While the vibration power for a set of harmonic force and velocity signals is well defined and known, it is not as popular yet for a set of stationary random force and velocity processes, although it can be found in some literatures. In this paper, the definition of the vibration power for a set of non stationary random force and velocity signals will be derived for the purpose of a time frequency analysis based on the definitions of the vibration power for the harmonic and stationary random signals. The non stationary vibration power, defined as the short time average of the product of the force and velocity over a given frequency range of interest, can be calculated by three methods: the Wigner-Ville distribution, the short time Fourier transform, and the harmonic wavelet transform. The latter method is selected in this paper because band-pass filtering can be done without phase distortions, and the frequency ranges can be chosen very flexibly for the time frequency analysis.
Three algorithms for the time frequency analysis of the non stationary vibration power using the harmonic wavelet transform are discussed. The first is an algorithm for computation according to the full definition, while the others are approximate. Noting that the force and velocity decomposed into frequency ranges of interest by the harmonic vvavelet transform are constructed with coefficients and basis functions, for the second algorithm, it is suggested to prepare a table of time integrals of the product of the basis Functions in advance, which are independent of the signals under analysis. How to prepare and utilize the integral table are presented. The third algorithm is based on an evolutionary spectrum. Applications of the algorithms to the time-frequency analysis of the vibration power transmitted from an excitation source to a receiver structure in a simple mechanical system consisting of a cantilever beam and a reaction wheel are presented for illustration.