PhD Defense by Sazuan Nazrah Mohd. Azam: Model Predictive Control Implementation for Modified Quadruple Tank System

Friday 11 January 2019, 13.00 – 16.00, The Technical University of Denmark, Building 101, room S10

Supervisor: Associate Professor John Bagterp Jørgensen, DTU Compute

Associate Professor Jakob Kjøbsted Huusom, DTU Chemistry
Professor Matti Kalervo Vilkko, Tampere University of Technology, Finland
Managing Partner Guruprasath Muralidharan, Smarta-Opti Solutions, India

Moderator: Assistant Professor Dimitri Boiroux, DTU Compute

Model based controller is one of the advanced control strategy that is currently common and extensively recognized in industry and academic, famously known as Model Predictive Control (MPC). MPC is a controller that utilizes the identified model of a system to predict its future behaviour, given a prediction horizon. The main idea is to minimize the cost function and taking into account the constraints. Then the first controller moves is implemented at a sampling instants over the control horizon, by implementing only the first move the optimal feedback is achieved and then the complete sequence will be repeated again, which is known as moving horizon concept. Nowadays the applications of MPC are not limited to the process control field, but also including other various fields. This work described comprehensively an outline for MPC implementation for a linear system on a lab scale system in a simple and constructive method.
Throughout this work, the Modified Quadruple Tank System is utilized as an example to assimilate the fundamental theory of Model Predictive Controller to an exemplification of a multi-input-multi-ouput system, an illustration of the real-world complex system applications which is widely used for education in demonstrating advanced control strategies. It is a simple process that is non-linear but demonstrates complicated interactions between the manipulated and controlled variables. The system consists of four identical tanks and two pumping systems. Flows through the pumps can be controlled in order to achieve desired setpoints of water levels in these tank s in occurrence of some unknown measurement noise and stochastic disturbance. The thesis shows on the modeling part of the system, realization of a linear discrete-time state space model, state estimation using Kalman filter and finally demonstrates the application of MPC in a methodical mannered.

READ MORE about this thesis in DTU Orbit.



Fri 11 Jan 19
13:00 - 16:00


DTU Compute


DTU, Building 101, room S10