PSI - Issue 3

Alberto Lorenzon et al. / Procedia Structural Integrity 3 (2017) 370–379 A. Lorenzon et al. / Structural Integrity Procedia 00 (2017) 000–000

376

7

Where U (t) is the vector of wind velocities at the nodal position, C D is the drag coefficient, ρ a is the air density and A i is the area of influence for each node. The vector of wind velocities can be written as the sum of a mean and a turbulent component. While the mean component is assumed to be constant in space and time, the turbulent component is simulated as a mono-dimensional zero-mean Gaussian stochastic process whose spectrum well approximates the theoretical one. A series of numerical simulations were then performed using a tridimensional geometric nonlinear finite element model and the stress cycles histograms for each simulation were evaluated using rain-flow cycle-counting methods. 3.3. Large roofs and stadiums In Flamand et al. (1996), the authors performed fatigue analyses for the design of a large steel stadium exposed to wind. The fatigue damage was calculated in time domain. A boundary layer wind tunnel provided the time series of the pressures on a 1:200 scaled fixed model of a steel stadium. Using a FEM model, the calculation of stress time series due to quasi-static and resonant wind components was then performed. A rain-flow counting method was then used to build the fatigue load spectrum and the total damage was evaluated using Miners law. In order to calculate the number of cycles for each stress group, the authors used site wind data measured over 30 years to calculate the probability of occurrence of every wind speed. 4. CFD applications to civil structures in literature On the basis of literature research on fatigue analysis of steel structures invested by the wind, it is clear that the wind tunnels are often used to obtain information on the effect of the wind on the structure. Such information consists on the aerodynamic and aeroelastic parameters in the case of analysis based in the frequency domain and in the case of analysis based in the time domain with a synthetic time series. In case of analysis based in the time domain with “real” time histories they consist in the time series of the pressures acting on the scaled model. Reported below are some examples from (the very numerous) CFD simulations in literature regarding large steel structures with the aim of observing whether such methods are able to provide the information necessary to perform fatigue analysis with one of the methods above. 4.1. High-rise, slender buildings - In Huang et al. (2007) the authors conducted a comprehensive numerical study of wind effects on CAARC standard tall building. The study used many different CFD turbulence models such as LES, RANS (many RANS closure models were used). The most advanced RANS model used in this test (namely the MMK model) was able to provide a good estimate of all pressure coefficients in most of the cases. The LES method gave results adherent to the results of experimental tests in the wind tunnel as well. It is noted in addition that the LES method is able to capture all effects of the irregular flow while the MMK model is not able to predict a part of the irregular motion. The difference between the two time histories is visible in Figure 1. The authors also provided information about CPU time showing that LES analyses required far more computational resources than RANS. - In Elshaer et al. (2016) the authors investigated the aerodynamic response of the on CAARC standard tall building using LES. In addition to the study of Huang and Li et al., a flow generator was implemented that allowed the introduction of an inflow boundary condition that satisfies the proper turbulence spectra and coherency. The pressures are in very good agreement with those from wind tunnels. The power spectral density obtained from the LES using the time history base moments match reasonably with the experimental measurements. - In Tominaga et al. (2008) the authors compared CFD results using various RANS and LES models applied to a high-rise building. Among the RANS models, Durbin’s v 2 -f has shown the best agreement with the experiment. LES model provided better results but the authors remark the influence of the inflow turbulence on the quality of the results.