The Sampling-Based Sensitivity Analysis Model for Yield Improvement in HDD Manufacturing

This paper proposes a design of a yield improvement analysis tool for disk drive manufacturing. The emphasis has been put on optimizing the HGA manufacturing process. The manufacturing data were obtained from the data warehouse for our study. We propose a design framework that consists of a set of various algorithms for automatic data analysis. The framework is verified by the engineers and analysts from a HDD manufacturing company. This research mainly focuses on 2 modules in the framework, which are the analysis of the critical parameters and the yield improvement. In the first module, the sensitivity analysis techniques are explored and adapted to identify the critical machine parameters and product specification that affect the yield variation. We also study and develop the methods used in the manufacturing for yield improvement analysis, namely, mean shift, variance reduction, and spec tightening in the second module. In addition, we present initial experimental results in ranking the parameters according to their impact to the yield. From the results, the top ranked parameters obtained from our algorithm are the same set that are often monitored by the experienced engineers.