Model based control of a water tank system
Deprecation warning
This website is now archived. Please check out the new website for Centre for Intelligent Systems which includes both A-Lab Control Systems Research lab and Re:creation XR lab.
On this site, please find older content as well, such as list of older completed projects and publications.
Neural network with a specific restricted connectivity structure is used to identify a model of a real-life process. Parameters of the identified model are used to design a controller based on dynamic feedback linearization. The designed neural network based controller is verified on mathematical model within MATLAB/Simulink environment and applied to the real-time control of a plant. The static error is eliminated retuning input signal in the steady-state mode. Liquid level tank system was chosen as a case study to illustrate the applicability of the proposed approach. Experimental results have shown a good performance of the proposed technique. The designed controller is capable of tracking the desired water level for all set points with high degree of accuracy and without significant over/undershoot.