【作者】 王宁；

【导师】 李铁军；

【作者基本信息】 河北工业大学 ， 机械工程， 2018， 硕士

【摘要】 随着科学技术的不断进步,机电设备日益复杂化,这增大了生产力并满足了人民日益增长的物质需要的同时,也带来了急需解决的问题:复杂机电设备的故障率变得越来越高。传统的基于内部传感信息的机电设备诊断系统通常为局部监测,且其成本大、智能性低、可移植性差。因此,融合外部信息的机电设备故障诊断系统研究势在必行。机电设备的声音能够表征其健康状态,对于设备的故障诊断具有重要作用。利用声源定位技术可以定位出故障的具体位置,这极大地加快了故障的定位及故障的初步分析。而主要利用外部信息进行故障诊断的巡检机器人多是识别仪表数据,其自主分析故障的能力较弱。基于以上考虑,本文着重研究了将声源定位融入巡检机器人对机电设备进行故障诊断的关键技术。本文首先根据不同维数的声音特征参数对识别率影响权重不同,将相关系数(Correlation Coefficient)与梅尔频率倒谱系数(MFCC)结合起来,提出一种基于相关系数的梅尔频率倒谱系数(CCMFCC)。实验表明,该方法更能体现机电设备的声音特征,提高工况识别率。其次,分析了信噪比对时延估计精度的影响,针对广义互相关算法抗噪性较弱,提出一种基于小波包分解和相关系数的广义互相关算法,并对其进行傅里叶插值,解决了时延估计对信号源信噪比及采样率要求高的问题,提高了时延估计的精度和鲁棒性。然后,推导了4元T字阵列声源定位算法的定位公式,并对其进行定位范围和误差分析。针对不足,提出一种7元12组T字阵列模型。该模型水平方向定位范围可达到0~360度,可实现对声源的全方位定位,定位精度不受声源方位的影响,定位范围和定位精度较单一T字阵列均有明显提升。最后,将声源定位融入移动机器人,搭建软硬件平台,对典型机电设备(钻铣床)在不同工况下的声音进行识别及定位。首先,利用CCMFCC进行声学特征提取;之后,利用改进的广义互相关算法对7元12组T字阵列的麦克风数据进行分析,用于故障声源定位。实验结果表明,本文提出的方法可很好的实现机电设备的故障诊断过程。更多还原

【Abstract】 With the continuous progress of science and technology,electromechanical equipment has become increasingly complex,which developing productive forces and meeting people’s growing material demand,also creating some urgent problems: complex electromechanical equipment has a high failure rate.The traditional electromechanical equipment diagnosis system based on internal sensing information has the disadvantage of local monitoring,high cost,low intelligence and poor portability.The sound of electromechanical equipment can characterize the health status of the equipment and play an important role in the fault diagnosis of the equipment.The use of sound source localization technology can locate the specific location of the fault,which greatly accelerate the positioning of the fault and the initial analysis of the fault.At present,the methods of fault diagnosis by using inspection robots are mostly based on machine vision,and their ability to independently analyze the capability faults is relatively weak.Based on these considerations above,this paper emphasizes the study of the key technology of integrating the sound source localization technology into inspection robots to diagnose the electromechanical equipment failure.In this paper,firstly,according to different dimensions of sound features have different contribution to identify the sound signal,combine Correlation Coefficient and Mel Frequency Cepstrum Coefficients(MFCC)and propose an improved Mel Frequency Cepstrum Coefficients(CCMFCC).The experimental results indicate that this method can better reflect the sound characteristics of electromechanical equipment and raise the working condition recognition rate.Secondly,this paper analyzed the influence of signal-to-noise ratio on the accuracy of time-delay estimation and studied the influence of time delay estimation error on the location result.The generalized cross-correlation time delay estimation is improved.A generalized cross-correlation time delay estimation based on wavelet packet decomposition and correlation coefficient is proposed and utilize Fourier-interpolation to solve the problem that time delay estimation needs high signal-to-noise ratio(SNR)and sampling rate,and raise the accuracy and noise immunity of time delay estimation.And then,we deduce the location formula of 4-element T-shaped array sound source localization algorithm,and analyze the range and error of localization.In view of the insufficiency,we propose a 7-element 12 groups of T-microphone array model.The horizontal positioning range of this model can reach 0 ~ 360 degrees,which can realize the all-directional positioning of the sound source.The localization accuracy is not affected by the sound source orientation.Compared with the T-shaped array,the range and accuracy of localization are obviously improved.Finally,we integrate the sound source localization into the mobile robot,set up the hardware and software platform.Performance tests carried out in a typical electromechanical equipment(bench drill)when different modes.The experimental results indicate that utilize 7-element 12 groups of T-microphone array model,the CCMFCC acoustic feature extraction method and the generalized cross-correlation algorithmcan which based on wavelet packet decomposition and correlation coefficient can realize the fault diagnosis of electromechanical equipment well.更多还原