基于BEMD与LSSVM的大型磨床磨削颤振在线检测方法研究

发布时间：2018-04-01 01:08

本文选题：磨削颤振　切入点：时变信号　出处：《浙江理工大学》2017年硕士论文

【摘要】：磨削加工是现代机械制造业中不可或缺的一种用来获取高精度、低粗糙度的零件加工表面的工艺方法。对磨床进行状态实时监测和故障识别诊断来确保磨床长期稳定可靠运行具有有重大现实意义和产业价值。需要注意的是,在加工过程中,磨床会进入颤振的状态,从而引发一系列负面影响。因此可靠的颤振监测和识别技术是必不可少的,以实现磨床振动状态的实时监测。以傅里叶变换为理论基础的传统时频信号处理方法不适用于非线性、非平稳和多维的磨床振动输出信号。二维经验模态分解(Bivariate Empirical Mode Decomposition,BEMD)扩展了EMD的能力,能将二维复值信号分解为一系列零均值的旋转成分。BEMD不仅能描述非线性动力学行为,而且能节约计算时间,并移除计算中由于假设和人为原因产生的失真。其在检测初始故障方面表现出更强的能力,能有效地分析并提取非平稳、非线性磨床颤振信号特征。本文以KD4020X16数控龙门导轨磨床为研究对象,根据磨床自身的动静态特性搭建了颤振检测试验平台,进行了磨削参数多水平试验。利用IEPE压电加速度传感器和配套的TST5912动态信号分析仪对振动信号进行采集和保存,得到不同磨削参数设定下的80组实验样本数据,其中包括45组平稳磨削振动信号和35组颤振磨削信号。本论文对实验过程中采集到的砂轮主轴X和Z方向的振动信号进行信号重构,进行BEMD处理得到多阶BIMF分量;利用基于相关系数的真实固有模态函数提取准则筛选出真实BIMF;提取出对颤振信号敏感的指标量—峰峰值、实时方差、峭度以及瞬时能量,分别进行求和与归一化处理形成颤振特征向量;最后以最小二乘支持向量机作为(Least Square Support Vector Machine,LSSVM)智能化模式分类器对随机选取的55组样本数据的特征量进行训练,得到颤振检测识别模型,以剩下的25组样本数据作为检验样本,对识别模型进行检验和判断,验证其准确率及可行性。证明了基于BEMD与LSSVM的方法具有较好的识别率。通过上述方法,建立了大型数控磨床磨削颤振检测软件,验证了其实时监测磨床振动状态的可行性。
[Abstract]:Grinding is an indispensable method in modern mechanical manufacturing industry to obtain high precision. Process method for machining surface of parts with low roughness. It is of great practical significance and industrial value to ensure the long-term stable and reliable operation of grinding machine by real-time monitoring and fault identification diagnosis of grinding machine. Grinding machines enter a flutter state, causing a series of negative effects. Reliable flutter monitoring and identification techniques are therefore essential. In order to realize the real-time monitoring of grinding machine vibration, the traditional time-frequency signal processing method based on Fourier transform theory is not suitable for nonlinear. The two-dimensional empirical mode decomposition extends the ability of EMD to decompose the two-dimensional complex signal into a series of rotating components with zero mean value. BEMD can not only describe the nonlinear dynamic behavior. Moreover, it can save calculation time and remove the distortion caused by assumptions and human causes. It has a stronger ability to detect initial faults and can effectively analyze and extract non-stationary. Based on the dynamic and static characteristics of the KD4020X16 CNC gantry guideway grinder, a flutter detection test platform is built in this paper, which is based on the dynamic and static characteristics of the grinder itself. The vibration signals were collected and saved by IEPE piezoelectric accelerometer and TST5912 dynamic signal analyzer, and 80 sets of experimental data were obtained under different grinding parameters. This paper reconstructs the vibration signals in X and Z directions of the grinding wheel spindle collected during the experiment, and obtains the multi-order BIMF component by BEMD processing, which includes 45 sets of stationary grinding vibration signals and 35 sets of chatter grinding signals. The real BIMF is selected by using the real inherent mode function extraction criterion based on correlation coefficient, and the peak peak value, real time variance, kurtosis and instantaneous energy sensitive to flutter signal are extracted. Finally, the least square support vector machine is used as the intelligent pattern classifier for least Square Support Vector machine to train the characteristic quantity of 55 groups of randomly selected sample data. The flutter detection and identification model is obtained. The remaining 25 sets of sample data are used as test samples to test and judge the identification model. The accuracy and feasibility of the method are verified. The method based on BEMD and LSSVM has a good recognition rate. Through the above method, a large NC grinding machine grinding chatter detection software is established, and the feasibility of real-time monitoring the vibration state of grinding machine is verified.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TG580.6

【参考文献】