Abstract: To perform wind-resistant reliability analysis of structures，it is necessary to simulate the stationary stochastic fluctuating wind field． Simulating the stochastic fluctuating wind fields through spectral representation method ( SＲM) often involves thousands of random variables，leading to the“curse of dimensionality”． On this basis， this paper proposes a new method for points selection in high-dimensional random-variate space to ensure the accuracy and efficiency of SＲM． The new method mainly consists of three steps: the decomposition of high-dimensional random-variate space，the points selection in the orthogonal subspaces，and the random pairing to obtain the representative points and assigned probabilities in high-dimensional random-variate space． The proposed method is used to investigate the horizontal wind field acting along the height of a 30-story frame-shear wall structure． The results show that the proposed method is of high accuracy and efficiency for the simulation in the second-order statistical sense and a complete probability set can be constructed by only using a small number of representative sample processes． Besides，the proposed method can be further combined with the probability density evolution method to elaborately evaluate the probabilistic response of wind-induced vibration of structures．

Key words: high-dimensional random space, fluctuating wind speed fields, spectral representation method, orthogonal subspace, Voronoi cells, random pairing