UC Berkeley

With rapidly rising demand for cloud storage, there is an increasing need for more efficient data storage options. Seagate Technology PLC unveiled the multi actuator drive technology in December 2017, a breakthrough that can almost double the data performance of the future generation hard disk drives.

A standard hard disk drive has an actuator arm controlling the position of eight read/write heads onto corresponding eight rotating disks. The multi actuator technology will equip drives with two actuator arms operating independently on the same pivot point, controlling four read/write heads each. Each actuator arm has a voice coil motor (VCM) connected in series with a micro actuator (MA). The read/write head is mounted at the edge of the MA. Each actuator arm has two operation modes, first - the track following mode - the actuator arm positions the read/write head onto a data track for data recording. This mode requires precision positioning of the read/write head for the smooth functioning of the hard disk drive. Second - the track seeking mode - the actuator arm sweeps through data tracks to settle onto a new desired track to initiate track following. This operation mode generally generates considerable vibration.

The multi actuator drive technology has induced new control challenges. Since, two actuator arms are operating independently on the same pivot timber, the control forces and torques generated by one actuator arm can adversely affect the performance of the other actuator arm. In this dissertation, feedforward data driven control design methodologies are presented to suppress the vibration imparted by the neighboring track seeking actuator arm onto a track following actuator arm. Also, joint feedback-feedforward data driven control design methodologies are presented to stabilize the closed loop system of the track following actuator, minimize the position error of the read/write head and, suppress the imparted vibration by a neighboring actuator arm.

Traditionally, in the model based design approach, models are fitted to frequency response measurements and, controllers are designed based on these models. In the data driven control design methodology, controllers are obtained directly based on the frequency response measurements of the system, avoiding model mismatch. Multiple frequency response measurements of the system and the imparted vibration are considered simultaneously in the design process, ensuring robustness of the control design.

Mixed H-2 H-infinity norm control design optimization problems, in the frequency domain, are formulated to obtain the feedforward and feedback controllers for the multi actuator drive. H-2 norm conditions are transformed into a locally convex objective function requiring an initial stabilizing controller for minimization. H-infinity norm conditions are transformed into necessary and sufficient convex constraints guaranteeing closed loop stability of the control system and providing an initial stabilizing controller for the H-2 norm minimization.

Application of input shaping techniques as an add-on feedforward suppression scheme for multi actuator drives is also discussed in this dissertation. Input shaping is used to suppress resonant frequencies of the coupling vibration from the excitation signal, minimizing the resultant vibration. The input shaping tool along with the data driven feedforward controller suppress up to 90 percent of the imparted vibration.