An Unsupervised Diagnosis for Process Tool Fault Detection: The Flexible Golden Pattern

摘要

The flexible golden pattern (FGP) algorithm uses a patented technology of empirical scoring to detect abnormal behavior for semiconductor processing equipment or a specific processing chamber during wafer production. This algorithm does not entirely rely on manual extraction of features from data acquired on each tool. It is able to automatically select good pattern indicators from raw (temporal) signal traces. It is able to diagnose unusual behavior disregarding specificity proper to a recipe or even a chamber or even a tool if the algorithm is calibrated for such a purpose. The algorithm does not need any complicated parameter settings; the diagnosis is established by comparison of the normal process behavior to the abnormal one.