Monitoring of Environmental Effects on Modal Estimates of Large Structures

Transcription

1 Monitoring of Environmental Effects on Modal Estimates of Large Structures Elsa CAETANO Associate Professor Carlos MOUTINHO Assistant Professor Álvaro CUNHA Full Professor Wei-Hua HU Post-Doc Researcher Filipe MAGALHÃES Assistant Professor Sandro AMADOR PhD Student Summary This paper aims at describing some applications recently developed by the Laboratory of Vibrations and Structural Monitoring ( of FEUP in which an accurate monitoring of the effects of environmental factors on modal frequency estimates was continuously performed, and appropriate statistical techniques were employed to remove or mitigate such effects, enabling a correct vibration based damage detection approach. The monitored structures are the Infante D. Henrique bridge, the Pedro e Inês and the FEUP stress-ribbon footbridges, and the suspension cable roof of Braga stadium. The paper focuses on the effects of temperature on modal parameters. Keywords: Continuous Dynamic Monitoring; Environmental factors; Temperature; Intensity of traffic, Modal frequency estimates; Damage detection. 1. Introduction The concern with different complementary aspects of the structural behaviour of bridges and special structures, such as the understanding and accurate modelling of the aerodynamic behaviour, the checking of vibration serviceability limits, as well as the verification of the efficiency of vibration control devices, or the ageing and structural degradation of large structures, has motivated an increasing interest of designers, constructers and owners by vibration-based structural health monitoring (SHM) programs. This recent trend has been enabled not only by important technological progresses of instrumentation systems for dynamic testing and continuous dynamic monitoring of large structures, but mainly by the development and software implementation of efficient strategies to efficiently and automatically extract useful information from the huge amount of data regularly collected by dynamic monitoring systems. A key aspect in this perspective is the influence that environmental and operational factors, like temperature, traffic intensity or wind speed can have on the variability of modal estimates, as these changes can be of the same order of magnitude of the modal variability induced by structural damage. In this context, this paper aims at describing some applications recently developed by the Laboratory of Vibrations and Structural Monitoring (ViBest, of FEUP, in which an accurate monitoring of the effects of those environmental factors on modal frequency estimates was continuously performed, and appropriate statistical techniques were employed to remove or mitigate such effects, enabling a correct vibration based damage detection approach. The monitored structures are the Infante D. Henrique roadway bridge, the Pedro e Inês and the FEUP stress-ribbon lively footbridges, and the suspension cable-roof of Braga stadium. The paper focuses specifically on the effects of temperature on modal parameters of those structures and the purpose is to evaluate the minimum level of damage that can be detected by means of continuous dynamic monitoring.

2 2. Infante D. Henrique bridge 2.1 The bridge The Infante D. Henrique Bridge (Figure 1), over Douro river at Porto, is formed by two mutually interacting fundamental elements: a very rigid prestressed reinforced concrete box beam, 4.50 m in height, supported by a very flexible reinforced concrete arch, 1.50 m thick. The arch spans 280 m between abutments and rises 25 m until the crown, thus exhibiting a shallowness ratio greater than 11/1. In the 70 m central segment, arch and deck join and define a box-beam 6 m in height. The arch has constant thickness and a linearly varying width from 10 m in the central span up to 20 m at the abutments. Fig. 1: Schematic view of Infante D.Henrique bridge, characterisation of monitoring system. 2.2 Continuous dynamic monitoring results The permanent dynamic monitoring system installed by ViBest/FEUP in this bridge has been in continuous operation since September 2007, and it is formed by 12 force-balance accelerometers, two 6-channel digitisers and one computer installed inside the box-girder, according to the scheme shown in Figure 1. The acceleration time series collected every 30 minutes at 50 Hz are immediately transferred to FEUP via Internet and processed using appropriate software (DynaMo) specifically developed for this purpose [1]. This software enables the automatic identification of modal parameters based on EFDD, SSI and p-lscf methods, and the application of statistical tools (multivariate linear regression or principal component analysis (PCA)) to remove the effects of environmental/ operational factors on the natural frequency estimates [2]. Frequency (Hz) Set/2007 Set/2008 Set/2009 Set/2010 Set/2011 Fig. 2: Estimates of first 12 natural frequencies in the period Sept 2007 Sept Figure 2 shows the time evolution of the 12 first natural frequencies identified on the basis of the p-lscf method in the period Sept 2007 to Sept Inspection of this figure may suggest that frequency estimates are very stable. However, a zoom of individual frequency plots evidences the occurrence of daily and seasonal effects, as shown in Figure 3 for the natural frequency of the second vertical bending mode during a period of observation of four years. Frequency variations with time stem from environmental (e.g. temperature, see Fig. 4) and operational (e.g. traffic intensity) factors, and though they are relatively small (variations in the range % are found), they can seriously disturb any attempt of damage detection based on natural frequency shifts. Therefore, it is important to mitigate the effect of such influences using appropriate statistical models and building suitable control charts using a novelty index to flag the occurrence of slight structural damage [1] [3]. Figure 5(left) presents, for instance, the time evolution of the first natural frequency estimates before and after the removal of the environmental and operational effects, during two years. Inspection of this figure shows the efficiency of that correction, most of the estimates becoming then located inside a very narrow frequency range with amplitude of about Hz. This means that relatively small damage can be detected in the future, provided that the consequent frequency shifts are higher than that order of magnitude. In the present

3 application, damage with potential to change the first frequency more than 1% could be detected. This conclusion was confirmed analysing different slight damage scenarios idealized numerically and applying appropriate control charts, as indicated in Figure 5(right) [1] Fig. 3: Time evolution of natural frequency of second vertical bending mode from 13/09/07 to 13/09/ Oct07 Jan08 Apr08 Jul08 Fig. 4: Variation of first natural frequency (dots) vs temperature measured at the top of section T3 in one year (solid line) Frequency (Hz) Oct07 Jan08 Apr08 Jul08 Oct08 Jan09 Apr09 Jul Frequency (Hz) Oct07 Jan08 Apr08 Jul08 Oct08 Jan09 Apr09 Oct08 Jan09 Apr09 Jul09 Jul09 Fig. 5: (left) Time evolution of the first natural frequency estimates before and after the removal of environmental/operational effects during two years; (right) Control chart used for damage detection. 3. Pedro e Inês footbridge 3.1 The bridge This is a 275m long slender steel footbridge composed by a 110m span central parabolic arch raising 9.40m above the water level, two lateral parabolic half arches with spans of 64m, an approaching span of 30.5m on the north bank and a transition span of 6m on the south bank (Fig. 6). The continuously supported deck has a width of 4m and is formed by a L-shaped box cross section whose top flange is a composite steel-concrete slab 0.11m thick. In the lateral spans, arch and deck generate a rectangular box cross-section 4m x 0.90m. In the central part of the bridge, each L-shaped box cross-section and corresponding arch meet to form a rectangular box cross-section 8m x 0.90m, creating a central square with 8mx8m at mid-span. The central arch is supported by two groups of vertical piles with a depth of about 30m. The stiffness provided by these foundations is low due to poor mechanical characteristics of soil layers beneath the river bed. The supports provided at the abutments and at the left bank intermediate column allow the longitudinal displacement of the deck and block transversal movements. Fig. 6: Lateral view of Pedro e Inês footbridge. Temperature (ºC) Frequency (Hz)

4 AV1 AV2 AV3 AV5 AV6 S3 S2 S1 30.5m 64.00m 55.00m 55.00m 64.00m 6.00m west AT4 east Fig. 7: Lateral and plan views of Pedro e Inês footbridge: location of accelerometers (AV1-AV3, AT4,AV5-AV6) and sections (S1-S3) instrumented with temperature sensors. 3.2 Continuous dynamic monitoring results The dynamic monitoring system installed by ViBest/FEUP at Pedro e Inês footbridge [4] [5], active since 2007, comprehends 6 uniaxial piezoelectric accelerometers located as schematically indicated in Figure 7. Five of these accelerometers measure vertical accelerations, whereas another one measures lateral vibrations at mid-span. These sensors are installed inside the metallic deck and are wired to an acquisition system located inside one of the concrete abutments of the structure. The acquisition system used is composed by a signal conditioner, and a digital computer incorporating an analogue-digital conversion board. The data acquisition is based on software developed with LabVIEW. A communication system is responsible to transmit data to FEUP using an ADSL line. Frequency estimates of a significant number of modes of vibration have been automatically obtained by the SSI-COV method based on the acceleration time series collected every hour, as shown for instance in Figure 8 for a period of three years. A zoom of the evolution of frequency estimates of the several modes shows that they increase in winter time and decrease in summer, which stems from temperature variations. Table 1 characterizes the range of some frequency estimates and their maximum relative variation with regard to the average value, which is in the interval 2.7%-3.9%. This naturally disturbs the detection of early structural damage. It is pointed that the presence of TMDs for several vertical modes in the range Hz causes difficulty in the identification of such modal parameters. Jun /07 Dec /07 Jun /08 Dec /08 Jun /09 Dec/ 09 Fig. 8: Identified natural frequencies from 1 st June 2007 to 31 st May Jun/ 10 Table 1: Variation of frequency estimates Average frequency (Hz) Frequency range (Hz) Maximum rel.variation (%) A detailed correlation analysis between natural frequency estimates, temperature and vibration levels can be found in [3]. This research shows that temperature is clearly the main factor inducing frequency changes. Figure 9 shows, for instance, the variation of the frequency estimates associated to the modes of vibration with mean frequencies of Hz and Hz, evidencing the existence of a linear trend of decrease of frequency with the increase of temperature, as well as the much less significant influence of the vibration level. This linear trend has been identified in all other natural frequencies analyzed.

5 Owing to this linear influence of environmental factors on the frequency estimates, Principal Component Analysis (PCA) has been applied with success to remove such adverse effects, enabling the reliable detection of small simulated structural damages at the foundations using 3 years of monitoring results [3]. Fig. 9: Identified frequencies vs temperature (red dots: estimates for high vibration levels). 4. FEUP stress-ribbon footbridge 4.1 The bridge This stress-ribbon footbridge (Figure 10) [6] establishes a link between the main buildings of FEUP and the students canteen. The two spans 28m and 30m long and the 2m rise from the abutments to the intermediate pier are the starting points for the definition of the bridge structural geometry. A continuous concrete cast-in-situ slab embedding four prestressing cables takes a catenary shape over the two spans, with a circular transition over the intermediate support, which is made of four steel pipes forming an inverted pyramid hinged at the base. The constant cross-section is approximately rectangular with external design dimensions of 3.80mx0.15m. South Accelerometers, signal conditioner and thermal sensors at 1/3 span Accelerometers, signal conditioner and thermal sensors at midspan North Unit 4 Unit 3 Unit 2 Unit 1 Longitudinal elevation NI Ethernet, DAQ devices, Power-line (b) Transverse elevation (c) Cross section Power supply Accelerometer, signal conditioner Fig. 10: Lateral view and schematic representation of FEUP campus footbridge.

6 4.2 Continuous dynamic monitoring results The continuous dynamic monitoring system implemented in FEUP stress-ribbon footbridge includes four sensor units. They are mounted separately on the lower surface of the bridge deck at both 1/2 and 1/3 of each span (Figure 10 (a) and (c)). Each unit comprises a vertical accelerometer, a signal conditioner and a thermal sensor. Acceleration signal conditioners and thermal sensors are connected via cable with National Instruments Ethernet data acquisition (DAQ) devices, which are incorporated in a steel box installed beneath the deck at the intermediate support (Figure 10 (b)). The NI Ethernet DAQ device is driven by a signal acquisition toolkit, generating a nearly real-time zipped acceleration and temperature files every 10 or 30 minutes, respectively, continuously. Signal files acquired under operational conditions are conveniently accessed via Internet. The continuous dynamic monitoring system has been operating from the 1 st of June 2009 up to now, except for occasional stops due to small technical problems. The identified frequencies of 12 modes in the range of 0-20Hz estimated by the SSI-COV method are shown in Figure 11 (this corresponds to all the first vibration modes, except the ones associated to frequencies of about 2.1 and 2.4 Hz, due to the perturbation resulting from the strong interaction with pedestrians). The variations of frequency estimates of higher order modes are quite clear. The tendency of increasing in winter time and decreasing in summer time reflects the seasonal environmental effects. Table 2 shows that the maximum relative variations of frequency estimates are in the range 15.3%-21.4%, which is significantly higher than in the two previous applications. Jun/09 Sep/09 Dec/09 Mar/10 Jun/10 Months Fig. 11: (Left) Identified frequencies from 1/6/09 to 31/5/10; (Right) Frequency of first mode vs temperature (black, red and blue dots: low, medium and high vibration levels, respectively). Table 2: Variations of frequency estimates Average frequency (Hz) Frequency range (Hz) Max. relative variation (%) A detailed correlation analysis between natural frequency estimates, temperature and vibration levels can be found in [3], showing that temperature is again the main factor inducing frequency changes. However, in this case, the influence of traffic intensity is slightly higher than in Pedro e Inês footbridge, as stressed by Figure 11 (right). Moreover, the trend of decrease of frequency estimates with the increase of temperature tends to vanish above 30ºC. Still, the existence of approximately linear correlation between the variations of frequency estimates of the several modes of vibration allowed the application of Principal Component Analysis (PCA) to remove the influence of environmental factors on the frequency estimates, enabling the reliable detection of small simulated structural damages at the connections of the deck with the abutments [3].

7 5. Braga Stadium suspension roof 5.1 The structure The Braga Stadium (Figure 12), in Portugal, is one of the football stadia constructed to host some of the matches of the 2004 European Football Championship. It was designed by the architect Eduardo Souto Moura, in collaboration with the engineering design office AFassociados [7]. The most interesting element of the stadium, from the structural engineering point of view, is, undoubtedly, the suspension roof, which consists of two concrete slabs supported by 34 pairs of full locked coil cables 0.35m distant. These pairs of cables span 202 m, with diameters ranging from 80 to 88mm and are spaced by 3.75m [8]. Fig. 12: Braga Municipal Sports Stadium: view of roof from the west side. Cross section. 5.2 Continuous dynamic monitoring results The dynamic properties of the suspension roof have been continuously monitored since March, 2009, when a dynamic monitoring system was installed by ViBest/FEUP on the west slab. This system is composed by a 6-channel 24-bit digitizer, a robust field processor and six force balance accelerometers. The system is programmed to collect and save one-hour long acceleration time series, which are then processed by autonomous identification software [9] [10]. Fig. 13: (Left) Time evolution of natural frequencies: (top) from 2009/03/25 to 2010/09/27 in the range of Hz; (bottom) from 2009/07/1 to 2009/07/10 of the modes 3, 4 and 5; (Right) Correlation of estimated frequencies and measured temperatures from 2009/07/01 to 2009/12/31.

8 Table 3: Variations of frequency estimates Average frequency (Hz) Conclusions Frequency range (Hz) Max. relative variation (%) Careful inspection of the monitoring results, as well as of temperature data available since the construction phase reveals that temperature plays also a major role in the variations of natural frequency estimates, as shown in Figure 13 (left). However, it s interesting to note that, probably due to the peculiar geometrically non-linear behaviour of the roof, the increase of temperature does not induce always a decrease of frequency estimates. Moreover, these variations have some nonlinear nature, as shown in Figure 13 (right). The variations of the natural frequency estimates obtained along almost two years of permanent monitoring of the roof structure are summarized in Table 3 and show relative variations no larger than 4.9%. The several applications developed by ViBest/FEUP and described in this paper show the influence of environmental factors, and particularly of temperature, on the variations of natural frequencies of bridges and special structures, as well as the feasibility of very accurate measurement of such influences based on appropriate continuous dynamic monitoring systems. Moreover, the application of suitable statistical tools, like multi-linear regression models or principal component analysis, enables the removal or mitigation of those influences, which is a key aspect for reliable vibration based damage detection based on the construction of suitable control charts. Upon removal of temperature, damage inducing variations of frequency higher than 1% could be detected. 7. References [1] MAGALHÃES F., Operational Modal Analysis for Testing and Monitoring of Bridges and Special Structures, Ph.D. Thesis, FEUP,, [2] MAGALHÃES F., CUNHA A. and CAETANO E., Vibration based structural health monitoring of an arch bridge: from automated OMA to damage detection, Mechanical Systems and Signal Processing, Vol. 28, 2012, pp [3] HU W.-H., Operational Modal Analysis and Continuous Dynamic Monitoring of Footbridges, Ph.D. Thesis, FEUP,, [4] CAETANO E., CUNHA A., MOUTINHO C. and MAGALHÃES F., Studies for controlling human-induced vibration of the Pedro e Inês footbridge, Portugal. Part 1: Assessment of dynamic behaviour, Engineering Structures, 32, 2010, pp [5] CAETANO E., CUNHA A., MAGALHÃES F. and MOUTINHO C., Studies for controlling human-induced vibration of the Pedro e Inês footbridge, Portugal. Part 2: Implementation of tuned mass dampers, Engineering Structures, 32, 2010, pp [6] CAETANO E. and CUNHA A., Experimental and Numerical Assessment of the Dynamic Behaviour of a Stress-Ribbon Bridge, Structural Concrete, 5, No 1, 2004, pp [7] FURTADO R., QUINAZ C. and BASTOS R. New Braga Municipal Stadium, Braga, Structural Engineering International, Vol.15, No.2, 2005, pp [8] CAETANO E., CUNHA A. and MAGALHÃES F., Numerical and experimental studies of Braga sports stadium suspended roof, J. of Struct. Infrastructure Eng., Vol.6, Issue 6, 2010, pp [9] MAGALHÃES F., CUNHA A. and CAETANO E., Installation of a Continuous Dynamic Monitoring system at Braga Stadium Suspended Roof: initial results from automated modal analysis, EVACES 09, Wroclaw, Poland, [10] AMADOR S., MAGALHÃES F., CAETANO E. and CUNHA A., Analysis of the influence of environmental factors on modal properties of the Braga Stadium suspension roof, EVACES'2011, Varenna, Italy, 2011.