基于声发射的滚动轴承智能故障诊断方法研究

发布时间：2018-06-18 21:09

本文选题：滚动轴承 + 故障诊断　；参考：《北京化工大学》2012年硕士论文

【摘要】：当前旋转设备正朝着高功率、智能化、一体化方向发展,有效地对机械设备进行状态监测,在故障早期发展阶段及时发现并予以相应维护,是大型旋转机械设备运行安全性和可靠性的保证。而滚动轴承作为旋转机械设备组的重要组成部分,因此针对于滚动轴承的状态监测、故障诊断和趋势预测的研究就具有极其重要的现实意义。滚动轴承故障诊断融合了机械动力学、现代测试技术、现代信号处理、数据挖掘和人工智能等多学科知识,主要包括信号预处理、特征提取、模式识别和趋势预测四个阶段,其主要任务是提取能反映设备状态的故障特征量,判别故障类型,预测分析故障特征量的发展趋势,根据故障严重程度制定适当的维修计划。本文采用EMD和小波分析两种预处理方法按照频段或频率将信号分解去除噪声成分和其它干扰信息；并提出一种基于“能量-香农熵比’的小波基选取方法,该方法将小波分解过程中的频带能量泄露降到最低。特征提取方面,提出了RMS序列及基于RMS序列的相对熵,其计算简单,抑噪能力强,可有效处理故障早期或信噪比较低情况下的声发射信号；并介绍了近似熵及其快速算法,分析阐述了计算过程中各参数对熵值和计算时间的影响。本文利用改进粒子群优化的神经网络对滚动轴承故障进行模式识别。基于适应度值对粒子群算法优化,通过对标准速度更新公式中各参数和公式本身进行改进,并且结合其它智能算法来完善其搜索策略,突出了粒子在不同阶段的全局和局部搜索能力,有效规避了搜索过程中粒子陷入局部最优点的可能性。最后,采取基于遗传算法的回归预测模型对故障发展趋势进行预测,利用遗传算法优化不同阶次回归模型中的系数。通过轴承外圈故障在强负载不良润滑下的剩余寿命预测实验对此预测算法进行试验验证。
[Abstract]:At present, rotating equipment is developing towards high power, intelligence and integration. It can effectively monitor the status of mechanical equipment, discover and maintain it in time in the early stage of fault development. It is the guarantee of safety and reliability of large rotating machinery. The rolling bearing is an important part of the rotating machinery, so it is of great practical significance to study the condition monitoring, fault diagnosis and trend prediction of the rolling bearing. The fault diagnosis of rolling bearing combines the knowledge of mechanical dynamics, modern test technology, modern signal processing, data mining and artificial intelligence. It includes four stages: signal preprocessing, feature extraction, pattern recognition and trend prediction. Its main task is to extract the fault characteristic quantity which can reflect the state of the equipment, to distinguish the fault type, to predict and analyze the development trend of the fault characteristic quantity, and to make the appropriate maintenance plan according to the fault severity. In this paper, two preprocessing methods, EMD and wavelet analysis, are used to decompose the signal to remove the noise components and other interference information according to the frequency band or frequency, and a wavelet basis selection method based on the "energy-Shannon entropy ratio" is proposed. This method minimizes the frequency band energy leakage in the wavelet decomposition process. In the aspect of feature extraction, the relative entropy of RMS sequence and RMS sequence is put forward, which is simple in calculation, strong in noise suppression ability, and can effectively deal with acoustic emission signal in early fault or low SNR, and the approximate entropy and its fast algorithm are also introduced. The influence of each parameter on entropy value and calculation time is analyzed. In this paper, the improved particle swarm optimization (PSO) neural network is used to identify the fault of rolling bearing. The particle swarm optimization algorithm is optimized based on fitness value. By improving the parameters of the standard speed updating formula and the formula itself, and combining with other intelligent algorithms, the search strategy is improved. It highlights the global and local searching ability of particles in different stages and effectively avoids the possibility that particles fall into local optimum in the search process. Finally, a regression prediction model based on genetic algorithm is used to predict the trend of fault development, and genetic algorithm is used to optimize the coefficients in different order regression models. The prediction algorithm is verified by the residual life prediction experiment of bearing outer ring fault under strong load and poor lubrication.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;TH133.3

【参考文献】