Self-organizing Map-based K-means Clustering for Stability Analysis of Product Quality in Packaging in Semiconductor Manufacturing

The “packaging” plays an important role in semiconductor manufacturing. Among the multiple test-passed steps in packaging production, the step of “wire-bonding” is the most complicated and critical one since there are many tuning parameters, such as force, current and time, etc. needing to be set-up by operators in order to perform good bonding. Several key perform indexes, for example, wire pull, ball shear and ball height, are used to measure the product quality. This paper presents the development of applying SOM (self-organizing map) based k-means to the data-based classification for product quality. Namely, the product quality can be divided into a number of categories in terms of huge historical quality related data. In addition, SOM based K-mean clustering model, algorithms and the case studies with real-production data are addressed. The concluding remarks and future work are given in the final section.