Economical and Reliable Adaptive Disturbance Cancellation

Efforts to apply active noise and vibration control commercially, while intensifying, continue to be hampered by the cost of the computers required to implement the active algorithms. Currently, the cost of wide-band control is prohibitive for lightly damped systems such as precision structures or highly reverberant acoustic spaces, even if the system has just one canceling actuator and one error sensor. The problem is compounded powerfully when multiple actuators and sensors must be coordinated. We propose to greatly reduce the computation cost and widen the applicability of adaptive cancellation by developing algorithms that require far fewer parameters to be adjusted, using infinite impulse response (IIR) system models and controllers.While the vast majority of work has been focused on finite-impulse response (FIR) modeling, IIR models have long been recognized as a potential solution to the high computational burden associated with FIR models. Many approaches have been tried, but their successful use has generally been restricted to systems with significant damping. We propose to solve the numerical ill conditioning and slow, unreliable convergence problems of current IIR methods by transformations of the system inputs and outputs that orthogonalize the correlation matrix of the inputs to the adaptive IIR system.