While designing metal buildings for wind drift, for simplicity of analysis and design, connection at base of column is considered as pinned which provides no rotational restraint. The actual behavior of the connection however, is partially rigid, that provides some rotational stiffness even in case of single row of bolts. Moreover, using a two-dimensional (planar) structural model for analysis ignores any load distribution provided by roof and wall sheeting. Simulation of true behavior of base connection and diaphragm stiffness can substantially reduce drift caused due to lateral forces thereby lessening the conservatism in traditional design practices. This thesis provides results obtained from full-scale experimental testing and analytical study for a metal building. A full scale load test was conducted to quantify the lateral stiffness of an existing metal building. A static lateral load, consistent in magnitude with the building's design wind pressure, was applied to the knee of a primary frame, and the resulting lateral displacements and column-base rotations for all primary frames were measured. The test procedure was repeated at several locations. The experimentally obtained results were then validated using two-dimensional and three-dimensional analytical models. The three-dimensional models explicitly simulated the primary and secondary framing, roof and wall diaphragms, and column-base stiffness. A couple of approaches have been proposed to model column-base plate connection varying in complexity and accuracy. Once validated, the FE model is utilized to quantify the relative stiffness contributions of the metal building system components to lateral drift. While performing analysis some other parameters were also studied. These consisted of effect of base plate thickness and length of anchor bolts on column-base rigidity. Also, effect of including shear deformations and considering the haunch (column-rafter junction) as rigid were studied. Another small but important part of the paper is comparison of wind pressures obtained using different procedure of ASCE 7-05 with database assisted design pressures. Once these parameters are quantified practical engineering guidelines are developed to incorporate the influence of secondary framing, roof diaphragms, wall cladding, and column-base stiffness and wind loads in metal building design. / Master of Science

A methodology to measure a characteristic time scale (duration) of peaks in pressure and velocity data is presented. This methodology is based on the use of the Morlet and Paul wavelets. Detailed descriptions of these wavelets and their implementation procedures are given. The results show that similar time scales or durations can be measured using either Morlet or Paul wavelets. To obtain consistent results data windowing might need to be applied. Using the Paul wavelet, durations of events measured in different wind tunnel simulations are obtained and discussed. / Master of Science

The full-scale value of the Reynolds number associated with wind loads on structures is of the order of 10^7. This is further complicated by the high levels of turbulence fluctuations associated with strong winds. On the other hand, numerical and wind tunnel simulations are usually carried out at smaller values of Re. Consequently, the validation of these simulations should only be based on physical phenomena derived with tools capable of their identification. In this work, two physical aspects related to extreme wind loads on low-rise structures are examined. The first includes the statistical properties and prediction of pressure peaks. The second involves the identification of linear and nonlinear relations between pressure peaks and associated velocity fluctuations. The first part of this thesis is concerned with the statistical properties of surface pressure time series and their variations under different incident flow conditions. Various statistical tools, including space-time correlation, conditional sampling, the probability plot and the probability plot correlation coefficient, are used to characterize pressure peaks measured on the top surface of a surface-mounted prism. The results show that the Gamma distribution provides generally the best statistical description for the pressure time series, and that the method of moments is sufficient for determining its parameters. Additionally, the shape parameter of the Gamma distribution can be directly related to the incident flow conditions. As for prediction of pressure peaks, the results show that the probability of non-exceedence can best be derived from the Gumbel distribution. Two approaches for peak prediction, based on analysis of the parent pressure time series and of observed peaks, are presented. The prediction based on the parent time series yields more conservative estimates of the probability of non-exceedence. The second part of this thesis is concerned with determining the linear and nonlinear relations between pressure peaks and the velocity field. Validated by analytical test signals, the wavelet-based analysis is proven to be effective and accurate in detecting intermittent linear and nonlinear relations between the pressure and velocity fluctuations. In particular, intermittent linear and nonlinear velocity pressure relations are observed over the nondimensional frequency range fH/U<0.32. These results provide the basis for flow parameters and characteristics required in the simulation of the wind loads on structures. / Ph. D.

The issue of the effects of free stream turbulence on the flow field over a surface-mounted prism is examined through experimental and numerical investigations. In the experimental studies, particle image velocimetry measurements are conducted in the ESM water tunnel at Reynolds number of $9,600$ and under two cases of turbulent inflow conditions. The results show that the mean flow separation, reattachment and parameters such as mean velocity, root mean square, Reynolds stresses and turbulent kinetic energy are affected by the turbulence characteristics of the incident flow. The instantaneous dynamics of the interactions between the separating shear layer and the solid wall and between the shear layer and the turbulence in the incident flow are detailed. In the numerical studies, large eddy simulations of the flow over a surface-mounted prism under two inflow conditions, namely, smooth inflow and isotropic homogeneous turbulence inflow, are performed. The use of a fifth-order scheme (CUD-II-5), which is a member of a family of Compact Upwind Difference schemes, in large eddy simulations of this flow is assessed. The performance of this scheme is validated by comparing the rate of temporal decay of isotropic turbulence with available experimental measurements for grid-generated turbulence. The results show that the spectra are sensitive to the method of flux vector splitting needed for the implementation of the upwind scheme. With van Leer splitting, the CUD-II-5 scheme is found to be too dissipative. On the other hand, using the Lax-Friedrichs vector splitting yields good agreement with experiments by controlling the level of artificial dissipation. This led us to recommend a new procedure, we denote by C6CUD5 scheme, that combines a compact sixth-order scheme with the CUD-II-5 scheme for large eddy simulation of complex flows. The simulation results, including flow patterns, pressure fields and turbulence statistics show that the CUD-II-5 scheme, with Lax-Friedricks flux vector splitting, provides high resolution of local flow structures. The results present new physical aspects of the flow topology over surface-mounted prisms. The effects of the incident homogeneous turbulence on the size of the separation region and suction pressures are determined by pointing out differences in the flow topologies between the two incident flow cases. / Ph. D.