打开文本图片集
摘要：针对生产现场机械设备零部件结构复杂、设备运行时背景噪声干扰严重等造成的监测诊断难题，以及传统盲信号处理算法在机械声信号处理方面的局限性，提出一种基于参考信号约束频域半盲提取的机械故障声学诊断算法。详细介绍了该算法的关键技术：以频域盲解卷积算法为基础，使用利于全局寻优的人工鱼群算法，构建适用于机械故障特征的改进多尺度形态学滤波器，以最大程度削弱背景噪声干扰；结合机械设备零部件结构参数构建参考信号，通过单元参考信号约束频域半盲提取算法，对降噪后的信号逐段进行复数盲分离；利用改进KL距离，解决复分量间次序不确定性问题，最终实现机械故障特征信号的提取与分离。实际声场环境中的滚动轴承故障声学诊断实验验证了该算法的有效性。
关键词：算法理论；参考信号约束；频域半盲提取；人工鱼群算法；声学诊断
中图分类号：TH1653文献标志码：A
收稿日期：2014-11-02；修回日期：2015-01-01；责任编辑：张士莹
基金项目：国家自然科学基金（51305186）；昆明理工大学校级人培基金（KK3201301026）
作者简介：羿泽光（1989—），男，河北唐山人，硕士研究生，主要从事状态监测与故障诊断方面的研究。
通讯作者：潘楠博士。E-mail：15808867407@163.com
羿泽光，潘楠，刘凤.基于参考信号频域半盲提取的机械故障特征声学诊断[J].河北科技大学学报，2015，36（4）：351-358.
YI Zeguang，PAN Nan，LIU Feng. Acoustic diagnosis of mechanical fault feature based on reference signal frequency domain semi-blind extraction[J].Journal of Hebei University of Science and Technology，2015，36（4）：351-358.Acoustic diagnosis of mechanical fault feature based on reference signal
frequency domain semi-blind extraction
YI Zeguang， PAN Nan， LIU Feng
（Faculty of Mechanical & Electrical Engineering， Kunming University of Science and Technology， Kunming， Yunnan 650500， China）
Abstract：Aiming at fault diagnosis problems caused by complex machinery parts， serious background noises and the application limitations of traditional blind signal processing algorithm to the mechanical acoustic signal processing， a failure acoustic diagnosis based on reference signal frequency domain semi-blind extraction is proposed. Key technologies are introduced： Based on frequency-domain blind deconvolution algorithm， the artificial fish swarm algorithm which is good for global optimization is used to construct improved multi-scale morphological filters which is applicable to mechanical failure in order to weaken the background noises； combining the structural parameters of parts to build a reference signal， complex components blind separation is carried out on the signals after noise reduction paragraph by paragraph by reference signal unit semi-blind extraction algorithm； then the improved KL-distance of complex independent components is employed as distance measure to resolve the permutation， and finally the mechanical fault characteristic signals are extracted and separated. The actual acoustic diagnosis of rolling bearing fault in sound field environment results proves the effectiveness of this algorithm.
Keywords：algorithm theory； reference signal constraints； frequency-domain semi-blind extraction； artificial fish swarm algorithm； acoustic diagnosis
