Linear Quadratic Gaussian Controllers for Feedforward Active Noise Control : Pushing Performance and Moving Towards Adaptive Control

Sammanfattning: Active noise control is a research area focused on using destructive interference of sound fields to attenuate undesired noise. Methods for active noise control are best suited for low frequency noise, as the complexity of the problem grows rapidly with frequency. Coincidentally, passive means of damping have the opposite quality in that they work better for higher frequencies and become bulky and impractical for low frequencies. Applications for active noise control range from fan noise in ducts, noise-cancelling headphones and noise in cars to propeller induced aircraft cabin noise. In this comprehensive summary, the underlying principles of active noise control are presented and the control problem is discussed. Several aspects of the control system are introduced to give an introduction to the research papers that are the basis of this licentiate thesis. The work behind the thesis is focused on a Multiple-Input Multiple-Output (MIMO) Minimal Mean Square Error (MMSE) Linear Quadratic Gaussian (LQG) feedforward controller. This controller is shown to achieve uniform damping in an extended region in space and push the upper frequency that can be controlled. The influence of different design variables has been investigated, and the properties of the control path analyzed with consideration of its ability to suppress noise of prescribed spectral properties over an extended region. In this context, it has been shown how to use the reproducibility of the primary noise path by the control path as an indication of achievable performance for a given control system. Finally, the controller has been adapted to follow changes in the primary noise statistics, an approach that seems promising to considerably raise the performance of the controller.