Identification of process models in closed loop
Abstract
In control engineering, there are two types of models known: Open loop and Closed loop. In the open loop model, the input is altered manually by an actuator based on the output desired. But in a closed loop system, the output is automatically adjusted to the setpoint using the feedback. The controller manipulates the input to reduce the error between the setpoint and the desired output, and this is done by tuning the controller appropriately. In the demonstration case, the process model is first constructed for analysis in both open and closed loop. A PRBS signal is applied to excite the systems to get the process data.
Identification of process models has advanced leaps and bounds in recent years. An excellent example would be the DSR toolbox, present in MATLAB, which provides system matrices in state space model of any process. It is proven that model identification is imperative, as it shows information of a process model in a compact form.
Open loop identification has established that the identified model shows the same information as the original process model. But when system identification is applied to a Closed loop model, it shows a strange behavior.
For a Closed loop, the identified model does not show the same behavior as the original process model. The input and output chose for identification are not just influenced by model process alone, but also with some external disturbance created by the controller. This disturbance is created by the controller in the Closed loop system.