Fuzzy-based risk priority number in FMEA for semiconductor wafer processes

Abstract

Semiconductor wafer manufacturing processes are considered to be complex and require considerable manpower and financial investment. Effective control to promote wafer yield is thus a very important issue. Many computational problems in traditional failure mode approaches, such as variable selection and data collection, are far too dependent on the experience of engineers, with a lack of specifically quantified values. Therefore, there are significant differences between the results of research and the actual processes. Although several researchers have revised the failure mode computation, the occurrence results continue to differ from the actual quantified values. The potential failure mode in semiconductor wafer manufacturing processes is effectively discussed using linguistic fuzzy variables to replace the severity and detection in the failure mode for re-calculation and sorting, along with the occurrence acquired from the wafer processes yield change. Based on the results of engineering experiments, the use of a risk priority number can effectively overcome the problem of a lack of objectivity in the traditional failure mode and effect analysis, as well as accurately distinguish the priority of the key failure factors so that the research results become more complete.