基于改进集合经验模态分解的信号处理方法研究

Research on Signal Processing Method Based on Improved Ensemble Empirical Mode Decompo

针对信号通过集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)处理后获取的缺陷特征不明显的问题,提出一种EEMD联合主成分分析法(Principal Component Analysis,PCA)的信号处理方法。此方法首先对信号进行EEMD处理;其次在重构信号时,采用相关系数法进行特征选取;最后采用PCA对缺陷信号分量进行处理,能够在有效抑制模态混叠现象的同时,加强信号特征。通过仿真验证得到,经改进后的信号处理方法的重构信号信噪比为29.5237dB,相较于单一EEMD方法提高了51.06%,均方根误差降低了51.02%。改进后的算法能够在去除噪声的同时,保留更多的信号特征,更适用于处理超声检测信号。

Aiming at the indistinct defect features obtained after signal processing through Ensemble Empirical Mode Decomposition(EEMD),this paper proposes a signal processing method that combines EEMD with Principal Component Analysis(PCA).With this method,the signal is first processed by EEMD,and then during signal reconstruction,feature selection is performed using the correlation coefficient method.Finally,PCA is applied to process the defective signal components,effectively suppressing mode mixing phenomena while enhancing signal features.Simulation results show that the reconstructed signal-to-noise ratio of the improved signal processing method is 29.5237dB,an increase of 51.06%compared to the single EEMD method,and the root mean square error reduces by 51.02%.The improved algorithm can eliminate noise while preserving more signal features,making it more suitable for processing ultrasonic detection signals.