基于特征频率筛选的声发射信号去噪算法  

Denoising Algorithm Based on Characteristic Frequency Selection of Acoustic Emission Signal

作　　者：陈文峰 向华[1,2] 许光达 柯西江 CHEN Wenfeng;XIANG Hua;XU Guangda;KE Xijiang(National Numerical Control System Engineering Research Center,Wuhan 430070,China;School of Mechanical Science and Engineering,Huazhong University of Science and Technology,Wuhan 430070,China)

机构地区：[1]华中科技大学机械科学与工程学院,武汉430070 [2]国家数控系统工程技术研究中心,武汉430070

出　　处：《机械工程师》2021年第5期31-35,41,共6页Mechanical Engineer

基　　金：国家科技重大专项“华中9型-新一代智能型数控系统研究开发及应用验证”(2018ZX04035002-002);国家自然科学基金“基于切削大数据的刀具全工艺流程寿命模型”(51675204);国家重点研发计划“多品种大批量混线加工智能工程集成技术研究和应用示范”(2018YFB170183)。

摘　　要：在实际的使用过程中,声发射信号中存在散布于有效信号频带内的高能峰,难以通过连续通带滤波器去除噪声,进而得到有效的信号成分。为解决该问题,将散布的高能峰概括为一种离散的频率分布,并将其视为特征频率,设计了一种去除离散频率分布噪声的算法——基于特征频率筛选的去噪算法(Denoising Algorithm based on Characteristic Frequency Selection,DCFS):采用基于去噪特征频率窗口滑动的频谱筛选方法实现频谱有效成分的提取,进一步对有效成分采用基于iFFT的去噪信号重构方法得到去噪信号。为验证算法的性能,采用互相关函数作为算法性能评价指标,对不同的刀具和材料的声发射信号进行实验,并与通用的带阻滤波算法和LMS(Least Mean Square)-自适应滤波算法进行去噪效果比较,结果显示DCFS的去噪效果更优。In actual use,there are high-energy peaks scattered in the frequency band of effective signal in AE signals,and it is difficult to obtain effective components through band-pass filtering.In order to solve this problem,scattered high energy peaks are generalized as a discrete frequency distribution and regarded as the characteristic frequency.Denoising Algorithm based on Characteristic Frequency Selection(DCFS)is designed to remove the non-processing signal.The spectral filtering method based on sliding window of denoising characteristic frequency is adopted to extract the effective components of the spectrum,furthermore,iFFT based denoising signal reconstruction method is adopted for the effective components to obtain the denoising signal.In order to verify the performance of the algorithm,the cross-correlation function is used as the performance evaluation method.The acoustic emission signals of different tools and materials are tested,and the denoising effect is compared with the general band-stop filtering algorithm and Least Mean Square(LMS)-adaptive filtering algorithm.The results show that the denoising effect of DCFS is better.

关 键 词：声发射 去噪技术 高能峰 互相关性 铣削加工 离散频率分布 

分 类 号：TN911.7[电子电信—通信与信息系统] TG54[电子电信—信息与通信工程]