Combined with meliorated wavelet packet algorithm

an algorithm of phase space reconstruction of multivariate data based on parameter optimization is proposed to cope with the faultiness of phase space reconstruction from univariate time series

which adopted the nearest neighbor divergence to ensure embedding dimension and the incorporation technology of self-interrelated function and pseudo-phase portrait to ensure time delay.Then the algorithm is applied to extract fault syndrome of indiscernibly of BW350 mud pump.The result shows that the fault eigenvalue achieved with the proposed method is uniform.It has practical value in fault judgment of difficult to identify bearing shell and joint cross.