Canonical polyadic decomposition enhanced stochastic wave-based spectral representation method for the simulation of nonstationary turbulent wind fields

Stochastic wave-based spectral representation method (SWSRM) has attracted massive attention and gained advancements in recent years as a widely utilized approach for stochastic wind field simulation. Nevertheless, when generating the nonstationary wind fields in multi-spatial dimensions, SWSRM is still limited by substantial memory usage and low computational efficiency due to the multiple operations on large matrices. In this work, a novel Canonical Polyadic Decomposition (CPD) enhanced SWSRM is proposed. The hybrid of interpolation and CPD decouple and downsize the evolutionary wavenumber-frequency joint spectrum (EWFJS) into multiple univariate vectors, which can mitigate memory consumption. Furthermore, a factorization method is introduced to decompose the triple index random phase angle into three independent random vectors, while the time-consuming three-dimensional FFT (3D-FFT) can be substituted with 1D-FFT. Numerical experiments validate the accuracy and effectiveness of the proposed method via multiple statistical metrics. Moreover, parametric analysis of computational time is conducted while the results further demonstrate the improving efficiency of the CPD enhanced SWSRM in the simulation of the nonstationary wind fields in two-spatial dimensions.