基于信号奇异性分析的轴承故障检测方法

Bearing's Fault Detection Method Based on Signal Wavelet Singularity Analysis

针对轴承振动信号利用小波单奇异点检测无法克服噪声影响的不足,提出利用小波模极大值分析信号奇异性变化进而进行轴承故障检测的方法。实验中对信号的模极大分形指数,模极大分形指数熵,Lipschitz指数以及Lipschitz指数熵等奇异特征进行分析比较,实验结果表明这些特征都能有效克服噪声影响实现故障检测,但模极大曲线数最能体现故障特征且检测效果最好。将该方法同基于小波包能量谱特征和小波单奇异点检测的方法进行比较,结果表明本文建议的方法在检测时间及检测率上都有显著提高。

Aiming at the disadvantage that the wavelet singular-point detection based method is more sensitive to the noise in bearing＇s vibration signal identification, the scheme of bearing＇s fault detection is proposed based on the number of singularity points using wavelet transform modulus maximum with constant length. In the test, the wavelet transform modulus maximums of fractal spectrum, fractal spectrum entropy, Lipschitz spectrum, and Lipschitz spectrum entropy are analyzed and compared. The testing results show that the number of singularity points with constant length can effectively reflect the characteristics of bearing＇s faults. The proposed method is compared with the method based on wavelet packet energy spectrum and wavelet singular-point detection based method in the experiments. The results show that the number of singularity points using wavelet transform modulus maximum with the constant length is particularly well adapted to describe fault characteristics and fault diagnosis, which outperforms the method based on wavelet packet energy spectrum in the aspects of detection time and detection rates.