Journal of Aerospace Engineering

This paper discusses the large eddy simulation technique as applied to vortex shedding from a cubic obstacle placed in a turbulent wind environment and presents results from a recent large-scale computation of this flow. The simulation was sufficiently resolved to capture the dynamics of the conical vortices on the roof of the obstacle and to predict the roof pressure footprint in good agreement with the experimental findings. The 3D visualizations of the vortex shedding and an analysis of the mean flow topology are also presented. The computations used a grid of ∼10⁷ nodes, and they were performed on a Cray T3E parallel machine. The turbulent inflow was taken from a separate precursor simulation targeted so that the obstacle represents a building placed in an urban wind environment.

The applicability of the k-ε model in the prediction of aerodynamic force and instability is investigated herein. To show that two-dimensional analysis by the k-ε model is different from ordinary two-dimensional analyses that simply neglect the spanwise velocity, an analysis of the aeroelastic vibration of a cross section with B/D = 2.0 is performed first using the latter method. In this case, motion-induced vortex oscillation was successfully simulated; however, galloping, in which the flapping motion of the separated shear layer plays an important role, could not be simulated. The result shows that physically reasonable flows can not be obtained by ordinary two-dimensional analyses, unless the spanwise momentum diffusion is incorporated correctly. On the other hand, the k-ε model, which incorporates this diffusion process by an eddy viscosity, enables two-dimensional analyses even in the high Reynolds number region. In this paper, applicability of the model is examined for rectangular cross sections with a wide range of B/D ratio, i.e., 0.6 ≤ B/D ≤ 8.0. Various typical aerodynamic features calculated using this model were found to be in good agreement with those obtained experimentally, particularly including discontinuities in Strouhal number at the critical cross sections of B/D = 2.8 and 6.0. Based on this result, an elastically supported B/D = 2.0 cylinder was analyzed. The motion-induced vortex oscillation and a coupling of the vortex-induced oscillation and galloping were successfully simulated, and their values were in good agreement with those measured in experiments conducted earlier.

For aerodynamic unstable oscillations, the dynamic properties of a structural system have a strong influence on its response behavior. The mass-damping parameter (the product of the mass ratio and the damping factor), known as the Scruton number, has usually been employed to evaluate the occurrence of aeroelastic instabilities. Buildings and structures have become very light of weight and susceptible to wind-induced oscillations. Consequently, we cannot suppose that the dominant frequency of response is almost equal to the natural frequency of the structural system. Taking into consideration that various kinds of unstable oscillations appear in the heaving oscillation of a rectangular cylinder, a variation in the mass ratio may cause peculiar behavior of the structure, even if the Scruton number is constant. Thus, we must clarify the dependency of various unstable oscillations on the Scruton number and the mass ratio. The purpose of this paper is to investigate the above issues using three-dimensional simulation of the fluid flow around an oscillating rectangular cylinder. The numerical method allows precise control and wide variation of the dynamic properties.

Physical and numerical simulations of the flow of atmospheric boundary layers around a number of building configurations, with pollutant releases from them, are described. Attention is concentrated on the issues that face the numerical modeler. Some inferences are drawn about the likely degree of uncertainty in the use of computational fluid dynamics for such cases, by using comparisons between the results obtained using different modeling choices, and between these results and experimental data. It is demonstrated that decisions concerning the precise details of the turbulence model may often be less important than those concerning details of the specifications of boundary conditions, the mesh construction, and the numerical schemes used to solve the modeled equations. All these factors lead to inevitable uncertainty about the quantitative accuracy of the calculation. The degree of uncertainty depends greatly on the particular quantity of interest and may or may not be significant in a practical case.

Wind tunnel tests, computational fluid dynamics simulations, and Gaussian models are employed for a dispersion study in nonuniform atmospheric boundary layer conditions with releases affected by neighboring building wake entrainment and topography. A vertical stack release and a horizontal crosswind release are subjected to both qualitative (plume trajectory and geometry) and quantitative (concentration) analysis. The results of the three modeling approaches (experiments, computational fluid dynamics simulations, and Gaussian models) are compared, their relative strengths are discussed, and conclusions are presented. It is suggested that a combined numerical (computational fluid dynamics) and physical (wind tunnel) modeling approach might benefit atmospheric dispersion studies in complex, nonuniform atmospheric environments.

For the efficient finite-element wind analysis, using the optimal mesh is one of the most important factors. The optimal mesh can be obtained when the errors of the solution are distributed uniformly over the entire domain. The study presents the development of a transition element for flow analysis, an element that has a variable number of midside nodes and can be effectively used in the adaptive mesh refinement by connecting the locally refined mesh to the existing coarse mesh through a minimum mesh modification. In the dynamic analysis of flow, the optimal mesh should be changed continuously in accordance with the changing error distribution; the proposed refinement/recovery scheme was found to be very effective for this purpose. The modified superconvergent patch recovery for the variable-node element is presented to estimate a posteriori error of the solution for the adaptive mesh refinement. The boundary conditions of the nodes generated by refinement process are different from those used in the ordinary finite-element method, in order to describe the singular point correctly. The numerical examples show that the optimal mesh for the finite-element analysis of flow around the structures can be obtained automatically by the proposed scheme.