Performance enhancement of hard disk drives through data-driven control design and population clustering

Abstract

This paper presents a frequency-domain data-driven control design method for improving the robustness and performance of a large population of hard disk drives. First, a multi-rate notch filter design method is presented to prevent the aliasing of post-Nyquist frequency resonant modes. The tradeoff between aliasing and system phase loss due to the multi-rate notch filters is investigated and discussed using a set of real HDD plant data. A data-driven servo control design framework for a dual-stage HDD system is then proposed, where the stability, robustness, and performance properties of the closed-loop system are simultaneously computed and optimized. The validity of the proposed notch filter and controller design methods is evaluated through a set of time-domain simulations. Finally, a clustering-based control optimization method is discussed to further improve the population-level system performance. Simulation results indicate that the proposed data-driven controller with frequency domain plant clustering is an effective approach for designing high-performance servo controllers for HDDs.