Вы находитесь на странице: 1из 6

Numerical Methods in Geotechnical Engineering Benz & Nordal (eds) 2010 Taylor & Francis Group, London, ISBN

BN 978-0-415-59239-0

Caisson movement caused by wave slamminga comparison of ABAQUS and FLAC analyses
L. Andersen, H.F. Burcharth & T. Lykke Andersen
Department of Civil Engineering, Aalborg University, Aalborg, Denmark

A.H. Augustesen
Department of Civil Engineering, Aalborg University, Aalborg, Denmark COWI A/S, Aalborg, Denmark

ABSTRACT: During wave slamming, caisson movement may occur as a combination of sliding along the caissonfoundation interface and local failure in the foundation and seabed. The paper presents a comparison between different techniques applied to the analysis of this movement. Thus, a finite-difference analysis has been performed by means of the commercial code FLAC. Similarly, ABAQUS has been employed for finiteelement analyses based on linear as well as quadratic spatial interpolations, assuming fully drained conditions and utilizing an elasticplastic model for the rubble foundation and the seabed. The results of the numerical codes are compared to an analytical solution in which the deformation of the subsoil has been disregarded.

INTRODUCTION

Wave slamming on a caisson breakwater may lead to movement of the structure. As shown in Figure 1, irreversible deformations may occur in different forms including sliding at the interface between the structure and the foundation, slip failure between the foundation and the seabed, and local collapse of the quarry rock constituting the foundation. Typically, a static design is carried out for breakwater. However, this demands a static pressure distribution equivalent to the dynamic pressure stemming from the wave. As indicated by the pressure distribution sketched in Figure 2 and the example results from measurements of the horizontal force resultant shown Figure 3, defining a proper level of this equivalent static load may not be possible. Thus, shock-like peaks exist in the pressure due to wave slamming, and different sampling frequencies in model tests may lead to very different conclusions regarding the design. Hence, a dynamic model is preferred. Burcharth et al. (2008) proposed a simple model based on the one-dimensional equation of motion. Due to its simplicity, the model can be used for lifetime analysis of a caisson; but it only accounts for sliding along the interface between the caisson and the rubble foundation. A more realistic modelling of the subsoil can be achieved by utilization of numerical methods. Employing a nonlinear finite-element model, Barqun (1998) included a visco-elastic soil model, and plastic deformation of the foundation and seabed was considered by Burcharth et al. (2009) as well as Kudella & Oumeraci (2009). However, the reliability of such numerical models must be verified

Figure 1. Failure modes for a caisson subjected to wave impact.

before their application in design. Hence, based on the commercial codes ABAQUS and FLAC, finiteelement and finite-difference solutions are compared with focus on their ability to quantify the displacement of a caisson due to horizontal sliding. The numerical models both employ a Lagrangian formulation accounting for elastic and plastic material behaviour as well as geometrical nonlinearity. Finally, a comparison is made with the result of the solution proposed by Burcharth et al. (2008).

COMPUTATIONAL MODELS

The overall geometry of the vertical breakwater is defined in Figure 4. The wall on top of the caisson has a width of 5 m at the base and 2.5 m at the top. The caisson is assumed to consist of reinforced concrete backfilled with sand. It has been found that the stress levels will not lead to failure in the concrete. Hence, the structure is modelled as a linear elastic material with the properties listed in Figure 4, where E, and

895

Figure 2. Dynamic part of the hydraulic pressure distribution from wave slamming on a caisson.

Figure 5. Simplified distribution of the hydraulic pressure on the caisson during wave slamming.

Figure 3. Horizontal force resultant from wave slamming on caisson measured in model scale at Aalborg University.

Figure 4. Geometry and material properties of the considered caisson, rubble foundation and seabed.

signify Youngs modulus, Poissons ratio and the total mass density, respectively. The foundation and the seabed are assumed to consist of quarry rock and sand, respectively. Linear response is assumed in the elastic regime, and the Mohr-Coulomb criterion is employed to identify stresses at initial yielding and strains during plastic deformation. The material properties are given in Figure 4, where and are the angles of internal friction and dilation, respectively, whereas c is the cohesion. The latter has been introduced to increase stability of the computational methods. Finally, an interface exists between the caisson and the foundation. Sliding occurs along this interface in accordance to Coulombs friction law. In the present case, a friction coefficient of = 0.6 is assumed (see Figure 4) corresponding to an interface friction angle

of 31 . Initially, contact is assumed between the structure and the foundation. However, during the dynamic response slip is allowed to happen in the numerical models, i.e. the tensile strength and stiffness of the interface are both zero. Rayleigh damping is applied, since it is available in ABAQUS as well as FLAC. The damping ratio = 5% is assumed at the frequency 1 Hz and only mass proportional damping is used. Added mass is disregarded, and the water (including the pore fluid in the foundation and subsoil) has not been modelled directly. Instead, a pressure is applied on the caisson surface.The transient part of this pressure is divided into two components as illustrated in Figure 5. The load time histories assumed for the components P1 (t) and P2 (t) are shown in Figure 6. Thus, further to the buoyancy provided by the hydrostatic pressure P0 , a quasi-static pressure, P1 (t), is applied with a uniform distribution on the front side of the caisson and a triangular distribution on its base. The latter part of the load is assumed to act in phase with the load on the front side whereas in reality a short delay occurs due to a finite velocity of the pressure wave travelling through the pore fluid. Finally, wave slamming is modelled by a shock load applied as a pressure, P2 (t), distributed uniformly on the upper half of the front side. The simplified model should be compared to the measured wave load reported in Figure 3 and the pressure distribution illustrated in Figure 2. In the present simulations, the maximum value of max P1 (t) is given by Pstatic /P1 = 2, where Pstatic is the value of P1 leading to sliding failure when applied statically in the absence of the shock load P2 . Based on Figure 4, the value Pstatic = 98 kPa is determined, provided a hydrostatic pressure of P0 = 10 kPa at the base of the caisson, i.e. at a water depth of 10 m. In order to study the influence of the peak height as well as the peak width of the shock load, four wave load time histories are analysed and compared. Thus, as listed in Figure 6, the short-term excitation from wave slamming has a duration of t2 = 0.2 or 0.4 s with the peak occurring after t1 = 0.1 or 0.2 s. The peak max max height is defined by P2 /P1 = 4 or 8. 2.1 ABAQUS models Two finite-element (FE) models are analysed by ABAQUS (Simulia 2009). Plane strain is assumed, utilizing four and eight-node quadrilateral elements with linear and quadratic spatial interpolation of the

896

Figure 6. Time history of the two components constituting the simplified wave load on the caisson.

Figure 8. Deformations (scaled 50 times) at ultimate failure in the ABAQUS models using linear (top) and quadratic (bottom) spatial interpolation.

Figure 7. Discretization in the ABAQUS models using linear (top) and quadratic (bottom) spatial interpolation. The dark, medium and light shades of grey indicate concrete, rubble and sand, respectively. The soil domain is 120 m 40 m.

displacement field, respectively. Figure 7 shows the mesh in the two models. Thus, the node distance is the same, whereas the mesh size is doubled in the second-order elements compared to the linear model. In order to establish the in situ stresses in the structure, foundation and subsoil, gravity is applied incrementally and stepwise. The full mass density is employed and the gravitational acceleration is set to g = 5 m/s2 in order to compensate for uplift on the soil skeleton from the water. This ensures a correct computation of the effective stresses as well as the dynamic response. Full gravity with g = 9.82 m/s2 is applied on the caisson and buoyancy is introduced by a pressure of P0 = 10 kPa as discussed above. The FE models are utilized to identify the failure mode in static loading. Horizontal fixities introduced

at the vertical edges of the soil domain, whereas vertical and horizontal fixities are employed along the bottom of the computational model. As illustrated in Figure 8, static failure occurs as a combination of sliding at the interface and local soil collapse at the heel of the caisson. The local deformations behind the heel are slightly different in the models with linear and quadratic spatial interpolation. However, the ultimate value of P1 is found as 95 kPa in either model, i.e. a value below the capacity related to pure translational sliding, Pstatic . Further, it has found that a reduction of the foundation height by a few metres leads to foundation slip failure at a pressure of the same magnitude. Thus the failure modes shown in Figure 1 occur at similar levels of the load. For this reason, the value Pstatic = 98 kPa is used in the dynamic analysis which is performed by ABAQUS/Standard, i.e. an implicit Lagrangian solution scheme is employed. 2.2 FLAC model A finite-difference method solution is computed by the commercial code FLAC3D (Itasca 2007). A 1 m wide strip is constrained in the y-direction to simulate plane strain. The model employs the discretization used in the linear FE model, i.e. the grid shown in Figure 7. However, the cells are not regarded as elements. Instead, the terminology zone is introduced for the space between four adjacent nodes. The volumes and edges of the zones are utilized for application of external forces and boundary conditions as well as evaluation of strains and stresses.

897

Figure 9. Vertical normal stresses due to gravity in the FLAC model. The dark and light shades of grey indicate high and low pressure, respectively.

Similarly to the FE model, gravity is applied incrementally. However, since FLAC builds on an explicit dynamic solver, a ramp function is introduced such that the load is increased linearly over time slowly enough to allow static equilibrium between exterior and interior forces to develop. Alternatively, the entire gravitational load is applied at once and time integration is performed until static equilibrium is reached, i.e. the correct stress distribution has been established. In this quasi-static part of the solution, mass scaling is employed to allow the use of larger time increments, leading to a faster solution. Figure 9 shows the distribution of the vertical normal stresses at the end of the gravitational step and after introduction of buoyancy on the caisson, i.e. just before the application of the transient load. The discontinuity of the stresses that can be observed at the interface between the structure and the foundation stems from the hydrostatic pressure applied at the base of the caisson. The ultimate capacity for static loading cannot be identified by application of a forced displacement, since the kinematics of the structure and subsoil at failure are unknown. It has not been possible to determine the ultimate capacity of the caisson for a static load distributed as P1 in Figure 5. However, the main purpose is to find the magnitude of the displacement as well as the deformation mode during wave impact. Thus, only the dynamic analysis has been carried out, based on the explicit solver with the true mass lumped at the nodes.

RESULTS AND DISCUSSION

Figure 10 shows the final deformations obtained in the FLAC simulations. A similar deformation is obtained in the ABAQUS simulations and will not be shown. max max For all combinations of t1 and P2 /P1 , a combination of the three failure modes illustrated in Figure 1 can be observed. Thus, significant plastic deformation occurs under the heel of the caisson due to the localized stresses stemming from the high contact pressure evolving during wave impact. Further, foundation slip failure clearly occurs at some stage during

Figure 10. Final deformation in the FLAC models. Dark and light shades of grey indicate high and low magnitudes of the strain, respectively. The displacement is scaled by a factor 10.

the response. This is different from the response to static loading where failure occurs as a combination of local soil collapse and sliding. Figure 11 shows the results of the ABAQUS and FLAC models for the different combinations of t1 and max max P2 /P1 . The horizontal displacements at the top left

898

Figure 11. Time history of the horizontal displacements at the base and top corners on the front side of the caisson. The peak max max of the wave-impact force occurs at the time t1 and the peak height is P1 + P2 . The grey lines indicate the horizontal sliding obtained by the simple analytical solution proposed by Burcharth et al. (2008).

corner (caisson top) and the lower left corner (caisson base) are plotted as functions of time. A comparison is made with the simple analytical solution proposed by Burcharth et al. (2008). Since the analytical solution only concerns the horizontal translation, it provides the same displacement at the top and base of the caisson. A number of observations can be made by inspection of Figure 11. Firstly, significantly different magnitudes of the horizontal displacement are predicted by the finite-difference and finite-element models with FLAC providing the smaller value. Thus, ABAQUS calculates a final displacement that is about a factor two higher than the FLAC model when quadratic spatial interpolation is employed. In the case of linear interpolation, the displacements obtained by the FE model are even greater. On the other hand the difference between the displacements at the caisson top and base is nearly the same in all models, suggesting that the rotation of the caisson is not influenced by the choice of model. The displacements obtained by the numerical models at the base of the caisson should be compared to the analytical solution indicated by the grey shaded line in Figure 11. Evidently, the simple model predicts a displacement that is much smaller than the values provided by ABAQUS for all considered combinations of

the peak duration and magnitude. In most cases, the analytical solution is also far below the displacement given by FLAC. The difference is most pronounced max max for t1 = 0.1 s and P2 /P1 = 4. Here the analytical approach provides a displacement smaller than 5 mm in contrast to FLAC which predicts a displacement of approximately 2 cm. The difference between the analytical and numerical results becomes less significant with increasing magnitude of the shock load. Especially for t1 = 0.2 s max max and P2 /P1 = 8 the results of the simple method and FLAC are close to being identical. This may be explained by the fact that all failure modes indicated in Figure 1 are active in the numerical models, whereas only sliding is included in the simple model. Thus, Figure 10 shows that plastic deformations occur at the foundationsubsoil interface for all load histories even for the relatively short shock load with low max max magnitude (t1 = 0.1 s and P2 /P1 = 4). Apparently for wave loads with a high peak value, sliding along the structurefoundation interface becomes the dominant mode of deformation and the simple model is more accurate than in the case of low-magnitude forces. Slightly different formulations of the contact at the structurefoundation interface in ABAQUS and FLAC may cause a significant part of the deviation between

899

the results provided by the two codes. In ABAQUS a masterslave definition is employed for the interface with the top of the foundation acting as master surface and the base of the caisson acting as the slave. Contact is then identified when a node from the slave surface lies on or passes the element edges on the master surface. Pressure over-closure is penalized by a spring with the stiffness 1010 N/m/m2 acting in the normal direction; but no elastic deformation can occur in the tangential direction. In FLAC, so-called interface elements are applied at the base of the caisson. Contact is established whenever a node from another domain (in this case the foundation) hits these interface elements. Then linear elastic springs counteract penetration in the normal direction and sliding or shear in the tangential direction. In the present analyses, the stiffness 1011 N/m/m2 has been utilized for both stiffnesses. It has been checked that there is no significant change in the results if the interface elements are instead applied on the top of the foundation or if the spring stiffnesses are changed by one order of magnitude. Another explanation for the deviation between the solutions obtained by FLAC and ABAQUS may be a poor degree of convergence. For this reason, a FLAC computation has been carried out with one-and-a-half times as many nodes in each direction. The result is nearly identical to those presented in Figure 11, indicating that the FLAC model is fully converged. On the other hand, an ABAQUS model using linear interpolation and a mesh size that is two thirds of the original mesh size provides the same elastic response but slightly higher displacements at failure. This indicates that the FE solution does not have the same degree of convergence. Furthermore, there is a great influence of the element type (interpolation order) on the outcome. Figure 8 brings some insight into this by clearly demonstrating that the quadratic and linear elements treat contact at the heel of the caisson in very different manners. 4 CONCLUSIONS

predict more displacement than a simple solution only accounting for sliding along the structurefoundation interface, since a combination of foundation slip failure, sliding and local soil deformation under the heel occurs in the refined models. Given that a consistent and fully converged result has been obtained with FLAC, this code is proposed for further analyses. However, a comparison with ABAQUS/Explicit simulations and results based on adaptive mesh refinement within the finite-element analysis may be relevant. The results presented in this paper are based on a model of finite extent. In ABAQUS, semi-infinite elements are available and FLAC allows the use of absorbing boundary conditions for dynamic analysis. In both cases, the idea is to provide boundary conditions that more realistically model the behaviour of unbounded soil. This will be considered in future analyses. Furthermore, it has been assumed that the soil is fully drained and the dynamic pressure on the base of the caisson develops immediately when the wave plunges on the front of the structure. This may not be realisticin particular not when the foundation and seabed consist of materials with low permeability. Hence, the next step is to include a dynamic pore pressure model in the numerical simulations. Finally, accumulated displacement after a series of wave impacts should be analysed, possibly using a model that accounts for liquefaction of the seabed.

REFERENCES
Barqun, G.G. 1998. Dynamic analysis of a vertical breakwaterextension to the Escombreras Basin Port of Cartagena, Spain. Madrid. Burcharth, H.F., Andersen, L. & Lykke Andersen, T. 2009. Analyses of stability of caisson breakwaters on rubble foundation exposed to impulsive loads. In Smith, J.M. (ed.), Proc. 31st Int. Conf. Coastal Eng., Hamburg, Germany, 31 Aug. 5 Sep. 2008: 36063618. World Scientific. Burcharth, H.F., Lykke Andersen, T. & Meinert, P. 2008. The Importance of Pressure Sampling Frequency in Models for Determination of Critical Wave Loadings on Monolithic Structures. In Proc. COPEDEC VII, Dubai, UAE. Itasca 2006. FLAD3D Fast Lagrangian Analysis of Continua in 3 Dimensions Users Guide. Minneapolis, Minnesota USA: Itasca Consulting Group, Inc. Kudella, M. & Oumeraci, H. 2009. Experimental and numerical study of the response of a sandbed beneath a caisson breakwater subject to cyclic wave load. In Smith, J.M. (ed.), Proc. 31st Int. Conf. Coastal Eng., Hamburg, Germany, 31 Aug. 5 Sep. 2008: 36193631. World Scientific. Simulia 2009. ABAQUS Version 6.9 Documentation. Providence, RI, USA: Dassault Systmes Simulia Corp.

Caisson movements due to shock loads from wave impact have been analysed by the finite-element code ABAQUS and the explicit finite-difference solver FLAC3D under the assumption of plane strain and elasticplastic response of the foundation and subsoil. It has been found that linear and quadratic interpolation in ABAQUS lead to a significant difference in the total displacements regardless of the load magnitude and load duration. The horizontal displacements achieved by the ABAQUS models are greater than the displacements obtained by FLAC. Both codes

900

Вам также может понравиться