High frequency pressure fluctuations on a two dimensional square cylinder in smooth and turbulent flows

• Main Author: Vasanji, Zainulabdeen
• Language: English
• Published: 2010
• Online Access: http://hdl.handle.net/2429/21168

There is a need to develop statistical design procedures for glass and cladding panels in buildings exposed to strong winds. Account must be taken of the high frequency, high amplitude pressure fluctuations observed occassionally on full scale buildings and building models in simulated atmospheric flows, and the non-Gaussian probability distribution of these pressure amplitudes. Most probability distributions of the pressure fluctuations can be modelled over the low-pressure range where the danger of damage is the highest, by a decaying exponential curve of the form F(x) = 1 – e[sup –ax[sup b]] Using this type of approximation, which finally rests on the determination of the two empirical statistical parameters "a" and "b", plus the usually measured mean and root-mean square pressure coefficients, the numerical risk of occurrence of any specified low pressure can be predicted for a particular time interval over which the high winds persist. The simple analysis developed here also requires as an input a non-dimensional Strouhal number which relates the length scale (panel size) and velocity scale (reference velocity) to the highest frequency of pressure fluctuation which is likely to affect the panel in Question. Wind tunnel measurements have been made of the pressure fluctuations on a simplified building model, a square prismatic cylinder, in smooth and turbulent uniform flows. Spectra and probability distributions have been obtained for various locations at various angles of the incident wind, where possible glass and cladding failure could occur. Careful calibration of the pressure measuring system ensured that a broad frequency range of pressure fluctuations was sensed. The measured probability distributions of the pressure fluctuations have been fitted by a Gaussian model for locations where the mean pressure coefficients are positive-going, and by a curve of the exponential type over other locations where the mean-pressure coefficients are definitely negative. The two statistical parameters required to fit the distributions in the latter case do not correlate well with the measured mean pressure coefficients. Once empirical values of the two parameters are obtained however, from experience or test, design of building elements to include the extreme values of negative (suction) pressures can be readily accomplished. Examples are given to demonstrate this procedure. === Applied Science, Faculty of === Mechanical Engineering, Department of === Graduate