Dynamical Bayesian Testing for Feature Information of Time Series with Poor Information using Phase-space Reconstruction Theory

摘要

A dynamical Bayesian testing method is proposed to examine feature information on performance variation of time series with poor information in advance. Sub-series of time series are obtained via a regularly sampling, a multidimensional information space is formed by phase-space reconstruction method, probability density functions of phase trajectories are acquired with bootstrap and maximum entropy theory, a referenced sequence from phase trajectories is found by minimum variance principle, the posterior probability density function is established according to Bayesian theory and the mutation probability is defined in the light of fuzzy set theory. At the given significance level, dynamical Bayesian testing for feature information on performance variation of the poor information process is put into effect with the help of the mutation probability. Experimental investigation on vibration acceleration of a rolling bearing for space applications presents that the method proposed can effectively detect feature information on performance variation of time series with the unknown probability distribution and trend for the early detection of the hidden danger, thus avoiding serious accident.