Abstract:
In Active Noise Control (ANC) systems, the online modelling of the Secondary Path is an important topic of research. This functionality, achieved by the injection of modelling noise, is vital to the successful stable operation for noise attenuation, particularly when the Secondary Path is time varying due to its acoustic nature. A notable weakness is that there is a lack of experimental data available. The research discussed in this thesis presents experimental data obtained by replicating existing online modelling methods in NI compactRIO-9082 hardware. This thesis investigates the online modelling performance and noise attenuation of Akhtar's MFxLMS (Modi ed - FxLMS) method. The experimental results produced demonstrated two things. Firstly the MFxLMS fails to operate on a higher control step size parameter compared to the FxLMS al-gorithm, most likely due to complications arisen from implementing in a physical environment. Secondly, the restriction of the modelling noise amplitude is a requirement for stable and accurate Secondary Path modelling. Two other existing algorithms are also implemented. Each algorithm has its own criterion for injecting and suspending modelling noise. Extensive testing of these two methods shows that one system is better at suspending modelling noise once filter convergence is achieved. This prevents the Secondary Path estimate from destabilising even though modelling accuracy is reduced. It is suggested that while modelling accuracy and modelling stability are different from each other, modelling stability is a more important factor to consider as modelling accuracy is dependent on stability. Two methods of online secondary path modelling are proposed as improvements from the ANC systems already tested in this research investigation. It is shown that the appropriate ne tuning of the FwFxLMS (Filtered-w FxLMS) algorithm improves noise attenuation convergence at the expense of operational stability. The second proposed method incorporates a third modelling filter, an idea initially proposed by Zhang. The experimental results show that the using an additional modelling filter improves the quality of the error signal used for the online modelling of the Secondary Path. This in turn improves the modelling accuracy of the Secondary Path and the overall stability of the ANC system.