A micro-view-based data mining approach to diagnose the aging status of heating coils

The aging status of heating coils dramatically influences the semiconductor process with respect to the operating fluency and the cost of non-warning breakdowns. Diagnosing the health status of heating coils is a critical issue in the manufacturing industry, but very few studies have dedicated their efforts to discovering relevant remedies. This study proposes a novel observation perspective that grasps the micro view of data information from real processes, and does not lose its effectiveness even if the recipes vary. The proposed aging measuring includes the following main procedures: firstly, acquiring specific process flows composed of recipe steps; secondly, analyzing the local areas corresponding to each process flow; thirdly, capturing sensing values and the corresponding setting information related to the specific process flows; fourth, generating a local feature in a time interval corresponding to each of the process flows according to the corresponding local areas; finally, generating an aging trend on the basis of the local features, and determining whether to send a replacement early-warning. The experiments show that the proposed approach to the aging diagnosis can effectively offer a replacement warning before the heating coils run down.