Probabilistic Structural Wind Load Effects And Reliability Estimation

Wind is a sort of strong spatially and temporally random nature phenomenon.To guarantee the building structures' safety and function in wind,the research on reliability of structures against wind is a conventional and vital theme in wind engineering community.Even sustained endeavor has been paid and fruitful research results have been achieved in describing characteristics of natural winds,understanding the interaction between winds and structures,and calculating the wind load effects,the devastating economic loss of civil engineering structures and even fatalities keep happening all over the world,which keeps reminding researchers of inadequacy in structure reliability against winds.In the paper?Past,present and future of wind engineering?published by Davenport in 2002,the topic of reliability to wind loading was stressed again specially.Based on the aforementioned reasons,the topic of this thesis is decided to be about structure reliability against wind loading.As mentioned above,wind is random,which leads to the randomness characteristics of wind induced pressures and structural responses.The core of structure reliability against wind is to quantify the randomness of wind and wind induced effects rationally.There are three aspects to be addressed:1)randomness of wind,which includes the random mean wind speed and direction,the mean wind speed profile,the specially and temporally correlated fluctuating wind field and its spectrum;2)uncertainties of aerodynamic coefficient from wind-structure interaction under given exposure,which includes the uncertainties of turbulence simulation in physical and numerical wind tunnel,and randomness of aerodynamic coefficient under different angles of attacking;3)probabilistic framework for considering the randomness and direction of mean wind speed and aerodynamic coefficient comprehensively.This thesis makes researches containing the above three aspects.The framework of this thesis is as following:Chapter I introduces the motivation and background of this dissertation.Chapter II investigates the influence of uncertainties in simulation of mean wind speed profile and turbulence in the physical wind tunnel tests.Although physical wind tunnel tests have been used frequently nowadays,there are possible discrepancies between the simulated wind field and the real one,which is caused by several reasons including the inadequate knowledge of the real wind characteristics,the more or less simulation errors by different setups,and the missing of low-frequency part in power spectrum due to limited size of the physical wind tunnel,etc.In this chapter,to investigate the influences of these discrepancies,pressure characteristics on a saddle type in various inflows are studied.The effects of mean wind speed profile,turbulence intensity and turbulence integral scale length,and missing of low frequency turbulence are investigated.The method of quasi-steady is explored to compensate the effects of missing low frequency turbulence on fluctuating wind pressures.Chapter III proposes a new convenient probabilistic method for estimation of extreme wind load effects coefficients which is validated by various non-Gaussian wind pressure processes from Chapter II.In wind tunnel tests for practical design,only short time history of wind pressures will be recorded from the point of economy,and wind pressure processes are featured with different non-Gaussian distributions.The probability distributions of extremes for non-Gaussian processes with short time history have been usually calculated by moment-based Hermite model translation process method in wind engineering.However,there are some non-Gaussian processes which cannot be applied or have larger errors by this method.After a detailed investigation of effects of skewness and kurtosis on extremes estimation,new method of defining statistical moments for different tails is proposed which improves the accuracy of the moment-based model.Chapter IV studies sampling error in estimation of probability distribution of extreme wind load effect coefficients by moment-based Hermite model investigated in Chapter III.The sampling errors exist in the estimations of skewness and kurtosis and peak factor of non-Gaussian process by moment-based Hermite method when limited length of data is used.There are no analytical formulas for estimating the peak factors by moment-based Hermite approach currently which means it cannot be theoretically answered how long of the samples to be recorded will be adequate in wind tunnel tests.In this chapter,formulations are provided to determine the sampling errors of the skewness and kurtosis with consideration of the dependence of data.The sampling error of peak factor is then calculated based on the first order second moment(FOSM)method.Its accuracy and effectiveness are examined by comparing the estimated standard error of peak factor with that directly derived from the long-term wind tunnel data in Chapter II.The results of this chapter benefit the determination of record length in wind tunnel tests in the future.Chapter V establishes a probabilistic framework for structure reliability estimation considering the uncertainty and directionality of wind load effect coefficient and mean wind speed and also the correlation between different limit state responses by combining the probability distribution of extreme wind load effect coefficient with probability distribution of mean wind speed.The reliability of claddings on a saddle type roof is estimated by the proposed framework.The codifications on directionality effect in ASCE 07,Australian/New Zealand Standard AS/NZS 1170.2:2002 and AIJ-RLB-2004 are discussed.An alternate approach of treating directionality effect is suggested.The disadvantages of disregarding the correlation between multiple limit state responses are pointed out including that risk of structures against strong winds may be underestimated.Chapter VI summarizes the conclusions of this thesis and future work is recommended.