Research Of Non-Gaussian Characteristics Of Bridge Wind Loads

With construction technology keeps improving, "long, big, light and flexible" have became the development direction of modern bridges. That makes bridges more beautiful and applicable, but also reduces bridges’stiffness, furthermore, it is inevitable that bridges become more sensitive to wind load at the same time. Wind load is a kind of stochastic process, and its extreme value is an important index of designing. Current procedures for estimating the peaks of wind load are based on the assumption that the response is Gaussian, but more and more test and observation have proved that edge area of bridge, especially where shape is continuously changing, the wind load is non-Gaussian. It means that it’s dangerous to design under the Gaussian assumption. Based on the background, this paper does some research about non-Gaussian features of wind load. Its main work includes several aspects as follows:(1) Introducing the development of non-Gaussian wind load and main calculation method of its extreme value, combined theoretical analysis and practical application, drawing some conclusions:calculation results of improved peak factor method are always bigger than actual; Gumbel method has terrible stability of its calculation results, because of the influence of correlation between peak values under short time interval; Sadek-Simiu method apples only to the stochastic process which has big skewness and small kurtosis.(2) Summarizing a few difficult problems about calculating the extreme value of Non-Gaussian wind load:A. First four statistic moments are the only statistic information we have, how to use them to perform the infinite variety of Non-Gaussian distribution? B. How to use only one real test sample data to portray the whole stochastic process? C. Enough peak value samples is the key factor to simulate Gumbel distribution, but how to get enough samples and avoid the adverse effect of auto-correlation property at the same time? D. How to maximize the applicable range of calculation method, to make it be applicable to as much kinds of stochastic processes as possible.(3) For solving the problems raised above, this paper proposes a new method based on simulation technology:utilizing Johnson transformation to produce Non-Gaussian sequence, which could possess any first four statistic moments to simulate Non-Gaussian distribution; a big number of simulation of wind pressure histories compose the sample space of stochastic process, and satisfy the number of Gumbel distribution simulation; using AR model to complete the simulation of auto-correlation property; the method proposed by this paper has adaptability, it can use different PDF and AR model to complete simulation according to different original test data, then calculate the peak wind pressure by classical extreme value theory, so it can be applicable to any kinds of stable stochastic process.(4) Through wind tunnel test results of main girder of bridge, making some conclusions of distributed characteristic of non-Gaussian area. Calculating the extreme value of wind load at measuring points by four traditional method and newly proposed procedure by this paper, and compare the results with the real test data. It proves that the new method is applicable to all kinds of wind load process. In addition, its calculation results are more accurate than other methods.