Parametric Uncertainty Identification and Model Reduction for Robust H2 Synthesis for Track-Following in Dual-Stage Hard Disk Drives

This paper presents a systematic, semi-automated method for identifying parameters and parametric uncertainty for a set of dual-stage hard disk drives. A modal analysis technique is selected to extract parameters from a batch of frequency response data. In order to avoid redundancy in modal parameters, a method of data fusion is incorporated into the modal analysis to find common parameters for multiple frequency response functions. Next, a model truncation methodology is presented as an alternative way to eliminate redundant parameters from a multiple-input-multiple-output system model. In both model reduction methods, optimization algorithms are employed to minimize the error associated with the reduced order model. Finally, convex optimization and singular value decomposition are employed to obtain a low-order, minimally conservative approximation of uncertain parameters. The result is a reduced order state space model with parametric uncertainty to be used in robust H₂ control synthesis for a track-following hard disk drive servo.