Power Spectrum is an important characteristic quantity of stochastic process in frequency domain. It can represent the power distribution of the stochastic process. However

because the power spectrum is only a numerical characteristics of stationary process in the frequency domain in nature

it is very difficult for it to involve all probability information of the original stochastic process. This shortcoming not only brings difficulties to the analysis of random vibration for the non-stationary process

but also causes that exact solutions of structure reliability wouldn’t be obtained only by means of numerical characteristics solutions even for the stationary process. So it is necessary to investigate the stochastic process at a more frontal level. In this paper

the idea of Fourier stochastic function is adopted to reflect the stochastic process

for the above purpose. Based on the relationship between the power spectrum and Fourier amplitude spectrum

the stochastic Fourier amplitude spectrum is defined. Then Davenport spectrum

which is used widely in wind engineering

is taken as an example to validate the proposed idea. The research finds that

when roughness length z

0

is characterized by log-normal distribution and basic wind speed U

10

at height of ten meters is characterized by extreme value-Ⅰ distribution

the stochastic Fourier amplitude spectrum can be constructed by the power spectrum. Finally

the stochastic Fourier amplitude spectrum is proved to be rational by comparing measured wind speed Fourier spectrum with the stochastic Fourier amplitude spectrum.