The selection of filtering bandwidth for modal parameters identification using Hilbert-Huang transform

The development and application of Hilbert-Huang transform are deeply restricted by Mode mixing and End effect. Band-pass filter is an effective approach to mode-mixing, but it brings new end points to the original signal, which would aggravate the end effect. After studying the filtering theory and lots of experiments, how to select more reasonable cut-off frequencies is explained, and a linear least-square fit procedure is used on the data-set with more physical significance. Numerical simulation results demonstrate that the proposed method is more effective.