Abstract: For accurately identifying the mode of AE signal, the noise in the AE signal must be removed. The denoising results by conventional means of filtering are unsatisfactory, the denoising method by threshold on wavelet coefficients shows unique advantages. Aiming at threshold selection risky problem in the denoising method by threshold on wavelet coefficients, Kmeans clustering method was used to classify the highfrequency coefficients by wavelet decomposition, determining the removal threshold for the wavelet coefficients corresponds to the noise, then the wavelet coefficients were reconstructed to achieve the denoising purpose. Hardthreshold method and softthreshold method was applied on denoising method by threshold on wavelet coefficients for AE signal, threshold generated by Kmeans clustering approach and threshold generated by the improved Donoho method were respectively used as the threshold for the denoising method by threshold on wavelet coefficients, experimental results show that in the three indicators of signal to noise ratio, root mean square error and smoothness, the proposed method is superior to the improved Donoho method.

Key words: Acoustic emission signal,    ,  Kmeans clustering,    ,  Improved Donoho method,    ,  Wavelet transform,    ,  Threshold,    , Denoising

摘要： 为准确识别声发射信号模式,必须剔除声发射信号中的噪声,传统滤波去噪方法效果并不理想,小波阈值去噪方法显示了独特的优势。针对小波阈值去噪方法中阈值设置风险问题,利用K灢均值聚类法对小波分解后的高频系数进行分类,确定去除噪声对应小波系数的阈值,然后进行小波系数重构达到去噪目的。采用硬阈值法与软阈值法对声发射信号进行小波阈值去噪,将基于K灢均值聚类方法生成的阈值和改进Donoho方法生成的阈值分别作为小波去噪阈值,实验结果表明,在信噪比、均方根误差和平滑度三个指标上,本方法优于改进Donoho方法。

关键词: 声发射信号 , K灢均值聚类 , 改进Donoho方法 , 小波分析 , 阈值 , 去噪