基于核主成分分析BP＿Ada Boost算法的数控铣床故障诊断

TOOL WEAR STATE MONITORING BASED ON WAVELET PACKET BP＿ADABOOST ALGORITHM

刀具是数控铣床加工过程的关键零部件,其长期处于高速加工状态极其容易出现故障。针对数控铣床加工过程中刀具的磨损状态数据少、诊断效率低、维护成本高、缺乏有效的诊断方法的问题,提出了利用小波包分析与核主成分分析提取特征,然后利用BP_AdaBoost算法对刀具磨损状态进行诊断的方法。通过在数控铣床的加工工件与其夹具间安装测力仪及安装加速度传感器,来采集立铣刀振动信号与切削力信号;然后对振动信号与切削力信号进行小波包分析处理,将处理好的信号进行核主成分分析(KPCA),降维以后作为立铣刀磨损状态的特征向量;最后利用得到的特征向量训练和验证BP_AdaBoost分类模型。实验结果表明BP_AdaBoost算法比SVM算法能更有效实现对数控铣床的刀具磨损状态的评估。

Tools are the key parts in the process of NC milling machine.They are in high-speed processing for a long time and are prone to failure.Aiming at the problems of less tool wear state data,low diagnostic efficiency,high maintenance cost and lack of effective diagnostic methods during CNC machine tool processing,A method of extracting features by wavelet packet analysis and kernel principal component analysis,and using BP_Ada Boost algorithm to diagnose tool wear state is proposed.The tool vibration signal and the cutting force signal are collected by installing an acceleration sensor on the machined workpiece of the numerical control machine tool and a force gauge on the workbench;Then the wavelet packet decomposition is performed on the signal to pass the signal through the low-pass filter and the high-pass filter of different dimensions,so that the conditional selection can be performed to form the energy value corresponding to the different frequency bands.The data after the dimension reduction of the kernel principal component analysis is taken as the characteristic parameter of the tool wear state;Finally,the eigenvectors are used to train and validate the BP_AdaBoost classification model.The experimental result shows that the BP_Ada Boost algorithm can effectively diagnose the wear state of the tool in CNC machine tools compared with the SVM algorithm.