中文摘要：

快速谱峭度图是滚动轴承故障诊断的有效方法,然而,该方法有可能将最佳频带分割到不同区域,造成对故障信息的识别能力不足,并且对随机冲击噪声的免疫力较弱。根据滚动轴承故障激发多个固有频率的特点以及平均降噪原理,提出多自由度小波包诊断方法,选取适当数量的小波包子带信号,将所选子带信号的频谱平均处理,充分利用信号的有用信息,增强了对白噪声及随机冲击噪声的免疫力。对滚动轴承故障仿真信号及实测信号分别应用快速有限冲击响应（FIR）,谱峭度方法,小波包谱峭度方法以及多自由度小波包诊断方法的对比分析,表明多自由度小波包诊断方法对白噪声及随机冲击噪声具有更强的免疫力,验证了该方法的有效性及实用性。

英文摘要：

The spectral kurtogram is effective for rolling bearings fault diagnosis.However,its performance is inadequate to identify fault information since it is possible to divide the optimal frequency-band into different bands.In addition,the method is vulnerable to random impulse noise.As rolling bearings fault can excite vibrations of multiple frequencies,a diagnosis method of multi-degree of freedom-wavelet packet（MDF-WP）is proposed using the average de-noising principle.Reasonable sub-band signals of the WP are selected,and their frequency spectrums are processed by average de-noising.Useful information in the signal is fully used so that immunity to white noise and random impulse noise is improved.Bearing fault simulation signals and the measured signals of the rolling bearings are processed by three methods：fast FIR spectral kurtogram,wavelet packet spectral kurtogram and the proposed MDF-WP method.The results show that the MDF-WP method is more robust to white noise and random impulse noise,verifying its validity and practicability.