基于品质可调小波去噪的低速滚动轴承故障诊断

Low-speed rolling bearing fault diagnosis based on quality adjustable wavelet denoising

摘要:

针对低速运行滚动轴承故障特征易被噪声湮没的问题，提出了一种基于可调品质因子小波分解(TQWT，Tunable Q-factor Wavelet Transform)的分层自适应阈值去噪方法，并将该方法与包络谱分析相结合，对低速轴承进行故障分析与诊断；首先，将采集到的轴承振动信号进行TQWT分解，得到分解后的各层小波系数； 然后，利用Sigmoid函数构造分层自适应阈值函数，并利用该阈值函数对TQWT的高频系数进行阈值去噪处理；最后，结合去噪后的高频小波系数和低频小波系数对信号进行重构，得到去噪后的轴承振动信号。通过仿真故障信号，模拟故障实验信号和实测故障信号分别进行了去噪实验分析。实验结果表明，经典的软阈值函数和硬阈值函数相比，本文方法能获得更好的去噪效果，在降低噪声干扰的同时有效保留了轴承的故障特征信息，去噪后信号的包络谱可以清晰的呈现故障的频谱特征，可观察到故障特征的多倍频峰值，且峰值附近干扰很少，有效提高了轴承早期故障的诊断精度。在仿真信号实验中，与软阈值、硬阈值函数相比，本文方法去噪后具有更高的信噪比(SNR， Signal-to-noise Ratio )和更低的均方根误差(RMSE, Root Mean Square Error)，与硬阈值函数相比，本文方法的SNR平均增加了4.1491，RMSE平均下降了0.1329； 与软阈值函数相比，本文方法的SNR平均增加了5.1118，RMSE平均下降了0.1505。

Abstract:

In view of the problem that low-speed running rolling bearing fault characteristics are prone to noise annihilation, a hierarchical adaptive threshold denoising method based on (TQWT，Tunable Q-factor Wavelet Transform) is proposed, and the method is combined with envelope spectrum analysis for fault analysis and diagnosis of low-speed bearings.First, the collected bearing vibration signal is TQWT decomposed to obtain the decomposed wavelet coefficient; then construct the hierarchical adaptive threshold function using Sigmoid function to threshold the high frequency coefficient of TQWT; finally, combined with the high frequency wavelet coefficient and low frequency wavelet coefficient to reconstruct the signal to obtain the denoising bearing vibration signal. Experimental results show that compared with the classical soft threshold function and the hard threshold function, the method in this paper can obtain better denoising effect.While reducing noise interference, the fault characteristic information of the bearing is effectively preserved.The envelope spectrum of the signal after denoising can clearly show the spectral characteristics of the fault, and the multi-frequency peak of the fault characteristic can be observed, and there is little interference near the peak.Which effectively improves the diagnosis accuracy of early bearing faults.In the simulation signal experiment, compared with the soft threshold and hard threshold functions, the method in this paper has higher signal-to-noise ratio (SNR) and lower root mean square error (RMSE) after denoising.Compared with the hard threshold function,the SNR of the proposed method increased by 4.1491 on average, and the RMSE decreased by an average of 0.1329; compared with the soft threshold function, the average SNR of the proposed method increased by 5.1118, and the average RMSE decreased by 0.1505.