摘要
针对MPCA方法在具有多时段的间歇过程中的故障监测效果不佳的问题,提出一种新的多时段建模方法,首先根据各时间片上的主元个数不同,对过程进行模糊划分,然后利用K均值算法对样本数据聚类得到精确划分,最后按照划分结果在各阶段建立代表性统计分析模型对整个过程进行监控。将该方法用于半导体蚀刻过程的故障监测,并与MPCA方法进行了比较证明该方法具有良好的监控性能,能够及时准确及时的监测出引起产品质量发生变化的故障。
Point at the poor effects of MPCA method for faults monitoring in batch processes with multiple periods, this paper proposes a new multistage modeling method, first,according to the different number of the principal component on the each time slice to fuzzy on the process of division,then using k-means algorithm for precise division of sample data clustering, and finally according to the classification results, establish the typical statistical analysis model at each stage to monitor the whole process. The method for fault monitoring semiconductor etch process, ,and are compared with the MPCA method proved that the method has a good monitoring performance and can accurately and timely monitoring the change caused by product quality failures.
出处
《传感技术学报》
CAS
CSCD
北大核心
2015年第6期798-802,共5页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(61203020)
江苏省自然科学金项目(BK20141461)
关键词
多向主元分析
间歇过程
时段划分
故障监测
蚀刻过程
multi-way principal component analysis
batch process
time division
fault monitoring
ech process
