Applying the Nest-Site Selection (NeSS) Algorithm for Yield Improvement in Disk Drive Manufacturing

This work applies a honeybees-inspired Nest-Site Selection (NeSS) optimization algorithm to a clustering problem. We study an actual data set collected from a hard disk drive (HDD) manufacturing process. The dataset contains attributes such as date of production, materials used, and manufactured location as well as machine parameters of the Head Gimbal Assembly (HGA) production lines. The HGA products can be classified as pass or fail based on the attributes and machine parameters. In this work, we first pre-process the data using the regular data cleaning techniques. Subsequently, a clustering technique is performed to group the machine parameters into two clusters based on a measure of data similarity such as distances between clusters or dense areas of data. In the beginning, data will be randomly assigned to the clusters. Then, in each iteration, the NeSS algorithm will reassign the data to more related clusters until a stopping criterion is reached. Finally, different profiles of the passed and failed HGA products will be created. The methodology and results in this work should help the data analysts in a HDD company to understand problems on the manufacturing line and improve yield by monitoring the right set of parameters and configuring the values to fit the “passed” profile.