In the current research on wavelet de-noising for rotor faults signal

the denoising effect often depends on the decomposing layer number

rotor rotating speed and signal sampling frequency.But

the de-noising process can not be realized automatically.For this reason

a new self-adaptive de-noising method for rotor fault signal was developed.Firstly

original data was re-sampled

secondly

re-sampled signal was decomposed to given layer number

finally

de-noising was achieved by Donoho soft threshold method.The de-noising process can be performed automatically without manual selection of decomposing layer number.Anumber of examples from simulation and experiments verify the efficiency of this new method.