基于改进InDBSCAN算法的批量钻削工序质量增量聚类分析

Incremental cluster analysis for batch drilling-quality based on improved InDBSCAN algorithm

针对批量钻削工序质量检测问题,采用声发射传感器采集工序加工过程中的声发射信号,提取其时域统计特征,构造工序过程信号的特征向量,根据密度带噪声的空间增量聚类算法(InDBSCAN)对工序过程中的声发射信号特征向量进行增量聚类,以分析批量工序质量。考虑到插入数据点在促成新类创建的同时可能引起已存在的不同类合并的情况,改进InDBSCAN算法。实验结果表明:改进的InDBSCAN算法使插入数据点的增量聚类更加合理,工序质量分布状况检测准确率达84.03%。

Aiming at monitor and analysis on batch drilling-quality,an acoustic emission sensor was used to collect the acoustic emission signal,extract statistic characteristics and then construct the signal characteristic vector.An improved incremental density based spatial clustering algorithm of time-domain applications with noise（InDBSCAN） was put forward to analyze the distribution law of batch drilling-quality indirectly.Take new data insertion into consideration.Because some of the original clusters could be remerged when the new cluster was created,and so the InDBSCAN algorithm was modified.The results show that the conclusion of incremental cluster analysis is more reasonable by the improved InDBSCAN algorithm and the detection accuracy of batch drilling-quality is up to 84.3%.