Adaptive redundant multiwavelet with applications to fault diagnosis

Signals of mechanical equipment faults in operation with obscure symptoms and weak features are always contaminated by stronger background noise. To solve the difficulty, a new method called adaptive redundant multiwavelet is proposed. Following Chui-Lian multiwavelet and two-scale similarity transforms, and taking the minimum envelope spectrum entropy as the optimization objective and genetic algorithms as the optimization tool, the redundant multiwavelet is adaptively constructed. Compared with the Fourier transform, Db6 scalar wavelet transform and CL3 multiwavelet transform, the applications to fault diagnosis rub-impact for a rolling element bearing of outer-race and a flue gas turbine unit of show the improved effectiveness of the proposed method.