Some issues with the intelligent control of industrial processes

Abstract

An intelligent controller was developed with neural networks for the control of a multicapacity liquid level system. The neural controller was trained off-line with data acquired from the knowledge of the water tank system dynamics. The performance of the controller was examined in relation to some of the robustness issues in the control of real-life plants. A comparison was made using a recently developed robust adaptive pole-assignment algorithm for the control of the same unknown plant.