Comparison of state space controllers with two different observers on an identified water tank system

Abstract

In this paper, identification of a water tank system is performed, and controller and observer for identified system are constructed. Identification is done via Linear Black Box Identification which consists of unknown parameters; need to be estimated from measured data. Firstly, PRBS Signal is built using data, which is achieved by analyzing step response of the plant and considering that the change in the input signal may not be too large since we are going to use a linear model to indentify a nonlinear process. Considered linear models are ARX, ARMAX and OE, whose results are compared using different criteria such as simulation, correlation, and pole-zero plots. And the chosen model which is an Armax model will be given explaining the reasons why it is chosen. In the next step, controllability and observability tests are performed for chosen Armax model and observer and controller designs are performed. Predictive and filtering observers are constructed to observe the system and state space feedback controllers with and without integrators are constructed to control the system. Depending on the achieved observers and controllers, response of real system and simulation are compared to find best controller and best observer for the water tank system.