Abstract: On the basis of Newland harmonic wavelet, an improved harmonic wavelet is put forward, it has a symmetrical and boxy frequency spectrum, and its expansions and translations will generate a series of wavelet functions which form an orthogonal basis of L2(R) space, the decomposition and reconstitution process of signal by the improved harmonic wavelet is easier than that by the Newland harmonic wavelet, moreover, the improved harmonic wavelet has the same excellent characteristics as the Newland harmonic wavelet. Using the simulation examples, the ability of extracting weak and local signal in strong noise is studied, in comparison to the DB8 wavelet, the effects of the improved harmonic wavelet are better. The improved harmonic wavelet is used to extract the super-harmonics and sub-harmonics components of the rotor faults signals because the harmonics frequencies characteristics reflect the rotor faults very well. The 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5 and 5.0 components are extracted by the improved harmonic wavelet from the rotor fault experimental signals including imbalance, misalignment, rubbing, and oil whirling faults, and the results verify the effectiveness and correctness of the new method.

Key words: Fault diagnosis, Harmonic wavelet, Rotor, Signal decomposition and reconstitution