基于多传感器的滚动轴承故障检测研究

发布时间：2018-04-15 22:27

本文选题：滚动轴承 + 故障检测　；参考：《河南科技大学》2015年硕士论文

【摘要】：滚动轴承是机械设备中最广泛应用的一种零部件,其运行状态可直接决定了整台机器的工作状态,传统的利用单一传感器识别滚动轴承故障的方式,能够采集到的故障信息有限,工作环境中的噪声干扰又比较大,常常会将微弱的故障信号淹没,甚至造成误判或错判,现阶段研究单一方法的比较多,综合研究的少。加速度和声发射传感器检测属于不同的技术检测手段,两者之间具有一定相关性和互补性,将它们有机地结合起来,对分析检测滚动轴承故障很有效果;因此,本文利用信息融合技术把加速度和声发射两种检测方法融合起来,来研究诊断滚动轴承的故障。本文系统全面地对滚动轴承故障的加速度信号和声发射信号的采集、特征提取和特征模型建立的过程进行了探讨,提出了基于BP神经网络的多信息融合方法,将该技术应用到滚动轴承故障的诊断中,提高了滚动轴承故障诊断的正确率。首先,根据实验对象选取了参数合适的加速度传感器和声发射传感器,并对传感器进行了校验,以确定传感器在正常的工作状态,以博峰轴承试验台为基础,搭建了加速度和声发射采集系统进行数据采集;然后分析了滚动轴承出现故障时候的振动机理,找出了滚动轴承发生不同故障时候相对应的理论特征频率,对安装在试验台上的加速度和声发射传感器所采集的信号进行了时域、频域分析,并运用希尔伯特振动分解的方法对加速度信号进行降噪,运用小波降噪的方法对声发射信号进行滤波降噪,对降噪后的信号包络解调,获取相对应信号的包络谱图,通过与故障轴承理论特征频率作对比,诊断轴承的故障类型;最后利用BP神经网络建立了基于多传感器的信息融合系统,并设计计算了故障信号的特征向量,经过归一化处理之后送入网络进行训练,直到达到所要求的误差范围以内,实现了对滚动轴承故障的诊断。本文对传感器技术、滤波降噪、包络解调以及神经网络的信息融合技术在滚动轴承检测方法的应用进行了积极的研究与探索,结合硬件平台对滚动轴承故障多信息融合监测进行了实验验证,实验数据表明:单一利用加速度传感器诊断的准确率是78%,利用声发射传感器判别的准确率是90%,而信息融合后的准确率提高到了94.1%,由此表明,通过多传感器的信息融合技术对滚动轴承进行故障诊断,可以提高故障诊断的正确率。
[Abstract]:Rolling bearing is one of the most widely used parts in mechanical equipment. Its running state can directly determine the working state of the whole machine.The limited fault information can be collected, and the noise interference in the working environment is relatively large, which often submerges the weak fault signals, and even results in misjudgment or misjudgment. At present, there are more single methods and less comprehensive research.Acceleration and acoustic emission sensor detection belong to different technical detection methods, and they have certain correlation and complementarity. It is very effective to analyze and detect rolling bearing faults by combining them organically.In this paper, the acceleration and acoustic emission detection methods are combined by using information fusion technology to study the fault diagnosis of rolling bearings.In this paper, the acquisition of acceleration signal and acoustic emission signal of rolling bearing fault, the process of feature extraction and the establishment of feature model are systematically discussed, and the method of multi-information fusion based on BP neural network is put forward.This technique is applied to the fault diagnosis of rolling bearing, and the correct rate of fault diagnosis of rolling bearing is improved.Firstly, the acceleration sensor and acoustic emission sensor with suitable parameters are selected according to the experimental object, and the sensor is calibrated to determine the normal working state of the sensor, which is based on the Bofeng bearing test bed.The acceleration and acoustic emission acquisition system is built to collect data, and then the vibration mechanism of rolling bearing is analyzed, and the corresponding theoretical characteristic frequency of rolling bearing when different fault occurs is found out.The signals collected by acceleration and acoustic emission sensors installed on the test bench are analyzed in time domain and frequency domain, and the acceleration signal is de-noised by Hilbert vibration decomposition method.The wavelet denoising method is used to filter the acoustic emission signal, demodulate the signal envelope, obtain the envelope spectrum of the corresponding signal, and diagnose the fault type of the bearing by comparing with the characteristic frequency of the fault bearing theory.Finally, the information fusion system based on multi-sensor is established by using BP neural network, and the eigenvector of the fault signal is designed and calculated. After normalized processing, it is sent to the network for training until it reaches the required error range.The fault diagnosis of rolling bearing is realized.In this paper, the application of sensor technology, filtering and noise reduction, envelope demodulation and neural network information fusion in rolling bearing detection is studied and explored.Combined with hardware platform, the multi-information fusion monitoring of rolling bearing fault is experimentally verified.The experimental data show that the diagnostic accuracy of single acceleration sensor is 78 and that of acoustic emission sensor is 90. The accuracy of information fusion is improved to 94. 1.The fault diagnosis of rolling bearing can be improved by multi-sensor information fusion technology.
【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TH133.33;TH165.3

【参考文献】