Comprehensive plug and play FD system realized Predictive Maintenance

This paper presents a comprehensive plug & play Fault Detection (FD) system that can be easily deployed fab-wide. By collaborating with Equipment supplier (ES) and Fault Detection and Classification (FDC) supplier, we have established FD models that can monitor each component of equipment comprehensively. The FD models convert recipe dependent parameters into independent ones defined by using set-point values and trigger variables obtained from the components. Since the FD models can continuously monitor tools regardless to recipe variation, it is suitable for applications of high-mix fab where various recipes are running in a tool. Applications of these models to Cu-CMP process enabled us to avoid unscheduled maintenances (UM) by predicting appropriate timing of Predictive Maintenance (PdM) and reduce up to 35% of scrapped wafers.