基于支持向量机的数控机床电主轴故障诊断研究

发布时间：2018-04-27 08:19

本文选题：电主轴 + 轴承　；参考：《合肥工业大学》2017年硕士论文

【摘要】：电主轴是数控机床核心的部件之一,其性能的优劣直接影响到机床的加工精度。对电主轴进行故障诊断能提高数控机床加工的可靠性,具有重大的意义。本文以数控机床电主轴为研究对象,通过经验模态分析和支持向量机等方法,开展其故障诊断方法的研究,并搭建故障诊断平台。主要工作如下:(1)对电主轴的结构进行了分析,并结合企业提供的维修历史数据,对电主轴的故障进行统计分析,将电主轴中故障频发、危害性较大的轴承故障作为本次研究的对象。(2)对轴承的故障机理,包括振动机理、失效模式、特征频率等进行了分析研究,并总结了各类典型故障的振动特性;将小波降噪、经验模态分析、峭度和相关系数判别方法结合,进行故障信号的特征频率提取,并通过在实验数据中的成功应用验证了该信号处理方法的有效性。(3)结合经验模态分析和自回归模型,优化了支持向量机特征向量的构建方法;通过数据预处理方法(归一化和降维)和参数寻优方法(穷举法、遗传算法和粒子群算法)的分析研究,将实验数据的诊断模型的测试准确率从87.8%提升至95.6%,验证了数据预处理和参数寻优能够有效提升诊断模型的性能。(4)最后,针对企业的一台龙门铣床的电主轴,搭建了故障诊断平台,验证了诊断方案的有效性。综上所述,本文提出的基于支持向量机的电主轴故障诊断方法,具有较好的实用性。
[Abstract]:Motorized spindle is one of the core components of CNC machine tools, and its performance directly affects the machining accuracy of machine tools. The fault diagnosis of motorized spindle can improve the reliability of NC machine tool, and it is of great significance. This paper takes the motorized spindle of NC machine tool as the research object, through empirical modal analysis and support vector machine and so on, carries out the research of its fault diagnosis method, and builds the fault diagnosis platform. The main work is as follows: (1) the structure of the motorized spindle is analyzed, and combined with the maintenance history data provided by the enterprise, the failure of the motorized spindle is statistically analyzed, and the faults in the motorized spindle are frequently occurred. As the object of this study, bearing failure mechanism, including vibration mechanism, failure mode, characteristic frequency and so on, is analyzed and studied, and the vibration characteristics of various typical faults are summarized. Empirical modal analysis, kurtosis and correlation coefficient discrimination are combined to extract the characteristic frequency of fault signal. Through the successful application in the experimental data, the validity of the signal processing method is verified. 3) combining with the empirical modal analysis and the autoregressive model, the feature vector construction method of the support vector machine is optimized. Through the analysis of data preprocessing methods (normalization and dimension reduction) and parameter optimization methods (exhaustive method, genetic algorithm and particle swarm optimization), The test accuracy of the diagnostic model of experimental data was raised from 87.8% to 95.6B, which proved that data preprocessing and parameter optimization could effectively improve the performance of the diagnostic model. Finally, for the motorized spindle of a gantry milling machine in the enterprise, the test accuracy of the diagnostic model was improved from 87.8% to 95.6%. A fault diagnosis platform is built to verify the effectiveness of the diagnosis scheme. To sum up, the method of motor spindle fault diagnosis based on support vector machine proposed in this paper has good practicability.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG659

【参考文献】