Abstract: Aiming at the modeling problem of tool wear under different machining parameters,a tool state recognition method based on joint distribution adaptation of multi-domain features was proposed to improve the tool state recognition model's generalization performance and recognition accuracy.For the data of sensor signals which under different machining parameters in the cutting process,the features were extracted in time domain,frequency domain as a well as time-frequency domain.Then the differences between the features were reduced by joint Distribution Adaptation Algorithm (JDA),and the adapted features were input to K-Nearest Neighbor (KNN) classifier to identify the wear state.The experimental results showed that the proposed method could identify the tool wear state under different machining parameters effectively,and the average recognition accuracy could be improved by more than 12%.It has great generalization and identification accuracy.

Key words: multi-domain feature, tool wear, joint distribution adaptation, K-nearest neighbor classifier

摘要： 针对不同加工参数下的刀具磨损建模问题,提出一种基于多域特征联合分布适配的刀具状态识别方法,以提高刀具状态识别模型的泛化性能与识别精度。对切削过程中不同加工参数下的传感器信号数据,提取时域、频域、时频域特征,通过联合分布适配算法(JDA)缩小特征间的差异,适配后的特征输入到K-最近邻分类器(KNN)进行磨损状态识别。实验结果表明,该方法能够有效识别不同加工参数下的刀具磨损状态,平均识别精度可提升12%以上,具有较好的泛化性能和识别精度。

关键词: 多域特征, 刀具磨损, 联合分布适配, K-最近邻分类器