An efficient approach to l(1) optimal control and its application to hard disk tracking control.

摘要

An approach to the tracking control problem of the head-suspension assembly of a hard disk drive is presented.; A dynamic model of the head-disk assembly is obtained from FEM analysis with a nonlinear spring model of the air bearing between the head and the disk. A control plant is built by discretizing a dynamic model which contains the first torsional and in-plane bending mode as well as the rigid body mode, modes which are the dominant residual error sources in tracking applications.; A robust tracking controller is presented that provides an improved head access time for the control plant. The controllers are based on an {dollar}lsb1{dollar} control design method using the linear programming optimization approach (FMV approach) in order to provide robust performance for finite time tracking. These {dollar}lsb1{dollar} controllers can successfully reduce the residual vibration of the flexible modes induced by the system's seek motion and can provide closed loop system robustness to input and structural uncertainties. The order of {dollar}lsb1{dollar} controller obtained by the FMV approach, however, for system with uncertainties becomes high.; As remedy to this deficiency, a new and efficient optimization approach for determining a low order {dollar}lsb1{dollar} optimal controller is presented. In the proposed method, the Q found from the closed loop impulse response equation {dollar}rm(Phi=H-UQV){dollar} is assumed to be a matrix of IIR filters and the coefficients of Q are used as optimization parameters to minimize {dollar}VertPhiVertsb1.{dollar} For multi-block problems, the Q domain optimization method requires only a small number of optimization variables and can converge to the desired optimal value more efficiently than is the case for linear programming optimization. Two examples are discussed that illustrate how this method can construct a lower order {dollar}lsb1{dollar} controller at the same optimal norm as that found from a linear programming optimization approach.