Neural Networks for Disturbance and Friction Compensation in Hard Disk Drives

In this paper, we show that the tracking performance of a hard disk drive actuator can be improved by using two adaptive neural networks, each of which is tailored for a specific task. The first neural network utilizes accelerometer signal to detect external vibrations, and compensates for its effect on hard disk drive position via feedforward action. In particular, no information on the plant, sensor and disturbance dynamics is needed in the design of this neural network disturbance compensator. The second neural network, designed to compensate for the pivot friction, uses a signum activation function to introduce nonlinearities inherent to pivot friction, thus reducing the neural network's burden of expectation. The stability of the proposed scheme is analyzed by the Lyapunov criterion. Simulation results show that the tracking performance of the hard disk drives can be improved significantly with the use of both neural networks compared to the case without compensation, or when only one of the networks is activated.