Improving the Consistency of the Failure Mode Effect Analysis (FMEA) Documents in Semiconductor Manufacturing

Digitalization of causal domain knowledge is crucial. Especially since the inclusion in data analysis processes helps to avoid biased results. To extract such knowledge, Failure Mode Effect Analysis (FMEA) documents represent a valuable source. Originally, FMEA were designed be exclusively produced and interpreted by human experts. As consequence, these often suffer from consistency issues. This paper argues that due transitive perception relations, discordant merged information cases are likely occur. Thus, we propose improve as step towards more efficient use knowledge. In contrast other work, this focuses on relations expressed documents. end, based an explicit scheme types inconsistencies derived perspective, novel methods enhance quality presented. Data improvement will significantly downstream tasks, root cause automatic process control.