Typhoon winds near its external eye wall are not only strong but also nonstationary with time-varying coherences. This study proposes a conditional simulation method for multivariate nonstationary typhoon winds with time-varying coherences for long-span bridges. The Kriging method for conditional simulation and the time-varying wind spectrum and coherence are first introduced. The decomposition of the time-varying spectrum matrix is then discussed in order to obtain the time-varying correlation functions. It is found that the conventional Cholesky decomposition and the proper orthogonal decomposition (POD) are not suitable and only the square root decomposition (SRD) is appropriate for the decomposition of the spectrum matrix with time-varying coherences. Based on the field-measured nonstationary typhoon wind time histories and the time-varying spectrum and coherence models established for the Stonecutters cable-stayed bridge during Typhoon Hato, the proposed conditional simulation method is finally applied to the bridge and validated by comparing the simulated typhoon wind time histories with the measured ones as well as the unconditional simulation results. The comparative results demonstrate that the conditionally simulated wind time histories are similar to the measured ones and that the proposed conditional simulation method can be used to provide a better typhoon wind field than the unconditional simulation method.

台风外眼壁附近的台风不仅很强，而且随着时变的连贯性也不稳定。本文针对大跨度桥梁，提出了一种具有时变相干性的多元非平稳台风风的条件模拟方法。首先介绍了用于条件模拟的Kriging方法以及随时间变化的风谱和相干性。然后讨论时变频谱矩阵的分解，以获得时变相关函数。发现传统的Cholesky分解和适当的正交分解（POD）不适合，只有平方根分解（SRD）适合具有时变相干性的频谱矩阵的分解。基于现场测量的非平稳台风风场历史，以及为台风哈托期间昂船洲斜拉桥建立的时变谱和相干模型，最终将所提出的条件模拟方法应用于桥梁，并通过对模拟台风进行比较进行了验证。风时间的历史记录以及无条件的模拟结果。比较结果表明，有条件模拟的风时历史与实测历史相似，并且所提出的有条件模拟法可以提供比无条件模拟方法更好的台风风场。最后将所提出的条件模拟方法应用于桥梁，并通过将模拟的台风风时历史与实测风时历史以及无条件的模拟结果进行比较，对其进行了验证。比较结果表明，有条件模拟的风时历史与实测历史相似，并且所提出的有条件模拟法可以提供比无条件模拟方法更好的台风风场。最后将所提出的条件模拟方法应用于桥梁，并通过将模拟的台风风时历史与实测风时历史以及无条件的模拟结果进行比较，对其进行了验证。比较结果表明，有条件模拟的风时历史与实测历史相似，并且所提出的有条件模拟法可以提供比无条件模拟方法更好的台风风场。