Feedforward Control Based on Neural Networks for Hard Disk Drives

We present a novel feedforward control based on neural networks to attenuate the effect of external vibrations on the positioning accuracy of hard disk drives. The neural network compensator, which is an add-on function on top of nominal feedback control, uses the accelerometer signals obtained from a sensor to detect external vibrations. Our feedforward control can be regarded as a nonlinear finite impulse response (FIR) that corresponds to linear FIR when the basis function of the neural network is linear. By neural network learning, the tracking performance of hard disk drives can be improved with no information on disturbance dynamics or sensor model. We have analyzed the stability of the proposed scheme by the Lyapunov criterion. Here, we give simulation results to demonstrate that our control scheme can eliminate the effect of external disturbances on positioning accuracy.