< Back to previous page

## Project

# Wind-Structure Interaction Simulations of Ovalling Vibrations in Silo Groups

With an increasing tendency in civil engineering to build ever taller skyscrapers and longer bridges with lighter materials, structures are more susceptible to wind-induced vibrations. In the design process, however, aeroelastic safety is often not considered. While the conceived structures are perfectly adopted to standard design loads, they are only seldom aerodynamically optimized. Many examples exist where wind-induced vibrations have led to catastrophic structural failure.

Wind-induced vibrations can be caused by either forced excitation or aeroelastic effects. In the first case, the forced vibrations originate fromturbulent fluctuations in the wind flow around the structure, e.g. fromnatural turbulence in the wind flow attacking the structure or due to periodic vortex shedding in the wake of the structure. The second, aeroelastic phenomena are typically self-excited and are due to displacements of the structure that result in interactions with the wind flow. It mustbe emphasized that it is often impossible to pinpoint one specific excitation mechanism as the single cause of observed wind-induced vibrationssince these are often initiated by a combination of different phenomena. In design codes, however, only forced excitation is mostly considered.It is therefore important to develop advanced techniques to study wind-induced vibrations that can incorporate all different excitation mechanisms at once.

To study the susceptibility of a structure to wind-induced phenomena, a coupled multiphysics problem considering both wind flow and structural dynamics has to be solved. Because of the complexity of such problems, simplified phenomenological and experimental models are often unsatisfactory while numerical methods have the advantage that all possibly interacting excitation mechanisms can be accounted forsimultaneously. With procedures available to couple a numerical model for the wind flow, i.e. computational fluid dynamics (CFD), and the structure, e.g. finite element (FE) models, it should be possible to study wind-structure interaction (WSI) numerically. The goal of this thesis is therefore to investigate whether for a realistic, complex WSI problem numerical simulations can adequately predict observed wind-induced vibrations. For this purpose, the specific case study of a silo group is considered where wind-induced ovalling vibrations were observed during a storm in October 2002 on several empty silos in a group of forty silos in the port of Antwerp (Belgium). The necessity of performing complicated numerical WSI simulations for this problem is examined.

First, it is investigated with both 2D and 3D CFD simulations how turbulent fluctuations in natural atmospheric winds can be simulated and how they are preserved in the flow field. In the 2D simulations realistic turbulence levels cannot be simulated in the oncoming wind flow but these calculations allow to study wake-induced effects on the aerodynamic pressures in the silo group for a changing angle of incidence of the wind flow, as the simulations are computationally significantly less intensive. In the 3D simulations, the turbulent wind flow can be modelled more realistically although some difficulties are encountered with the preservation of turbulence levels in the wind flow attacking the silo group as well.

Due to the particular geometry of the silo group and the highly turbulent character of the wind flow, validation of the CFD simulation results is difficult. In addition to the silo group, the flow around a singlesilo is therefore considered. The better documented 2D flow around a single cylinder is validated with results from the literature, while for the validation of the 3D wind flow simulations, a wind tunnel test is performed for the single silo configuration. The influence of the blockage effect in the experiment is investigated and the pressure distribution on the silo surface shows reasonably good agreement in experiment and simulation. Qualitative validation is furthermore performed wherever possible for the silo group by comparing the flow pattern with geometrically similar flows e.g. through tube bundles of heat exchangers or around surface mounted prisms.

The typical ovalling eigenmodes of a silo are determined from the FE model of the structure and subsequently, the modal projection of the aerodynamic pressure distributions from the 2D and 3D CFD simulations is determined. This approach allows to assess at which locations in the silo group ovalling vibrations are excited through direct forcing by the transient wind loads. For all angles of incidence in the 2D simulations, a harmonic decomposition methodology is introduced as an alternative for the 3D modal projection. The effect of a different turbulence modelling approach in the 2D and 3D simulations on theaerodynamic pressures on the silo surfaces is significant. While the 2Dsimulations are found to be inadequate for the present purposes, the location of the ovalling vibrations in the silo group is predicted well bythe 3D simulations.

Finally, the FE model of the structure and the 3D CFD wind flow simulations are considered together as asingle problem. To evaluate the importance of incorporating WSI for theprediction of ovalling vibrations, the structural response due to external aerodynamic forces is first calculated in one-way coupled simulations. The applied transient wind loads are determined a priori in the 3D CFD simulations. In this approach, only forced excitation is considered asa possible excitation mechanism of wind-induced vibrations. In the two-way coupled WSI simulations by contrast, the structural and fluid solverare fully interacting and information is exchanged at the interface between the two solvers in every time step. Therefore, in this approach aeroelastic phenomena are also accounted for. Both one-way and two-way coupled simulations are performed for the single silo configuration and for the entire silo group. The structural response in the different coupled simulations in terms of excited eigenmodes is compared by considering modal deformation energy. Different frequency components can be clearly distinguished in the modal deformation energy of the structural response which are related to physical phenomena in the turbulent wind flow.

After qualitative comparison of the results in the one-wayand two-way coupled simulations, it cannot decisively be concluded whether ovalling vibrations of a single silo are due to forced excitation orif aeroelastic effects are important. Because the coupled WSI simulations are at the limit of practical feasibility of present day computational power, no grid independent solutions could be obtained. Nevertheless, the results of low resolution simulations for the silo group arrangementindicate that aeroelastic effects have a sigificant impact on the structural response of the silos. It is therefore concluded that complicated and grid independent WSI simulations are in general required to realistically predict the onset of ovalling vibrations in silo groups with numerical techniques.

Wind-induced vibrations can be caused by either forced excitation or aeroelastic effects. In the first case, the forced vibrations originate fromturbulent fluctuations in the wind flow around the structure, e.g. fromnatural turbulence in the wind flow attacking the structure or due to periodic vortex shedding in the wake of the structure. The second, aeroelastic phenomena are typically self-excited and are due to displacements of the structure that result in interactions with the wind flow. It mustbe emphasized that it is often impossible to pinpoint one specific excitation mechanism as the single cause of observed wind-induced vibrationssince these are often initiated by a combination of different phenomena. In design codes, however, only forced excitation is mostly considered.It is therefore important to develop advanced techniques to study wind-induced vibrations that can incorporate all different excitation mechanisms at once.

To study the susceptibility of a structure to wind-induced phenomena, a coupled multiphysics problem considering both wind flow and structural dynamics has to be solved. Because of the complexity of such problems, simplified phenomenological and experimental models are often unsatisfactory while numerical methods have the advantage that all possibly interacting excitation mechanisms can be accounted forsimultaneously. With procedures available to couple a numerical model for the wind flow, i.e. computational fluid dynamics (CFD), and the structure, e.g. finite element (FE) models, it should be possible to study wind-structure interaction (WSI) numerically. The goal of this thesis is therefore to investigate whether for a realistic, complex WSI problem numerical simulations can adequately predict observed wind-induced vibrations. For this purpose, the specific case study of a silo group is considered where wind-induced ovalling vibrations were observed during a storm in October 2002 on several empty silos in a group of forty silos in the port of Antwerp (Belgium). The necessity of performing complicated numerical WSI simulations for this problem is examined.

First, it is investigated with both 2D and 3D CFD simulations how turbulent fluctuations in natural atmospheric winds can be simulated and how they are preserved in the flow field. In the 2D simulations realistic turbulence levels cannot be simulated in the oncoming wind flow but these calculations allow to study wake-induced effects on the aerodynamic pressures in the silo group for a changing angle of incidence of the wind flow, as the simulations are computationally significantly less intensive. In the 3D simulations, the turbulent wind flow can be modelled more realistically although some difficulties are encountered with the preservation of turbulence levels in the wind flow attacking the silo group as well.

Due to the particular geometry of the silo group and the highly turbulent character of the wind flow, validation of the CFD simulation results is difficult. In addition to the silo group, the flow around a singlesilo is therefore considered. The better documented 2D flow around a single cylinder is validated with results from the literature, while for the validation of the 3D wind flow simulations, a wind tunnel test is performed for the single silo configuration. The influence of the blockage effect in the experiment is investigated and the pressure distribution on the silo surface shows reasonably good agreement in experiment and simulation. Qualitative validation is furthermore performed wherever possible for the silo group by comparing the flow pattern with geometrically similar flows e.g. through tube bundles of heat exchangers or around surface mounted prisms.

The typical ovalling eigenmodes of a silo are determined from the FE model of the structure and subsequently, the modal projection of the aerodynamic pressure distributions from the 2D and 3D CFD simulations is determined. This approach allows to assess at which locations in the silo group ovalling vibrations are excited through direct forcing by the transient wind loads. For all angles of incidence in the 2D simulations, a harmonic decomposition methodology is introduced as an alternative for the 3D modal projection. The effect of a different turbulence modelling approach in the 2D and 3D simulations on theaerodynamic pressures on the silo surfaces is significant. While the 2Dsimulations are found to be inadequate for the present purposes, the location of the ovalling vibrations in the silo group is predicted well bythe 3D simulations.

Finally, the FE model of the structure and the 3D CFD wind flow simulations are considered together as asingle problem. To evaluate the importance of incorporating WSI for theprediction of ovalling vibrations, the structural response due to external aerodynamic forces is first calculated in one-way coupled simulations. The applied transient wind loads are determined a priori in the 3D CFD simulations. In this approach, only forced excitation is considered asa possible excitation mechanism of wind-induced vibrations. In the two-way coupled WSI simulations by contrast, the structural and fluid solverare fully interacting and information is exchanged at the interface between the two solvers in every time step. Therefore, in this approach aeroelastic phenomena are also accounted for. Both one-way and two-way coupled simulations are performed for the single silo configuration and for the entire silo group. The structural response in the different coupled simulations in terms of excited eigenmodes is compared by considering modal deformation energy. Different frequency components can be clearly distinguished in the modal deformation energy of the structural response which are related to physical phenomena in the turbulent wind flow.

After qualitative comparison of the results in the one-wayand two-way coupled simulations, it cannot decisively be concluded whether ovalling vibrations of a single silo are due to forced excitation orif aeroelastic effects are important. Because the coupled WSI simulations are at the limit of practical feasibility of present day computational power, no grid independent solutions could be obtained. Nevertheless, the results of low resolution simulations for the silo group arrangementindicate that aeroelastic effects have a sigificant impact on the structural response of the silos. It is therefore concluded that complicated and grid independent WSI simulations are in general required to realistically predict the onset of ovalling vibrations in silo groups with numerical techniques.

Date:1 Oct 2009
→
19 Sep 2013

Keywords:Structural dynamics, Fluid-structure interaction

Disciplines:Structural engineering, Other civil and building engineering

Project type:PhD project