MPC of MIMO Inverse Response System

Abstract

A novel QTP multivariable laboratory process that consists of four interconnected water tanks is presented as experimental setup. All the conventional and advanced controllers are implemented on QTP system. Present work is divided in two phases, first one is to develop MPC algorithm to control SISO as well as\ MIMO process. Model predictive control techniques are widely used in the process industry. They are considered methods that give good performance and are able to operate during long periods without almost any intervention. To be very specific MPC has three principle advantages over traditional controller one is it may inherently handles multivariable process very effectively, it may provide good performance even process has subjected to inverse response and process having dead time also could be nicely handle by MPC control algorithm. Reason behind the success of MPC control algorithm is, it has considered specific\ process objective function to be minimize and according to that it would give optimal control moves to manipulated input as a result controlled variable does not fluctuate much and it may settle down within less settling time with minimum overshoot. In present work code for MPC is developed for SISO and MIMO system and implemented on QTP setup successfully. In second phase Fault Tolerant Control algorithm is presented with artificial intelligence fault correction methods. FTC is a robust type of controller in which irrespective of faults in actuators or sensors or it could be in process itself in terms of leak in a\ tank or some change in\ parameters taken as a fault may be taken care by controller to keep process value to its set point very effectively. Two methods are proposed for FTC correction one is using artificial neural network and other is fuzzy based correction. Both correction algorithms, FTC gives satisfactory performance for small amplitude of fault but it is proven from simulation studies in case of large amplitude of fault fuzzy based AIFTC gives batter performance than ANNFTC. At the end validation of ANNFTC is done having comparison with simple PI type of controller.