The Semiconductor manufacturing process for computer chips is a complex sequence of many hundreds of processing steps on silicon wafers.  At a few points in the process, sample wafers are inspected by inspection equipment and defect images reviewed by technicians.  While the basic findings from the inspection tool and technician are fed automatically into a database, other general observations are recorded simply as a comment of several sentences capturing key interpretations of the data. Because of various uses of the information, the comments are entered into multiple databases.

Lean mapping efforts highlighted several improvement opportunities related to these comments: they should be recorded only once and the format of the comment should be standardized. By applying TRIZ functional analysis to the problem, a radically different and innovative improvement was developed.  This improvement demonstrates Altshuller’s key premise that correctly reformulating the problem is vital to an inventive solution. A semantic analysis of all the words of typical comments revealed that these comments had specific functions with certain values to customers of the information.  So the recommended improvement was to eliminate the comment altogether and replace it with a set of categorical choices.  These new fixed field categories are much higher in value to the customers due the new analytic possibilities associated with categorical data.  This paper will show the benefit of this work to the specific semiconductor application as well how the improvements could be applied to any general situation where observations are recorded in sentence structures such as the medical clinical environment.

Details

Category: Abstracts

Read more: Reformulating a Semiconductor Information Problem with TRIZ