模具修复锤击力信号的dbN小波阈值降噪方法

Denoising Method of dbN Wavelet Threshold for Die Repair Hammer Force Signals

针对模具修复锤击力信号采集噪声干扰较大的问题,运用dbN(db为Daubechies的简写,N表示小波阶数)小波对锤击力信号进行小波分解低频系数信号重构。保留信号的低频系数,舍弃高频系数,重构后信噪比为10.077 8,均方根误差为0.663 3,初步实现了模具修复锤击力信号的降噪。在dbN小波分解的基础上进行阈值化处理,处理后的信噪比SNR明显提高,最大值为44.231 3 dB,均方根误差RMSE明显降低,最小值为0.012 5。实验结果表明两种方法都能实现对锤击力信号的噪声滤除,其中软阈值法还能较大程度的还原原始信号在突变点处的细节特征,避免了信号的失真,保证了模具修复锤击力信号后续计算准确性。

In order to solve the problem that the noise of the hammer power signal in mold repair is large,the dbN wavelet(abbreviation of Daubechies,N is the wavelet order)is used to reconstruct the hammer force signal according to the obtained low frequency coefficient.The low-frequency coefficients of the signal are reserved and the high-frequency coefficients are discarded.Ultimately,the signal-to-noise ratio after noise reduction is 10.077 8 and the mean square error is 0.663 3,the noise reduction of hammer power signal in mold repair is initially achieved.Simultaneously,the thresholding is done on the basis of dbN wavelet decomposition.Experimental results show that the SNR is significantly increased after threshold treatment,up to 39.85 dB;the RMSE is significantly reduced and the minimum value is 0.449 8.Comprehensive macro-waveform characteristics indicating that the two methods can achieve noise filtering of hammering force signal.Besides,soft threshold methods can also be a greater degree of reduction of the original signal at the mutation point for the details characteristics,to avoid the signal distortion and ensure that the subsequent calculation accuracy of hammering force signal in mold repair.