Yield improvement analysis with parameter-screening factorials

This paper presents a technique for the critical parameter analysis of the disk drive manufacturing process. The objective of the work is to improve the manufacturing yield by tuning the parameters that significantly affect the yield. Several techniques were studied including the sensitivity analysis framework, which is currently used at several disk drive plants. From our initial experiments, we found that the sensitivity analysis results were not sufficiently good and the interactions between parameters were not identified. We then designed a new technique based on factorial designs, the parameter-screening factorials algorithm. Our method can work with a large number of inputs within reasonable computing time, and can identify both the parameter and the interaction effects. The results can be obtained more quickly and are better in comparison with the currently used technique. Moreover, by applying the technique to the full list instead of the pre-selected list of the manufacturing parameters, we discovered that the parameters watch list previously identified by the experts should be adjusted to include some extra parameters. After the results were validated by the experts, we designed software that automates the critical parameter analysis process. The software should greatly benefit the daily yield analysis at the disk drive manufacturing plant greatly.

: We design a critical parameter analysis technique based on principal component and factorial designs concepts. The algorithm allows a large number of inputs to be processed at once lifting the limitation of input size exists in most software used today. The interaction between parameter can also be studied. The algorithm can accurately identify both the parameters and the interaction effects of the attributes that affect the yield. The results suggested that the current parameter watch list used by plant engineers should be extended to include some extra meaningful parameters, which were left out initially due to the limitation of the currently used analysis software.Figure optionsDownload as PowerPoint slideHighlights
► The Parameter-Screening Factorials algorithm is designed based on the principal component concept and factorial design.
► The algorithm can identify the critical parameters and their interaction effect in the disk drive manufacturing process.
► The algorithm can work with a large number of inputs within reasonable computing time.
► When adopted in the actual manufacturing, the algorithm has successfully picked out extra critical and meaningful parameters that were not originally in the engineer's watch list.