Calibration of finite element models of concrete ... - ScienceDirect.com

Available online at www.sciencedirect.com Available online at www.sciencedirect.com

ScienceDirect ScienceDirect

Procedia Engineering 00 (2017) 000–000 Available online at www.sciencedirect.com Procedia Engineering 00 (2017) 000–000

ScienceDirect

www.elsevier.com/locate/procedia www.elsevier.com/locate/procedia

Procedia Engineering 199 (2017) 110–115

X International Conference on Structural Dynamics, EURODYN 2017 X International Conference on Structural Dynamics, EURODYN 2017

Calibration of finite element models of concrete arch-gravity dams Calibration of finite element models of arch-gravity dams using dynamical measures: theconcrete case of Ridracoli using dynamical measures: the case of Ridracoli Giulia Buffia,* , Piergiorgio Manciolaaa, Laura De Lorenzisbb, Nicola Cavalagliaa, Fabrizio Comodiniaa, a,* Giulia Buffi , Piergiorgio Laura De Lorenzis , Nicola Cavalagli , Fabrizio Comodini , c a d a , VittorioManciola Gusellaaa,,Marco Mezzi , Wolfgang Niemeier , Claudio Tamagnini Andrea Gambi c a d a Andrea Gambi , Vittorio Gusella , Marco Mezzi , Wolfgang Niemeier , Claudio Tamagnini a Department of Civil and Environmental Engineering, University of Perugia, Via G. Duranti, 06125 Perugia, Italy a Department of Civil and Environmental University of Perugia, ViaWeg, G. Duranti, 06125 Perugia, Italy Institute of Applied Mechanics, TechnischeEngineering, Universität Braunschweig, Bienroder 38106 Braunschweig, Germany b Institute of AppliedcRomagna Mechanics, Technische Bienroder Germany Acque SocietàUniversität delle Fonti Braunschweig, S.p.A., Piazza del Lavoro Weg, 4712238106 Forlì,Braunschweig, Italy d Acque Società delle Fonti S.p.A.,Braunschweig, Piazza del Lavoro 47122 Forlì,38106 Italy Braunschweig, Germany Institut für Geodäsie undcRomagna Photogrammetrie, Technische Universität Pockelsstraße, d Institut für Geodäsie und Photogrammetrie, Technische Universität Braunschweig, Pockelsstraße, 38106 Braunschweig, Germany b

Abstract Abstract Accurate and reliable predictions of the dynamic behaviour of dams is essential to ensure their correct management and the safety Accurate and reliable predictions thecontext, dynamic behaviour of dams is essential ensure their management and the safety of the downstream population. In of this structural monitoring and testing to procedures forcorrect their dynamic characterization are of the downstream population. In this context, structural monitoring and testing procedures for their dynamic characterization are essential tools for the calibration of numerical models of dams. This paper presents some results of an ongoing research program essential for the definition calibrationofofthe numerical models of dams. This paper of presents results ofdam: an ongoing research program aimed at tools an accurate geometric and structural properties a largesome arch-gravity the Ridracoli dam in the aimed at an accurate definition thefirst geometric structuralaproperties of a large arch-gravity the Ridracoli dam in has the Emilia-Romagna region, Italy. Inofthe part of and the research, detailed survey carried out by an dam: Unmanned Aerial Vehicle Emilia-Romagna region, Italy. In the part of the research, a detailed survey carriedThe outdense by anpoint Unmanned Vehicle allowed the detailed reconstruction of first the three-dimensional geometry of the structure. cloud, Aerial as provided by has the allowed the detailed reconstruction three-dimensional geometry of element the structure. dense point cloud, provided byrock the aerial survey, has been the base for of thethe definition of a high-fidelity finite model,The including the dam, theassurrounding aerial survey, has beenreconstruction the base for the definition of a high-fidelity element model, dam, the with surrounding rock mass, with a detailed of the site topography, and the finite reservoir water, whoseincluding dynamicthe interaction the structure mass, with aby detailed of the site topography, and reservoir water, whose dynamic interaction with the tests, structure is modelled meansreconstruction of acoustic elements. A large program ofthe structural monitoring, including a number of vibration has is modelled by means acousticdam elements. A large program of The structural monitoring, including a numberaccelerometers, of vibration tests, has been performed on the of Ridracoli during the last thirty years. dynamic monitoring system includes located been performed on the Ridracoli dam during the last thirty years. The dynamic monitoring system includes accelerometers, located in the structure and in the foundation rock mass, strain gauges and hydrodynamic pressure cells. The forced vibration tests were in the structure and in the foundation mass, strain hydrodynamic cells.characteristics The forced vibration carried out in correspondence to the rock maximum water gauges level, inand order to identify pressure the dynamic of the tests dam.were The carried out properties in correspondence the maximum water level, in order to identifybythe dynamicmodel characteristics ofwith the the dam. The mechanical of the damtomaterial and of the foundation rock are calibrated comparing predictions results mechanical properties oftests the dam and of the recordings foundation acquired rock are calibrated by comparing modelconsidering predictionsthe with the results obtained from vibration andmaterial from acceleration under recent seismic events, actual water obtained from vibration teststests. and from acceleration acquired seismic events, the actual of water levels registered during the The finite elementrecordings model obtained willunder allowrecent the simulation of theconsidering seismic performance the levels registered during theearthquakes. tests. The finite element model obtained will the simulation thevertical seismicjoints performance of the dam under different design The assessment of the effects of theallow reservoir level and ofofthe on the dynamic dam underofdifferent design earthquakes. The assessment of the effects of the reservoir level and of the vertical joints on the dynamic response the structure will be analysed. response of the structure will be analysed. © 2017 The Authors. Published by Elsevier Ltd. © 2017 2017 The Published Ltd. © The Authors. Authors. Published by by Elsevier Elsevier Ltd. committee of EURODYN 2017. Peer-review Peer-review under under responsibility responsibility of of the the organizing organizing committee of EURODYN 2017. Peer-review under responsibility of the organizing committee of EURODYN 2017. Ke Keywords: dam; UAV; FE model; dynamic analysis Ke Keywords: dam; UAV; FE model; dynamic analysis * Corresponding author. Tel.: 075-5853623; fax: 075-5853892. * Corresponding Tel.: 075-5853623; fax: 075-5853892. E-mail address:author. [email protected] E-mail address: [email protected] 1877-7058 © 2017 The Authors. Published by Elsevier Ltd. 1877-7058 2017responsibility The Authors. of Published by Elsevier Ltd. of EURODYN 2017. Peer-review©under the organizing committee Peer-review under responsibility of the organizing committee of EURODYN 2017.

1877-7058 © 2017 The Authors. Published by Elsevier Ltd. Peer-review under responsibility of the organizing committee of EURODYN 2017. 10.1016/j.proeng.2017.09.169

N.,Buffi Comodini Gambi A., Gusella V., Mezzi M.,110–115 Niemeier W., Tamagnini C. / Procedia111 2 Buffi G., Manciola P., De Lorenzis L., Cavalagli Giulia et al. F., / Procedia Engineering 199 (2017) Engineering00 (2017) 000–000

1. Introduction Primary aspects in dam management are the safety of downstream population and the efficient maintenance of the structure [1]. In fact, dam break events for large dams can have far-reaching catastrophic consequences, as demonstrated by documented case-histories and numerical studies [2]. Non-invasive techniques for the evaluation of the conservation state of the dam structure have been thoroughly investigated [3]. Under the assumption of linear behavior of the structure, modal analysis can be performed on dams by means of 2D or 3D Finite Element (FE) models to identify the natural frequencies and the corresponding mode shapes of the structure [4]. This information can then be used to identify the occurrence of a local degradation process [5], since a variation of the material properties over some portion of the structure can determine a corresponding change in the natural frequencies under the same dynamic input. For structures such as segmented arch dams built on valleys with complex topography, the detailed definition of the structure geometry – including the joints between the different segments – and of the foundation rock mass is of primary importance in defining the natural frequencies of the system. Moreover, specific attention must be paid to the mechanical response of expansion joints, as their stiffness affects significantly the overall response of the structure and the occurrence of inelastic sliding can cause internal forces redistribution under strong seismic events [6,7]. Therefore, a detailed geometrical reconstruction of the dam system can be very helpful in the definition of the dynamic properties of the dam and in the detection of their spatial variations. In this respect, the limited accessibility of large dams does not allow the use of traditional survey techniques, while the use of Unmanned Aerial Vehicles (UAV) is well suited for this purpose [8]. In particular, the “Structure from Motion” (SfM) technique can be used for the reconstruction of 3D objects from 2D images. However, the applications of such methodologies on large dams are still relatively scarce, and a reliable terrestrial survey is still necessary [9]. The present work aims at describing the preliminary calibration of a masonry dam FE model. Its geometry was derived from a UAV survey. The paper is organized as follows. Firstly, the case study of the Ridracoli dam is presented, the monitoring system of the dam and the UAV survey are illustrated. Secondly, the FE model construction is presented, paying attention to the joint, rock foundation and fluid-structure interaction modeling. Then the comparison of the 3D model with the vibration tests and with a seismic response is presented. Finally, the first results and conclusions are illustrated. Nomenclature E’ Drained Young’s modulus G’ Drained shear modulus Eu Undrained Young’s modulus Gu Undrained shear modulus Es Static elastic modulus

Ed Ν ρ B f

Dynamic elastic modulus Poisson’s coefficient Bulk density Bulk modulus Frequency

T H p

Period Dam high Prescribed pressure

2. The Ridracoli dam: monitoring system and UAV survey The Ridracoli dam is an arch-gravity dam in simple concrete (height 101 m and crest 432 m long) that closes a wide U-shaped valley in the center of Italy. The dam was completed in 1982 and it was completely filled for the first time in 1986. The dam body has a double curvature shape, symmetric with respect to the key section. It rests on a foundation saddle, which extends along the entire perimeter of the excavation. The body of the dam is not continuous, but divided in 27 vertical sections, which work as independent static elements in order to reduce the deformations within the dam body, during reservoir level changes, temperature variations or other actions, and to avoid cracking of the structure. The ashlars are connected by the insertion of neoprene elements that are disposed around the entire outer perimeter of the transverse section and of the tunnels that cross longitudinally the dam and the foundation. 2.1

Monitoring Systems

The control of the behaviour of the structure, the foundation rock, the reservoir banks and the downstream rocky slopes is achieved through a complex monitoring system activated since the construction phase. Most of the instruments and measuring points are located in five radial sections of the structure: the key section, the two external lateral ones and the two intermediate lateral ones [10]. Hydrostatic level, concrete, water and air temperatures, displacements, uplift pressures, deformations of concrete and rocks, and stresses in the dam body are periodically acquired and analysed by a dual control system, on-line and off-line. They verify independently the effective behaviour

Buffi G., Manciola P., De Lorenzis L., Cavalagli N., Comodini F.,/ Gambi A.,Engineering Gusella V., Mezzi M., Niemeier 112 Giulia Buffi et al. Procedia 199 (2017) 110–115W., Tamagnini C. / Procedia 3 Engineering00 (2017) 000–000

of the structure and make data available for comparison with models. The centralized acquisition of 284 instruments out of a total of 985 is performed in the guardhouse. The dynamic monitoring system is composed by a seismometer, accelerometers, strain gauges and pressure hydrodynamic capsules. The seismic motion at the foundation rock is recorded by an accelerometer located near the left side of the dam, at the same level of the crest, and in the foundation rock at 40 m depth. The structure response is monitored by two accelerometers placed near the foundation base and near the crest of the dam. The dynamic monitoring system has a seismometer that acquires all seismic events. It triggers the entire seismic monitoring system when the recorded acceleration at the crest exceeds 0.2 m/s 2 along the upstream-downstream direction. 2.2

Unmanned Aerial Vehicle (UAV) survey

The Ridracoli dam has been chosen as a benchmark to compare traditional surveying technologies and the innovative UAV photogrammetry technique [11]. The survey involved traditional tools such as Total Station, GPS Station and Laser Scanner to geo-reference and validate the dense point cloud by UAV. The comparison showed the accuracy of the UAV survey and its efficiency in the reconstruction of complex geometries. Available information from the original design blueprints have been used to integrate the UAV dense point cloud under the reservoir level and near the foundation base, where the access was impossible. Digital Elevation Model (DEM) of the ground, a bathimetry of the lake and some additional information gathered by UAV have been employed to reconstruct a highfidelity model of the topography in a large area around the dam. 3. FE Model of the Ridracoli dam The integrated dense point cloud is the base for the extraction of significant lines that can describe solid elements with different mechanical characteristics or boundary conditions (Fig.1a). The high accuracy of the survey allows the modeling of some important details such as the vertical joints, with their thickness (Fig.1b), and the dam body – foundation joint, the spillways, the stilling basin and the right and left sides of concrete blocks. The different aspects of the 3D FE model are discussed below. 3.1

Rock mass and reservoir modeling

The proper choice of the sizes of the rock and water volumes interacting with the dam under seismic conditions to be included in the FE model is related to the infinite reservoir condition for the basin and to the proper modeling of absorbing boundary conditions for the rock mass. The first condition requires that the area should be extended enough to ensure that the pressure waves generated by the fluid-structure interaction reaches the end of the basin in a time longer than the duration of the seismic event [12]. This condition is achieved by considering a distance of the fictitious water boundary from the dam approximately equal to 3÷4H, where H is the maximum height of the structure [13]. In the present work, such distance has been assumed equal to 500 m (5H). In the FE model, the rock mass boundaries have been placed along a vertical cylindrical surface at the same distance as the water boundaries, as shown in Fig.2a.

Fig. 1. (a) UAV survey of the Ridracoli dam structure, in white the significant lines that detect the vertical contraction joints; (b) solid elements of a vertical joint and of an ashlar.


3.2

Fluid-Structure Interaction (FSI)

The effect of the Fluid-Structure Interaction (FSI) on a dam cannot be neglected. It can be taken into account by simplified methods (i.e. Westergaard added mass approach), arbitrary Lagrangian Eulerian methods (i.e. Acoustic fluid, simultaneous or partitioned solution of solid and fluid domain), Eulerian fluid methods (i.e. Immersed boundary method), Lagrangian fluid methods or by multiphase flow methods (i. e. Combination of volume-of-fluid technique and immersed boundary formulation including or not the wave generation and propagation) [14,15]. In dynamic analysis, where the effect of pressure waves on the structure is of main interest and the fluid velocities are negligible, the acoustic equations are suitable to describe the dynamic behaviour of the reservoir water. In the FE model, FluidStructure Interaction (FSI) is modelled through acoustic fluid elements which allow the modeling of a liquid medium undergoing small pressure variations and its interface conditions. In these elements, the pressure is proportional to the (small) volumetric strain of the water [16]. The acoustic field is strongly dependent on the assumed boundary conditions. For dam reservoirs, some physical boundary conditions in acoustic analysis to simulate FSI are: • prescribed pressure p, in order to simulate a free surface with zero dynamic pressure (atmosphere); • acoustic-structural coupling, where the motion of an acoustic medium is directly coupled to the motion of a solid, i.e. dam structure-water interaction or rock mass-reservoir interaction (no bottom absorption); • impedance conditions, to simulate an acoustic-structural boundary with impedance, where the displacements are linearly coupled but not necessary equal. This last condition can be used to model rock mass-reservoir interactions with damping or radiation damping, to simulate a wave traveling outwards from the analysis region. 3.3

Interfaces and boundary conditions

The base of the FE model, placed at a depth of 150 m (1.5H) from the base of the dam, has been assumed as a rigid, perfectly rough boundary. The vertical contraction joints are modelled as independent solid elements (Fig.1b), connected to the adjacent ashlars of the dam and to the foundation base by surface-based “tie” constraints, that make the translational and rotational degrees of freedom equal for the pair of surfaces. The same interaction constraint is used to constraint the base of the dam (foundation) to the rock mass. 3.4

Mechanical characteristics and general aspects of modeling

The mechanical proprieties (Tab.1) of the concrete dam are taken from the technical reports describing the design of the structure. As for the rock mass properties, a large site investigation has been conducted in the dam area,

Fig. 2. (a) Global FE model of the Ridracoli dam; (b) detail of the FE model on the dam structure.

Buffi G., Manciola P., De Lorenzis L., Cavalagli N., Comodini F.,/ Gambi A.,Engineering Gusella V., Mezzi M., Niemeier 114 Giulia Buffi et al. Procedia 199 (2017) 110–115W., Tamagnini C. / Procedia 5 Engineering00 (2017) 000–000

including plate loading tests in exploratory tunnels and a series of geophysical investigations, aimed at measuring the compression and shear wave velocities of the rock [17]. In the FE model, the rock mass, as well as the concrete, has been assumed as homogeneous and linear elastic, based on the results of the site investigation. The elastic properties of the rock have been derived from the results of geophysical investigations, which provide the stiffness of the rock at very small strains – as appropriate for the modal analysis discussed in the following. Moreover, in order to analyze the dynamic properties of the dam structure, the mass density of the soil is assumed equal to zero. In the current FE modal analysis a dynamic modulus of elasticity for the concrete greater than the static one is used [18]. Finally, a bulk modulus B equal to 2.15∙106 kPa has been adopted for the reservoir water. The 3D FE model has been constructed using the commercial FE platform Abaqus Standard (Fig.2). Four-noded trilinear tetrahedral elements C3D4 are used to discretize the rock mass volume and the concrete structure. Acoustic four-noded linear tetrahedral elements AC3D4 are employed for the water reservoir volume. The mesh size adopted depends on the zone to be discretized: a relatively coarse discretization has been used for the rock mass, while a comparatively much finer mesh has been adopted for the dam body, to achieve a sufficient level of accuracy. Also the discretization employed for the vertical joints is finer than the one used for the ashlars of the dam (Fig.2b). Table 1. Mechanical properties of rock mass, concrete and of water. E’ (kPa)

Eu (kPa)

Es (kPa)

Ed (kPa)

G’ (kPa)

Gu (kPa)

ν

ρ (t/m3)

B (kPa)

Rock Mass

4.80∙10

2.17∙10

-

-

1.92∙10

8.70∙10

0.25

2.635

-

Concrete

-

-

3.07∙10

Water

-

-

-

7

7

7

3.97∙10

7

-

7

6

-

-

0.20

2.470

-

-

-

-

1.000

2.15∙106

4. Modal analysis and comparison with experimental results Every structural model aims at reproducing the real behavior through the numerical simulations. At the end of an iterative process upon which the model properties are changed to reduce the “distance” between predictions and measurements, a better understanding of the current conservation state of the structure can be achieved. In the following, the results of a preliminary calibration exercise of the Ridracoli dam model (in terms of dynamic properties) are briefly shown. A modal analysis has provided the dynamic proprieties of the model, in terms of natural frequencies and vibration modes. The dynamic structural identification can be performed using the excitation of the structure. The modalities can be different, the most reliable consists in the application of vibrators on the structure (no balanced rotating masses, hydraulic pistons.). The Ridracoli dam was tested with the vibrodyne in 1987. This procedure allows the generation of forces adjustable in both frequency or amplitude. The hydrostatic level during the test was 551.8 m a.s.l., close to that of maximum regulation. The response of the structure was acquired by seismometers applied in different sections and elevations on the dam body. A different, cheaper excitation source is represented by

Fig. 3. (a) Power Spectral Density (PSD) function of the recorded signal; (b) natural frequencies of the dam structure derived by the vibrodyne test, the finite element model and the power spectral density function.


environmental vibrations. In this case the signals can be affected by noises and some corrections are needed. For the comparison the seismic event of June 2011, during which the reservoir level was at 551.8 m a.s.l., has been also considered. The response of the structure was acquired by an accelerometer placed near the dam crest. To extract the experimental natural frequencies, the Power Spectral Density function is used (Fig.3a). As shown in Fig. 3 b, the natural frequencies extracted by the vibrodyne test (blue columns) and by the FE model (red columns) are similar, except for the higher modes (8,9,10) on which the fluid vibration effect is elevated and not distinguishable by the vibrodyne excitation. It might also be noted that the natural frequencies provided by the PSD function (green columns) are only partly identified. It is due to the signal noise, to the reduced structure acceleration related to the selected seismic event and probably to the accelerometer position, on the key cantilever close to the dam crowing. 5. Conclusions The high level of accuracy of the UAV survey allows the detailed geometry reconstruction of large, complex and difficult to access structures. This feature is thus well suited in the surveying of large dams. In this work, a high fidelity geometrical model for the Ridracoli dam has allowed the construction of a detailed 3D FE model which includes ancillary works (i.e. stilling basin, concrete blocks), the surrounding rock mass, and the reservoir water, modeled using acoustic elements. Significant structural details such as vertical contraction joints are also incorporated in the model. The results of a preliminary series of dynamic simulations have demonstrated that the model is capable of reproducing accurately the relevant natural frequencies of the real structure, measured independently from the interpretation of a vibrodyne test and of the observed response during the earthquake event of June 2011. Further studies are ongoing to evaluate the influence of the joints discontinuity as well as those of the reservoir level on the dynamic response of the structure, and of the non-linear behaviour of the concrete. References [1] W. H. Tang, B. C. Yen, Dam Safety Inspection Scheduling, J. of Hydraulic Engineering, 1991. [2] C. Biscarini, S. Di Francesco, P. Manciola, CFD modeling approach for dam break flow studies, Hydrol Earth SystSc 14 (2010) 705–718. [3] G. Hearn, R. B. Testa, Modal Analysis for Damage Detection in Structures, J. of Structural Engineering, 1991. [4] S. W. Alves, J. F. Hall, System identification of a concrete arch dam and calibration of its finite element model, Earthquake Engineering & Structural Dynamics 35 (2006) 1321–1337. [5] C. Ratcliffe, A Frequency and Curvature Based Experimental Method for Locating Damage in Structures, J. of Vibration and Acoustics, 2000. [6] M. T. Ahmadi, M. Izadinia, H. Bachmann, A discrete crack joint model for nonlinear dynamic analysis of concrete arch dam, Computers and Structures 79 (2001) 403-420. [7] J. R. Mays, L. H. Roehm, Effect of vertical contraction joints in concrete arch dams, Computers & Structures, 47 (1993), 4–5, 615-627. [8] A. Ellenberg, A. Kontsos, I. Bartoli, A. Pradhan, Masonry Crack Detection Application of an Unmanned Aerial Vehicle, International Conference on Computing in Civil and Building Engineering, 2014. [9] M. Naumann, M. Geist, R. Bill, F. Niemeyer, G. Grenzdörffer, Accuracy comparison of digital surface models created by unmanned aerial systems imagery and terrestial laser scanner, Remote Sensing and Spatial Information Sciences, XL-1/W2 (2013). [10] Consorzio Acque Forlì Ravenna, La diga di Ridracoli, Alpina Spa, 1985. [11] G. Buffi, P. Manciola, S. Grassi, M. Barberini, A. Gambi, Survey of the Ridracoli Dam: UAV – Based Photogrammetry and Traditional Topographic Techniques into the inspection of Vertical Structures, Geomatics, Natural Hazards and Risk, Submitted 02/2017. [12] R. Hellgren, Influence of Fluid Structure Interaction on a Concrete Dam during Seismic Excitation-Parametric analyses of an Arch DamReservoir-Foundation system, 2014. [13] B. Sevim, A. C. Altunşik, A. Bayraktar, M. Akköse, Y. Calayir, Water Length and Height Effects on the Earthquake Behavior of Arch DamReservoir-Foundation SystemsBaris¸ Structural Engineering, 2011. [14] G. Hou, J. Wang, A. Layton, Numerical Methods for Fluid-Structure Interaction – A Review. Comp. Physics, 12 (2012) 2, 337-377. [15] C. Rydell, T. Gasch, F. Facciolo, D. Eriksson, R. Malm, Interaction between structure and water in seismic analysis of nuclear facilities. International Association for Structural Mechanics in Reactor Technology, SMiRT 22 Conference 2013, San Francisco, USA. [16] Abaqus Documentation, 2012. [17] G. Oberti, F. Bavestrello, P. Rossi, F. Flamigni, Rock mechanics investigations, design and construction of the Ridracoli dam, Rock Mechanics and Rock Engineering, 19 (1986), 3, 113–142. [18] Neville, A. M., 1963. “Properties of Concrete”.