摘要：
针对容栅传感器检测的转动轴扭振信号掺杂的环境噪声干扰和自身的电磁噪声干扰使得信噪比低、微弱信号难提取的问题,提出了一种基于小波-EEMD-Adaline自适应线性神经网络去噪方法.该方法对信号进行小波、EEMD、Adaline网络消噪处理,采用三级去噪、噪声过滤、对消来逼近原始信号.用典型加噪超声信号、Doppler信号、Block信号对该方法进行有效性验证,与EEMD、基于小波分解的EEMD去噪效果相比较.实验结果表明,后两种方法信号去噪的SNR提升小(均不到20),而本文方法SNR(RMSE)提升(减小)明显,对于9 dB的Doppler信号SNR提升达90,RMSE从1.038 5降至0.009 5.对容栅电路实测大噪声微弱信号去噪,结果表明,该方法去噪性能更优,去噪后信号光滑性好,波动稳定性强.%The capacitance of torsional vibration signal of rotating shaft detected by capacitive grid sensor is dis-turbed by environment noise and electromagnetic noise,which make the signal-to-noise ratio low and the faint signal hard to be extracted.This paper presents a wavelet-EEMD-Adaline adaptive linear neural network denoising method. The proposed method carries out wavelet denoising, EEMD denoising and Adaline denoising on the signal, using three-level denoising, noise filtering and cancellation to approximate the original signal. The effectiveness of this method is validated by using typical noise-added ultrasonic signals, Doppler signals and Block signals, compared with the denoising effect of EEMD denoising method,EEMD based on wavelet decomposition denoising method. The results of the verification experiment show that while the SNR of signal de-noising in the latter two methods is small ( both less than 20) ,the SNR( RMSE) of the proposed method,which SNR is improved by 90 and RMSE decreases from 1.038 5 to 0.009 5 for Doppler(9 dB)signal,increases(decreases)significantly. At last,the denoising results of weak signal with large noise measured by capacitive sensor circuit show that this method has better denoising per-formance,and the signal is smooth and stable after denoising.
