Adaptive control of a disk drive arm's microactuator using neural networks

Accurate positioning of a magnetic disk drive's read/write head over ultrathin tracks is limited by the vibrations of the disk drive arm. Microactuators placed at the free end of the disk drive arm have been built and used to place the read/write head finely in conjunction with the voice-coil motor rotating the disk drive arm coarsely. This dissertation proposes to predict the vibration of the disk drive arm by computing the drag forces acting on the disk drive arm. A neural network trained by a Kalman filter adapts the prediction of the vibrations to keep up with the change in the dynamics of the disk drive.; Von Karman's approximate solution of the velocity profile in the boundary of a rotating disk is used to compute the drag forces acting on the disk drive arm. The vibrations of the arm are computed by modelling the disk drive arm as a constant cross sectional area beam and solving the Euler-Bernoulli beam equation for transverse loads. Solving the fourth order linear partial differential Euler-Bernoulli beam equation exactly during a track seek duration of less than 20 milliseconds is impractical. Hence the partial differential equation is approximately solved by finding an equivalent set of difference equations and solving the difference equations instead of the differential equation.; An extended Kalman filter is used to predict the deflection of the beam using the difference equations and Von Karman's velocity profile in the boundary layer above the rotating disk. Application of geometry is used to compute the displacement of the microactuator necessary to correctly place the read/write head once the deflection has been predicted.; A three layer neural network is used to predict the deflection residuals of the first Kalman filter in order to improve the prediction of the deflection. A second Kalman filter is used to periodically train the neural network so that fixed weights can be found such that performance specifications are met at the current state of the disk drive's dynamics. Computer simulations show that the proposed adaptive control can be used to help increase track density.