Vibration Signal-Assisted Endpoint Detection for Long-Stretch, Ultraprecision Polishing Processes

Abstract The research reported in this article is concerned with the question of detecting and subsequently determining endpoint a long-stretch, ultraprecision surface polishing process. While detection has attracted much attention for several decades chemical-mechanical planarization semiconductor wafer processes, uniqueness process under our investigation calls novel solutions. To tackle challenges, we develop both an offline model online method. functional regression that relates vibration signals to roughness, whereas procedure change-point method detects energy turning points signals. Our study reveals number insights. shows clearly progresses three states, including saturation phase, over which action could be substantially shortened. real-time when break cycle institute follow-up inspection, rather than letting machine engage overpolishing too long. When implemented properly, sets insights corresponding methods lead substantial savings time significantly improve throughput such processes without inadvertently affecting quality final polish.