Johnson curve model based CDF mapping method for estimating extreme values of the measured non-Gaussian wind pressures including the non-stationary effect investigation

Abstract

The study about the extreme value estimation of non-Gaussian wind pressure signals is crucial for structural safety or health monitoring. This is because the long tail region of the non-Gaussian distribution will lead to the unexpected extreme value which results in structural damage or fatigue damage in the practical engineering. In a general view, the existing extremes value estimation method can be divided into two main categories: the CDF mapping method and the peak factor method. The former is relatively complicated but accurate, considered highly dependent on the accuracy of the non-Gaussian parent distribution fitting; The latter is relatively convenient but passable, only requiring the skewness and kurtosis of the standardized non-Gaussian processes used for the input of the polynomial transformation. In this work, firstly, a Johnson curve model-based CDF mapping method (JCCDF) is proposed to combine the advantages of both approaches. By employing the measured non-Gaussian wind pressures, the convenient and accurate JCCDF is verified with the other CDF mapping methods, and a simplified direct JCCDF-based extreme value variables mapping method is developed. Through observing the calculated extreme values in the long-tail region of different methods, it can be figured out that the results of the peak factor method are rather less, and the computed values based on the observed extreme method are close to that of the CDF mapping method. In the investigation about the influence of non-stationary characteristics on the extreme value estimation, it can be spotted that the computed extreme values in the long-tail region of all methods are getting greater with different degrees after considering the time-varying first and second-order statistical moment. This point reveals that the non-stationary characteristics need to be taken into account.