PSO-DVMD-WT的变形信号去噪方法研究

The denoising method of deformation monitoring data based on PSO-DVMD-WT algorithm

为了滤除变形数据中含有的白噪声,该文提出一种基于粒子群优化算法的双重变分模态分解-小波阈值去噪模型。首先利用VMD对变形数据进行初次分解,初次分解层数K1由频谱图波峰个数确定,根据相关性分析将分量分为噪声分量和信号分量;然后针对信号分量出现模态混叠的现象,首次分解的信号分量再次进行粒子群优化的VMD分解,得到二次信号分量和二次噪声分量;对二次VMD分解得到的噪声分量进行小波阈值降噪;最后重构实现噪声的有效剔除。模拟实验结果显示,利用本文方法去噪得到的均方根误差降低至0.4180 mm、信噪比提升至10.1740 dB,对比小波阈值、总体经验模态分解(EEMD)、VMD等方法,降噪效果有明显的提升。在实际变形数据去噪中,相比于其他去噪方法,本文方法能够很好地抑制模态混叠的现象,且均方根误差降低至0.1510 mm、信噪比提升至23.8210 dB,验证了本文方法在实际应用中的有效性。

In order to remove the white noise in the deformation data,a dual variational modal decomposition-wavelet threshold denoising model based on particle swarm optimization(PSO)was proposed.First,the variational mode decomposition(VMD)algorithm was used to decompose the deformation data for the first time,The number of the first decomposition layer K1 was determined by the number of wave peaks in the spectrum,The component was divided into noise component and signal component according to correlation analysis.In view of the phenomenon of modal aliasing of signal components,VMD decomposition of the first decomposed signal components is carried out again by particle swarm optimization,The second signal component and the second noise component were obtained,and the noise component obtained from the two VMD decomposition was denoised by wavelet threshold value,and finally the noise was effectively eliminated by reconstruction.Simulation results showed that the root-mean-square error obtained by de-noising with the method in this paper was reduced to 0.4180 mm and the signal-to-noise ratio was increased to 10.1740 dB,Compared with wavelet threshold,ensemble empirical mode decomposition(EEMD),VMD and other methods,the noise reduction effect was significantly improved.The method in this paper could effectively suppress the phenomenon of modal aliasing,and the root-mean-square error is reduced to 0.1510 mm and the signal-to-noise ratio was increased to 23.8210 dB,which verified the effectiveness of the method in practical application.