The multiple reference principal component least mean squares algorithm: a projection based approach

The principal component least mean squares algorithm (PCLMS) is an elegant adaptive control algorithm for cancelling a tonal disturbance signal in active control applications, such as active noise control and active vibration isolation control. The algorithm removes the spatial correlation between the actuator inputs and the error sensor outputs to enable fast convergence of the adaptive controller. However, a drawback of the PCLMS algorithm is that it can only suppress a disturbance signal which contains a single frequency component. The contribution of this paper is that we present a numerically robust projection based approach in which the PCLMS is extended with the ability to suppress a disturbance signal which contains multiple frequency components. The potential of the algorithm is demonstrated by multi-tonal control on a realistic model of a real-time vibration isolation set-up. The algorithm is shown to outperform the traditional filtered-x least mean squares algorithm.