Research On The Turbulent Inflow For Large Eddy Simulation Of Fluctuating Wind Loads

Most engineering structures stand within atmosphere boundary layer, the near ground wind field and the flow around buildings could be recognized as incompressible turbulence. Considering the turbulent incoming flow and the time-dependent pressure on structures, large eddy simulation has become the reasonable choice of numerical wind resistant research.One preconditions of large eddy simulation in wind engineering is the turbulent inflows, which should have the properties of near ground wind field. Compared with other subjects, the turbulent inflow for wind engineering has its own properties, such as high Reynolds number, high turbulent intensity and differs with each other when developed on different terrain. At present, many wind engineering research based on large eddy simulation use laminar inflow, or use the improper turbulent inflows which do not have the properties of wind field, this may lead incorrect results. To get the inflow condition for wind engineering, the near ground wind field is preliminary simulated in this thesis, velocity and pressure time history are recorded and then set as the inflow condition for fluctuating pressure simulation. There are some difficulties in this procedure:1.The flow's Reynolds number of wind field is very high, so a stable and effective code is necessary when carry out the procedure.2.Simulation of fully developed atmosphere boundary layer needs long computational domain, which is very consumptive, so effective simulation methods should be established.3.Flow properties in different terrains differ with each other, several simplified model should be established to simulate wind field in different terrains.4.The grid system and Reynolds number of wind field simulation often different with that of building pressure simulation, so preserving the properties of turbulent inflow in the later simulation domain can not be ignored.The corresponding solution methods to these difficulties were proposed in this thesis, then the turbulent inflows which have the properties of wind field were successfully obtained and used for fluctuating pressure simulation; the work carried out in each chapter could be summarized as: In Chapter 2, Equations which describe fluid flows were elaborated and non-dimensionalized. A large eddy simulation code was compiled on Cartesian gird using finite differencing method. To simulate high Reynolds number flow, a third order upwind scheme which could adjust numerical viscosity with flow condition was developed for convective term. The one dimensional convective-diffusion equation was used to validate the upwind scheme in Chapter 3, and the results shows the upwind scheme have good accuracy and can suppress numerical oscillation in high convective-diffusion process. In addition, the two dimensional lid-driven flow and the three dimensional flow around rectangular were used to validate the salability and accuracy of the code.In Chapter 4, to simulate the fully developed atmosphere boundary layer efficiently, a streamwise periodic boundary condition was used to decrease the computational domain and CPU time. This periodic condition may induce mean velocity profile unstable, so a constant power law-distributed pressure gradient was proposed and introduced into NS equation to keep the profile stable. Because the structural similarity between flat plate boundary layer and atmosphere boundary layer, the flow developed on flat plate boundary layer was appreciatively used as near-ground wind flow. Streamwise periodic condition was used to simulate a flat plate boundary layer, and the results shows: velocity profile of the boundary layer is similar with wind profile on A-type terrain of China; velocity spectrum is consistent with Karman spectrum in low-frequency region, and has discrepancy in high-frequency region, finer mesh can improve the high-frequency quality of velocity history. Velocity and pressure histories in the boundary layer were recorded, and then introduced as turbulent inflow into a validation section, which has a "virtual obstacle", to validate the effectiveness of the inflow. The result shows, turbulent inflow adjusts well when mesh and Reynolds number change, and statistics quantities in validation section can be accepted by wind engineering.In Chapter 5, Numerical roughness and bars were introduced into boundary layer to increase the turbulent intensity. A quasi-periodic condition was developed into rough wall boundary layer simulation to maintain the velocity profile. The simplified models of B-type, C-type and D-type terrain of China were simulated. Compared with wind standards of Chinese, Australian and American, mean velocity profiles agree well with that recommended by Chinese standard, and turbulent intensity profiles are between the values recommended by Chinese and Australian standard, but much smaller than the value of American standard. Velocity spectrum is still consistent with Karman spectrum in low-frequency region, while has discrepancy in high-frequency region. Velocity and pressure histories were recorded and then introduced as turbulent inflow into the validation section, to validate the effectiveness of inflow when mesh and Reynolds number change. The result shows that mean velocity profiles are well preserved, and the turbulent intensities reduce slightly because of the disappearance of numerical roughness and bars, the reduction is especially obvious in outer region of boundary layer. Meanwhile, turbulent intensities are basically not influenced by the change of Reynolds number; velocity spectrums in the validation section agree well with those of turbulent inflows, and are not influenced by Reynolds numbers. Compared with the inflow, integral length scale in the validation section changes slightly when mesh grid change.In Chapter 6, three turbulent inflows with different turbulent intensities were generated on flat wall and rough walls boundary layer, respectively, and then they were introduced into the pressure simulation of flat roof building. Compared with wind tunnel tests, satisfied mean and rms pressure coefficient were obtained. Reynolds number and turbulent intensity are the two key factors which influence the distribution and value of pressure on bluff body shaped buildings. When Reynolds number becomes greater, the mean and rms pressure coefficients in separate regions, i.e. roof and side walls, increase slightly. With the increase of inflow's turbulent intensity, the suction pressure regions on the roof and side walls move towards to the leading edge, and the values become larger as well, and rms coefficients on the roof and side walls are become larger when inflow turbulent intensity increases. Influenced by the characteristic turbulence induced by bluff body building, the pressure histories on the roof show the Non-Gaussian property, especially in flow separate regions, i.e., leading edge and side edges, this phenomenon basically agrees with wind tunnel test. In addition, gust wind loads on the roof were predicted using gust factor method, and higher gust factor is recommended when pressure histories show Non-Gaussian property.