提出一种变分模态分解-排列熵的去噪方法,分析并设定排列熵中关键参数和阈值,进而通过排列熵来确定变分模态分解的分解层数值,将分解的各模态进行重构以实现对振动信号的去噪。通过仿真测试来验证该方法在正交性、完备性、信噪比和效率方面的优越性,最后对系统采集的实际振动信号进行去噪处理。实验结果表明,与现有的经验模态分解-相关系数和完全经验模态分解-相关系数方法相比,所提方法对触网、车轮碾压和雨淋三种振动信号具有最优的去噪信噪比(含噪信号与降噪值之比),分别为32.5358 dB、30.5546 dB和29.3435 dB,耗时也较少,分别为1.4432,1.6320,1.2349 s,信号模式识别准确率最高,均在99%以上。

In this paper, a denoising method of variational mode decomposition and permutation entropy is proposed, the key parameters and thresholds in permutation entropy are analyzed and set, and then the decomposition layer value of variational mode decomposition is determined by permutation entropy, and the decomposed modes are reconstructed to achieve denoising of vibration signals. The advantages of this method in orthogonality, completeness, signal-to-noise ratio, and efficiency are verified by simulation tests. Finally, the actual vibration signals collected by the system are denoised. The experimental results show that, compared with the existing empirical mode decomposition-correlation coefficient and full empirical mode decomposition-correlation coefficient methods, the proposed method has the best denoising signal-to-noise ratio (the ratio of noise signal to noise reduction) for three kinds of vibration signals (contact, wheel rolling, and rain), which are 32.5358 dB, 30.5546 dB, and 29.3435 dB, respectively, and the time-consuming is also less, which is 1.4432 s, 1.6320 s, 1.2349 s, respectively, and the accuracy of signal pattern recognition is the highest, all above 99%.