Efficacy of Interpolation-Enhanced Schemes in Random Wind Field Simulation over Long-Span Bridges

Abstract

The spectral representation method (SRM) has been widely used to simulate the random wind field for engineering structures. In the modeling process, the Cholesky decomposition of the cross power spectral density (CPSD) matrix is a critical step in the simulation. However, the decomposition of the CPSD matrix is a time-consuming procedure and the required decomposition becomes a computational challenge as the number of points to be simulated as well as frequency segments becomes larger. Such requirements are often noted in buffeting analysis of long-span bridges because they cover a large expanse. This paper examines the efficacy of the interpolation-enhanced scheme (IES) in improving the computational efficiency of SRM by reducing the number of Cholesky decompositions in the random wind field simulation of long-span bridges. Although the interpolation technique has been used to reduce the computational efforts of SRM, there is no literature on the selection of appropriate parameters concerning the interpolation, which is a critical consideration in the context of modeling efficiency and accuracy in applications. Hence, three vital issues related to the selection of interpolation functions, distribution of interpolated points, and interpolation intervals are discussed via a parametric study. The results show that IES can significantly improve the modeling efficiency of SRM, and the improvement becomes increasingly noteworthy as the number of simulation points increase. Hermite and spline-based interpolations are two interpolation functions of choice that can reduce the computational efforts and minimize the attendant computational error. The exponential and quartic polynomial transform are recommended for making the distribution of interpolated points adapt to the low-frequency concentrated contents of each element in the H matrix. In addition, reasonable suggestions are also provided for the selection of interpolation intervals. A comparison between the IES and the POD-based approach as well the explicit decomposition method demonstrates the effectiveness of IES in terms of both the computational efficiency and accuracy.