Improving the noise robustness of reset controllers with an observer

The performance of linear control is limited by the waterbed effect, while the performance demands of industry are continuously increasing. Using non-linear control in the future is therefore unavoidable. Being non-linear, reset control is able to achieve better performance than linear control. However, the lack of robustness to measurement noise withholds most reset controllers from being implementable in industry. The high frequency content of the noisy measurement signal causes excessive resets, which degrade the reset controller performance. In this thesis, two control loop architectures are studied while using an observer to filter out the measurement noise. Prior research is exploited to derive stability conditions and the performance of the two loop architectures is analyzed. The results show that the tracking performance of a CgLp-PID and CR-CgLp-PID controller with noisy measurements is improved by reducing the pseudo-sensitivity over a broad frequency range. The overshoot in the transient response is removed and the performance is not influenced by modelling inaccuracies in the observer when reset feedback line filtering is used. The presented architectures thus improve the reset control systems, making them well implementable in practical setups.Mechanical Engineering | Mechatronic System Design (MSD