基于局部均值分解的滚动轴承故障诊断系统研究与应用

发布时间：2018-07-10 13:09

本文选题：局部均值分解 + 端点效应　；参考：《中北大学》2017年硕士论文

【摘要】：针对大型设备逐渐趋于复杂化、一体化、智能化,错综复杂的设备之间的关联与耦合作用愈来愈强,极大影响了设备运行状态监测与故障诊断的有效性,继而给机械故障诊断领域带来了巨大挑战。论文以滚动轴承故障特征提取与智能诊断系统研究为主要研究内容,将局部均值分解(Local Means Decomposition,LMD)作为核心技术,结合非线性动力学理论与人工智能分类技术对上述背景下的滚动轴承故障诊断系统展开研究。针对局部均值分解存在端点效应与模态混叠现象,论文对LMD算法上稍作改进,首先采用基于局部波形积分匹配方法来抑制端点效应,该方法通过三点积分曲线法在信号内部搜索最佳匹配波形,在局部信号端点处采用扩展波形抑制端点效应。针对LMD存在模态混叠问题,提出基于总体均值分解与频率截止方法抑制模态混叠,该方法采用功率谱分析求得原始信号中频率成分最小的信号,再向原始信号中加入等幅值的高斯白噪声,对混合信号进行反复LMD分解,将得到的分量瞬时频域与信号最小截止频率对比,以此作为分量迭代终止条件。通过仿真与实验数据分析,验证所提方法不仅能够改善LMD在端点效应与模态混叠的问题,对于低频伪分量的抑制也有较好的效果。为了简化故障诊断流程,论文在改进的LMD算法的基础上,采用模糊熵对故障特征进行量化处理,从多个角度对原始信号进行深度剖析,提取全面表征故障特征的特征向量,结合具有极强的非线性分类能力的概率神经网络(probabilistic neural network,PNN)实现故障模式识别。最后论文在对LMD自时频信号分析处理方法研究的基础上,利用人机交互能力强的LabView与超强运算分析能力的Matlab进行混合编程,研究开发一套具有高效、准确的滚动轴承智能诊断系统,搭建一套集在线数据采集、数据分析与故障诊断于一体的滚动轴承故障诊断。
[Abstract]:In view of the complex, integrated, intelligent and complicated equipment, the relationship and coupling between the equipments is becoming stronger and stronger, which greatly affects the effectiveness of the monitoring and fault diagnosis of the equipment operating state. Then it brings great challenge to the field of mechanical fault diagnosis. In this paper, the fault feature extraction and intelligent diagnosis system of rolling bearing are the main research contents, and the local mean decomposition (LMD) is taken as the core technology. Combined with nonlinear dynamics theory and artificial intelligence classification technology, the rolling bearing fault diagnosis system under the above background is studied. Aiming at the existence of endpoint effect and modal aliasing in local mean decomposition, the LMD algorithm is improved in this paper. Firstly, the local waveform integral matching method is used to suppress the endpoint effect. In this method, the best matching waveform is searched inside the signal by three-point integral curve method, and the extended waveform is used to suppress the endpoint effect at the end point of the local signal. Aiming at the problem of modal aliasing in LMD, a method based on population mean decomposition and frequency cutoff is proposed to suppress modal aliasing. The power spectrum analysis is used to obtain the signal with the smallest frequency component in the original signal. The Gao Si white noise with equal amplitude is added to the original signal, and the mixed signal is decomposed repeatedly. The instantaneous frequency domain of the component is compared with the minimum cut-off frequency of the signal, which is used as the iterative termination condition of the component. The simulation and experimental data analysis show that the proposed method can not only improve the end-point effect and modal aliasing of LMD, but also have a good effect on the suppression of low-frequency pseudo-components. In order to simplify the process of fault diagnosis, based on the improved LMD algorithm, the fuzzy entropy is used to quantify the fault features, and the original signals are analyzed in depth from many angles, and the feature vectors representing the fault features are extracted. Fault pattern recognition is realized with probabilistic neural network (probabilistic neural) which has strong nonlinear classification ability. Finally, on the basis of the research on the analysis and processing method of LMD self-time-frequency signal, a set of high efficiency is developed by using LabView, which has strong human-computer interaction ability, and Matlab, which has super ability of operation and analysis. An accurate intelligent diagnosis system for rolling bearings is established. A set of on-line data acquisition, data analysis and fault diagnosis are built for fault diagnosis of rolling bearings.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【参考文献】