基于聚类的轮廓数据质量监控方法研究

Profile Monitor Based on Clustering Method

轮廓线的变点识别是质量管理的研究热点之一,当前研究多以轮廓整体变化为识别对象,而对局部变化问题研究相对较少,且更少有在发现变异时间的同时能够寻找到变化区域在个体轮廓曲线上位置的系统方法。本文针对轮廓线局部变化识别问题,提出基于小波变换和聚类分析的方法。通过仿真性能评价,并与现有方法进行比较,结果显示本方法能够在更小的差异度检测出变化并准确定位变化区域。在文章的末尾,本文采用了一个实例对该方法的效果进行验证。

Change-point detection of profiles has been a hot topic in quality management. Most of the literatures focus on the global change detection while research on local change problem is relatively few, and the ones that find change time as well as change area are extremely less. To solve local change problem, our paper proposes a method based on wavelet transform and cluster method. In this paper, we do a performance evaluation using Matlab and compare its result with WANOVA method. It shows that our method can detect change-point and locate change area under smaller difference between profiles.