Design of Control Schemes to Adapt PI Controller for Conical Tank Process

The non-linearity and constantly changing cross section of the conical tank are the immense challenge in process control and it cannot be effectively controlled by the linear PID controllers. Hence, an attempt is made here to adapt the PI controller suitable to control the conical tank by two methods namely an adaptive PI controller and neural network based PI controller, which are designed by choosing level as the scheduling variable. The first method combines the output of multiple linear PI controllers, each describing the process dynamics behavior at a specific level of operation. The global output is insertions of individual PI controller outputs weighted depending on the current value of the measured process variable. In the second method, a neural network is used to calculate the PI controller parameters based on the scheduling variable that corresponds to major shift in process dynamics. The real time implementation of these two control schemes is done in simulink toolbox using VMAT-01 data acquisition module for the conical tank process. The experimental results show that the proposed control schemes have good setpoint tracking and disturbance rejection capability.