机床主轴振动特性与工件材料类型关联性研究

Study on relevance between vibration characteristics of machine tool spindle and material types of machined workpieces

为了能够实现机床运行状态的远程监控并对运行状态实施智能判断,必须准确分析判断机床当前加工的材料类型。利用小波包能量法进行特征提取,利用BP AdaBoost技术建立了集信号采集、处理、分析与判断的混合模型,对切削振动信号特征进行分析与识别,从而推断出加工材料的类型。该方法通过采集机床在切削加工过程中主轴的三向振动信号,利用小波包分解得到振动信号不同频段的能量值并将其作为特征向量,最后通过BP AdaBoost对特征向量进行训练和验证,得到准确的判断结果。结果表明,基于BP AdaBoost神经网络相比单一BP算法和PNN算法有较高的辨别率,其平均准确率可以达到92.12%,能够实现对机床加工材料种类的判别。

In order to realize remote monitoring of the running status of the machine tool and implement intelligent judgment on its running state,it is necessary to accurately analyze and judge the types of materials currently processed by the machine tool.The method of wavelet packet decomposition(WPD)is used for feature extraction.The BP AdaBoost technology is used to establish a hybrid model of signal acquisition,processing,analysis and judgment.This model is used to analyze and recognize the features of cutting vibration signals,so as to infer the types of machined materials.In the model,the three⁃direction vibration signals of the spindle are collected in the process of cutting,and then the energy values of the vibration signals at different frequency bands are obtained by WPD and taken as the eigenvectors,which are trained and verified by BP AdaBoost to acquire accurate judgment results.The results show that,in comparison with the model based on BP algorithm and that based on the PNN algorithm,the model based on BP AdaBoost combining the neural network has higher discrimination rate(the average accuracy rate can reach 92.12%),and can fulfill the discrimination of the types of materials currently processed by the machine tool.