Session: 10-01 Tribology, Dynamics and Servo control of Nano-Micro systems

Paper Number: 118829

118829 - Predicting Drive-to-Drive Frequency Response Function Variation 

For reliable Hard Disk Drive (HDD) operation, the servo-controller must be robust against drive-to-drive mechanical plant transfer function, i.e., Frequency Response Function (FRF) variation. In this study, we combine a Machine Learning–based data analysis of measured, high-volume population FRFs of an HDD (Product A) with the numerically simulated FRF of another, similar HDD (Product B) to predict the drive-to-drive FRF variation of Product B. Firstly, we obtain measured population FRFs of Product A from HDD high-volume manufacturing builds. Next, we employ the LSCF (Least-Squares Complex Frequency-domain) method to decompose each FRF from the population of Product A into a set of decoupled, single degree-of-freedom modes. We utilize a clustering algorithm to identify modes with similar characteristics and accordingly assign a mode-shape label to every mode from every FRF in the population. Bivariate analysis is then performed on the clustered modes to select “boundary drives”, which represent the extreme response of Product A. Finally, we apply the observed variation from the boundary drives of Product A to the numerically simulated FRF of Product B (computed via Finite Element Analysis) to predict the drive-to-drive FRF variation of Product B. Our approach would enable generation of a synthetic population boundary FRF dataset for HDD servo-control design optimization and validation against drive-to-drive FRF variations.

Presenting Author: Siddhesh Vivek Sakhalkar Western Digital

Presenting Author Biography: Dr. Siddhesh Sakhalkar is a Mechanical Simulation Engineer at Western Digital in San Jose, California, USA. In this capacity, he utilizes advanced computational techniques and machine learning/statistical tools to optimize the design and dynamics of next-generation hard disk drives that power the world’s cloud data centers. He holds a Ph.D. in Mechanical Engineering from the University of California, Berkeley. His doctoral research involved developing numerical models to study nanoscale lubricant flow and heat transfer at the interface between the magnetic recording head and the storage disk in hard disk drives.