EEMD结合改进PCNN模型的气体泄漏信号降噪

EEMD Combined with Improved PCNN Model for Noise Reduction of Gas Leakage Signal

针对气体泄漏声波信号降噪的问题,提出了一种集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)与改进脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)相结合的降噪方法,对采集到的气体泄漏声波信号进行降噪处理,同时与EEMD和变步长自适应滤波(Least Mean Square)降噪算法进行比较。算法首先对信号做EEMD的分解,将原信号分解为9个模态分量,然后对这些分量做相干性的计算,根据分量各自的含噪情况调整参数分别做PCNN降噪,最后将信号重构。原始信号由NI仪器的cDAQ采用声传感器测得。实验结果表明,上述方法能够有效地剔除气体泄漏信号中包含的各种噪声,降噪后信噪比为16.64,均方根误差为0.0209,为后续信号分析减少了干扰,上述方法为气体泄漏声波信号的特征提取与分析提供了新的思路。

Aiming at the noise reduction problem of the sound signal of gas leakage,a set empirical mode decom⁃position(EEMD)method combined with the improved pulse-coupled neural network(PCNN)was proposed to re⁃duce the noise of the collected gas leakage signal and compare it with EEMD and variable step size adaptive filtering.First,the signals were decomposed into 9 modal components by decomposes the original signal into EEMD,then the coherence of these components was calculated,PCNN noise reduction for the noise components was performed,and finally the signals were reconstructed.The original signal was measured by the cDAQ of the NI instrument using an a⁃coustic sensor.The experimental results show that this method can effectively eliminate all kinds of noises contained in gas leakage signals,the signal-to-noise ratio is 16.64 and the root-mean-square error is 0.206 after noise reduction,which improves the signal quality and provides a new idea for feature extraction and analysis of gas leakage signals.