar to [5]. Table 1 lists the currently available OMA/ODS features of the software. The software can be run as a stand-alone tool and was named Good Vibrations. Figure 2 shows a picture of the current stand-alone version developed in MatLab [6]. Within the current project state-of-the-art EMA (experimental modal analysis) algorithms were added to the software (s. Table 1). Most of the algorithms are well known form literature. Additionally, a new EMA algorithm (POLYFSVD) developed by Herfert/Gollnick [3] was added. This algorithm is very fast and produces very good results. The software can import ASCII, BINARY universal file data as well as Artemis, Diadem, DasyLab and other file formats. Results can be exported in the same formats and as MatLab files. Table 1. Features of Good Vibrations. Operational Modal Analysis (OMA) Algorithms with Merging Strategies [7] FED Frequency Eigenvalue Decomposition (Peak picking) EFED Enhanced Frequency Domain Decomposition COV Covariance driven (SSI) PC/UPC Principal Component and Unweighted Principal Component (SSI) CVA Canonical Variant Analysis Operational Deflection Shape (ODS) Algorithms ODS-Freq. ODS in the frequency domain ODS-Time ODS in the time domain Experimental Modal Analysis (EMA) Algorithms ERA Eigensystem Realization Algorithm ERA-DC ERA with Data Correlation LSCE Least-Squares Complex Exponential Poly-LSCE Poly-Reference LSCE LSF Least-Scares Complex Frequency POLYMAX Poly-Reference Least-Squares Complex Frequency POLYFSVD Poly-Frequency Singular Value Decomposition

4 Figure 2. EMS/OMA Software Good Vibrations 4. SHAKER DEVELOPMENT The consulting partner Heiland & Mistler [4] developed a mobile shaker (Butterfly). Butterfly is a servohydraulic shaker for artificial excitation and is our proprietary development. The shaker was made for worldwide mobile application. Typical fields of application are: Simulation of train pass-bys on tracks e.g. to determine the insertion loss (please read our brochure Determination of Insertion Loss) Dynamic excitation of bridges Load tests of tower buildings Vibration excitation of massive structures like hydro dams/gravity dams Simulation of ground-borne vibrations on long distance Vertical and horizontal vibrations can both be simulated by simple alteration. Specially developed for shaking structures with low frequencies, Butterfly (s. Figure 3) shows high performance in an extraordinary frequency range of 0, Hz. The shaker moves its mass elements (weight: kg) with a maximum force amplitude up to 20 kn. Every load sweep is driven by a highly sophisticated sensor system. Within the BaSnet project the shaker was especially modified and further developed for horizontal excitations as will be shown in chapter 6. The partner also develops a software tool for the optimal positioning of the sensors and planning of shaker tests.

5 Figure 3. Butterfly Shaker [4]. 5. CAD/FEM/OMA/EMA PLATFORM In order to combine the briefly described components a platform was developed by the author. The platform is based on CAD/FEM tool called S3D developed by Prof. Rothe [8] and the author during the last years. This tool originally combined CAD standard features for drawing of 2D and 3D models with a FEM kernel developed by Rothe/Schmidt [9] in the late 1980 th. The FEM kernel was developed for the linear and non-linear (material and geometrical) static and dynamic analysis of civil engineering structures. It has a large element library ranging from 1D to 3D elements, contactelements, et al. The features of the FEM kernel are similar to ADINA. Within the BaSnet project this CAD/FEM software tool is further developed so that all parts of a monitoring system are included. The platform shall be used to plan and perform monitoring tasks with pre- and post-processing features as FEM and OMA/EMA. The results from all modules (FEM, measurements, OMA, EMA) can be overlayed so that it can be used as a planning and an evaluation tool. Since the platform S3D is developed in C++ with Qt it can be used on nearly all computer platforms. Figure 5 shows the platform with the model of the Thyssen/Krupp Tower as example. Figure 5. Platform S3D for FEM/EMA/OMA Analysis of Thyssen/Krupp Tower.

6 6. EXAMPLE OF EVALUATION To evaluate the current state of the components developed in BaSnet (sensor nodes, sensor network, Good Vibrations, Butterfly EMA/OMA, S3D-FEM/EMA/OMA) two buildings were tested (UNI- Tower Magdeburg and Thyssen/Krupp Tower Bochum) with the wireless monitoring system described above. Due to size limitations for this paper only the Thyssen/Krupp Tower shown in Figure 6 is presented. The structure was tested under ambient and shaker induced loadings. In order to proof the accuracy of the sensor nodes the measurements were done with a standard cable based system (~ 3 km cables!) and at the same time with a wireless sensor network. At the time when the test was performed the sensor nodes could not be run with GPS-synchronization nor with the new radio modules because of a delay in the hardware production. That means the sensors could not be synchronized. The network was set up as an ad hoc WLAN TCP/IP network with five sensor nodes. The comparison of the test results obtained by this wireless sensor network with that of the cable based system will show how the lack of synchronization influences the EMA/OMA results. Figure 7 presents the cable based test setup on the left and the sensor network setup with five sensor nodes at the right. At each sensor node three geophones (velocity sensors) were attached and in addition there was a temperature sensor attached to one sensor node. The measurements were performed under three different loading conditions during a period of two days. 1. Ambient excitation during and Shaker induced excitation at the northern end of the 11 th floor 3. Shaker induced excitation at the southern end of the 11 th floor North Quelle: Wikipedia / Foto: Cschirp South Figure 6. Thyssen/Krupp Tower in Bochum. Shaker on 11 th floor of the tower cross section of each floor. The ambient conditions during the two days are summarized and the test runs with shaker excitations are summarized in Table 2. In Figure 9 a load-time function (sweep from 0.6 to 12) that were induced by the shaker (s. Figure 8) and the corresponding vibrations are shown. As expected, the vibrations measured by both systems were the same but the cable based system showed a better signal to noise ratio. Though the wireless sensor nodes were not synchronized the OMA and EMA results calculated from their data were not significantly different from the results of

7 the cable based system. For higher natural frequencies the time lack between the sensor nodes will certainly become more significant. Figure 7. Sensor Setups; left cable based system 25 sensors, right: wireless network system 15 sensors Table 2. Ambient testing conditions and shaker load patterns. Test Run Conditions Ambient Wind mainly from west, max squall velocity 8 m/s, temperature 0-3 C Ambient Wind mainly from west, max squall velocity 5 m/s, temperature 2-5 C Sweep M8 1.3 to 12 Hz, 0.04 Hz/s, force controlled ~3kN, Sweep M9 0.6 to 1.5 Hz, 0.01 Hz/s, displacement controlled 40 mm Single Frequency M10 1,06 Hz, displacement controlled 45 mm Single Frequency M Hz, displacement controlled 45 mm Single Frequency M Hz, displacement controlled 25 mm Single Frequency M Hz, displacement controlled 8 mm Single Frequency M Hz, displacement controlled 6mm Sweep M to 5 Hz, 0.01 Hz/s, force controlled Figure 8. Horizontal shaker with ~2.5 t on 11 th floor of Thyssen/Krupp Tower Figure 8 illustrates the shaker excitation and the corresponding structural responses. Some results of the data evaluations are presented in Figure 10 and Table 3. The first bending modes in x and y direction are identified as 1.06 Hz and 4.35 Hz for the x-direction. It can be seen, that the natural frequencies obtained from the ambient measurements (OMA) and the shaker tests (EMA) differ up to

8 4 %. Interesting to notice is, that the OMA analysis give much clearer stability diagrams as the EMA analysis. Furthermore some mode shapes not could be identified by EMA whereas by OMA significantly more natural frequencies and corresponding mode shapes could be identified. The damping for the first mode shape was evaluated from the shaker test at 1.06 Hz.The damping is calculated form the logarithmic decrement of the decay process is 0.95 %. The damping values calculated by the different OMA/EMA algorithms are not consistent and differ up to 200 %. The first natural frequencies calculated form EMA are lover compared to that calculate form OMA. This may be explained by the fact that the average factor between velocity amplitudes (~0.25 mm/s, 11 th floor) under ambient excitation and those during shaker excitation (~2-4 mm/s, 11 th floor) is about Obviously the stiffness of the structure is lower for higher excitation levels. Table 3. Comparison of EMA and OMA results. Mode Shape south EMA north OMA Frequency Frequency Frequency [Hz] [Hz] [Hz] 1.bending y torsional bending x bending z bending x torsional To evaluate the OMA results calculated by Good Vibrations the OMA analysis was performed using Artemis [5]. Table 4 displays the comparison. It can be said that Artemis and Good Vibrations show similar performance and produce nearly the same results. Artemis is a bit faster but since Good Vibrations is still in beta phase and implemented in Matlab. A significant performance improvement is expected after it is converted to C++ in the near future. Shaker Force [N] M8 shaker-sweep 1,3 12 Hz, x-direction Velocities [mm/s] cable based system, y-direction level 12 (south) wireless system, y-direction level 12 (south) cable based system, y-direction level 12 (north) wireless system, y-direction level 12 (north) cable based system, x-direction level 12 (north) wireless system, x-direction level 12 (north) (south) Figure 9. Measurement Results for sweep test and single frequency test

9 1,07 Hz 1,35 Hz 1,48 Hz 2,21 Hz 1. bending mode y 1. torsional mode 1. bending mode x 1. bending mode z 4,27 Hz 4,37 Hz 5,06 Hz 7,61 Hz 2. bending mode y 2. bending mode x 2. torsional mode 3. bending mode x Figure 10. Identified Mode Shapes from EMA/OMA Table 4. Comparison of OMA results - Artemis and Good Vibrations Artemis Good Vibrations Frequency Damping Frequency Damping Mode Shape [Hz] [%] [Hz] [%] 1.bending y torsional bending x bending z bending y bending x torsional bending x CONCLUSIONS AND PROSPECT The application example shows that the newly developed wireless sensor network system works well. The results of cable based and wireless system coincide well. The mode shapes and natural frequencies determined by the OMA/EMA tool are close to that obtained by Artemis. The hardware sensor nodes and shaker - could be proved as reliable and applicable. The next steps of the project will be some further tests with the final sensor hardware (radio module, GPS synchronization) and the final integration of all developed tools (CAD/FEM/EMA/OMS, sensor software, etc.) in one platform.

10 ACKNOWLEDGEMENTS The author gratefully thanks the Arbeitsgemeinschaft industrieller Forschungsvereinigungen "Otto von Guericke" e.v. (AiF) [11], for granting the above presented research project. A particular acknowledgement is for all members of our research team for their help in preparing the material for this paper. REFERENCES [1] ESYS GmbH, [2] Andreas C. Wolf, Matthias Mahlig, Benchmarking of WSN Solutions and IEEE PSSS based Solutions, FSGN 2010, [3] Gesellschaft für angewandte Informatik, GfaI, [4] Dr.-Ing. D. Heiland, Dr.-Ing. M. Mistler, Baudynamik Heiland & Mistler GmbH, [5] Palle Anderson, Identification of Civil Engineering Structures using Vector ARMA Models, Phd. Thesis 1995, Structural Vibration Solutions A/S, Artemis Modal, hht:// [6] [7] M. Dohler, P. Andersen and L. Mevel, Data Merging for Multi-Setup Operational Modal Analysis with Data-Driven SSI, Proceedings of IMAC XXVIII, Jacksonville, Florida, February 1-4, 2010 [8] [9] Schmidt, Th. (2010). Development of a Wireless Bridge Monitoring System with Energy Harvesting Modules. Proceedings of IMAC XXVIII, Jacksonville, Florida, February 1-4, [10] König, G., Rothe, D., Schmidt, Th.: SNAP - Ein nichtlineares Finite Element Programm, in: Erwin Stein (Hrsg.), Nichtlineare Berechnungen im konstruktiven Ingenieurbau, Springer Verlag [11] Arbeitsgemeinschaft industrieller Forschungsvereinigungen "Otto von Guericke" e.v.(aif)