Process start/end event detection and dynamic time warping algorithms for run-by-run process fault detection

In semiconductor/FPD (flat panel display) manufacturing environments, APC (advanced process control) is a vital task for enhancing yield. The APC technology improves the productivity of equipments based on two main control mechanisms: fault management and R2R (run-to-run) recipe correction. This paper focuses on FDC (fault detection and classification) for the fault management, and proposes a run-by-run process fault detection method. The method consists of a process event detection algorithm to segment process data, and a pattern matching technique to classify the segmented data as normal or abnormal state. Experiments using the data collected from a RIE (reactive ion etching) process show that the proposed method is able to recognize normal/abnormal states with high accuracy.