基于传感器阵列的机械故障声源定位系统  

Mechanical Failure Sound Source Localization System Based on Sensor Array

作　　者：李铁军[1] 王宁[1] 赵义鹏 刘今越[1] LI Tie-jun;WANG Ning;ZHAO Yi-peng;LIU Jin-yue(School of Mechanical Engineering, Hebei University of Technology, Tianjin 300130, China)

机构地区：[1]河北工业大学机械工程学院,天津300130

出　　处：《机械设计与制造》2019年第4期166-170,共5页Machinery Design & Manufacture

基　　金：国家自然科学基金资助项目(51175145)

摘　　要：传统的机械故障诊断系统需要在被监测本体上安装传感器,只能定点定性监测,无法大范围监测,且可移植性差、智能性弱。针对此问题设计了基于四元声学传感器阵列的故障声源定位系统,能实现故障声源的识别及定位。首先,通过声学传感器采集机械设备运行过程中的声音信号,提取其Mel频率倒谱系数。然后,经BP神经网络对声音信号进行识别判断,若为故障声则采用广义互相关算法计算其时间延迟,进行定位。利用该系统对台式钻床空转、正常、磨损、崩刃四种工况进行识别定位测试,实验结果表明,该系统的工况识别准确率可达到89%,故障声源定位精度误差在6cm以内,具有较好的故障声源识别及定位功能。Traditional mechanical devices fault diagnosis systems need to install sensors on the monitored machine, so that they can only qualitatively monitor in a fixed point. It usually has poor portability and intelligence. A failure sound source localization system based on acoustic sensor array is designed for this problem. Firstly, acoustic sensors are employed to collect the sound signal after the machine is run, the Mel frequency cepstrum coefficient feature vectors of the sound signal are extracted at the same time. Secondly, BP neural network is used to distinguish mechanical failure sounds from the collected acoustic signals. Then the difference of arrival time of the acoustic signals was calculated on the generalized cross-correlation method to locate the position of mechanical failure. Lastly, performance tests carried out in a bench drill when four modes, the idling sound, normal sound, wear sound and the chipping sound are existed respectively. The experimental results indicate that the working state of bench drill can be recognized with a high accuracy of 89%, and the positioning error is less than 6cm. This system has good performance for the identification and localization of the mechanical failure sound.

关 键 词：故障检测 梅尔频率倒谱系数 BP神经网络 广义互相关 声源定位 

分 类 号：TH16[机械工程—机械制造及自动化] TP277[自动化与计算机技术—检测技术与自动化装置]