基于电流分析法的车床刀具状态监测技术研究

发布时间：2018-05-20 20:52

本文选题：刀具 + MCSA　；参考：《西南交通大学》2017年硕士论文

【摘要】：在现代生产加工中,车削加工占有了很大一部分比例,而刀具状态是对车削加工质量影响最大的因素之一。因此,对刀具状态进行监测,是提高生产质量重要手段,通过监测可以对刀具磨损进行预警,以便临界磨损时更换,从而大大提高生产效率,降低次品率。目前,刀具的状态监测手段包含了振动、切削力和电流监测等,其中电流分析法相比其他两种具有成本低、不需要对机床额外改造等优点,而被广泛运用于实际生产中。机床电机电流分析法(MCSA,Motor Current Signal Analysis)是一种基于电流信号的状态监测方法。本文以刀具磨损试验平台为对象,通过对主轴的驱动电流进行分析,确定车床切削状态,以及刀具的磨损状态。本文中,包含的主要内容如下:(1)通过对电流信号做小波包分解,并对各频率带能量进行对比,可以选择出包含刀具状态的频率带,对该频率带进行特征提取得到刀具敏感特征,在对特征筛选之后,便能够从中获取敏感特征作为刀具磨损判别的指标,最后采用分类器对其进行分类。(2)通过线性判别分析模型,对电流特征进行降维,重新得到一组包含了全部特征的新指标,优化后的特征具有最大的类间距离和最小的类间距离。通过对其进行B样条拟合,可以得到其状态曲线,建立磨损状态模型。(3)采用SVM对刀具是否磨损进行判别,通过刀具之前的特征,确定刀具临界磨损时的特征阈值,在将要磨损时发出报警,提示更换刀具。试验结果表明,通过提出的方法,能够很好的识别出刀具磨损状态,最终的识别准确率到达了 90%以上。
[Abstract]:In modern production and processing, turning takes up a large proportion, and cutting tool state is one of the most important factors affecting the quality of turning. Therefore, the monitoring of tool condition is an important means to improve the quality of production. The tool wear can be forewarned by monitoring, so as to change critical wear, thus greatly improve the production efficiency and reduce the rate of defective products. At present, the tool condition monitoring means include vibration, cutting force and current monitoring. Compared with the other two methods, the current analysis method has the advantages of low cost and no need for the extra modification of the machine tool, so it is widely used in practical production. Motor Current Signal Analysis) is a state monitoring method based on current signal. In this paper, the cutting state of lathe and the wear state of the tool are determined by analyzing the driving current of the spindle on the tool wear test platform. In this paper, the main contents are as follows: (1) by decomposing the current signal with wavelet packet and comparing the energy of each frequency band, we can select the frequency band which contains the tool state, and extract the feature of the frequency band to get the tool sensitive feature. After screening the features, the sensitive features can be obtained as the discriminant index of tool wear. Finally, the current characteristics are reduced by linear discriminant analysis (LDA model), which is classified by classifier. A new set of indexes containing all the features is obtained again. The optimized features have the maximum inter-class distance and the minimum inter-class distance. By B-spline fitting, the state curve can be obtained, and the wear state model. (3) the tool wear is judged by SVM, and the characteristic threshold of the tool critical wear is determined by the characteristics before the tool. Issue an alarm when the tool is about to be worn and prompt to change the cutting tool. The experimental results show that the tool wear state can be recognized well by the proposed method, and the final recognition accuracy is more than 90%.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG51

【参考文献】