JUSTIN MILES

Modeling of a Dynamic Airflow System 

Overview
          Within its undergraduate research program, USC’s Department of Chemical Engineering is developing a process simulation model for a dynamic airflow process.  When completed, the project will serve as a tutorial for undergraduate students learning process controls theory.  Also, the finished model will be used to test the capabilities of an improved controller currently being developed here at the university. 

About the Project 

Process controls theory is an important aspect of every industrial process.  It is the basis for the development of complex computer programs designed to regulate industrial processes.  These systems are the “brain” of an industrial process.  They carry out the vast majority of all operations within the plant, as well as ensuring that product quality remains high.  Control systems are also designed with inherent safety measures capable of shutting down reactors and stopping flow into tanks to prevent overflow in order to maintain a safe plant environment.  Simulation models of industrial processes can be used to test that a control system will function properly once installed in the plant.  Also, by using the control system to run a simulated model of the plant, one can identify flaws in a control system and correct them before the system is installed in the actual plant.  With the help of Dr. E.P. Gatzke, I will be using various computer programs such as MATLAB and Simulink to create an accurate, multi-variable simulation model for an airflow system which will be used to test a new controller, as well as serving as a hands on tutorial for undergraduate chemical engineering students such as myself. 

The system consists of two large and two small tanks arranged into two trains [see Figure 1].  Each train consists of a large tank connected to both of the smaller tanks.  Air flows into the system through two control valves fitted with regulators (designated a and g on the figure).  Airflow exits the system through the two smaller tanks, which can be opened to the atmosphere by the two small hand valves designated as d and j.  Airflow into each of the four tanks and between the four tanks is determined by the position of six more hand valves, designated b, c, e, f, h, and i.  Our system is set up to measure pressure in each of the four tanks.  I hope to create a mathematical model which, when entered into MATLAB and Simulink, will accurately predict these pressures based on the positions of the valves and the differential pressures between the tanks.

            Once I have completed the process simulation model, it will connected to a modern type of controller known as an Model Predictive Control (MPC).  What makes an MPC unique is that it will take the mathematical model associated with a system and use it to calculate the optimal way to bring a process to a new set point.  However, what really makes this type of controller stand out, is that it continuously recalculates the optimal route to the new set point.  This phase of the research will relate more directly to process and industry.  It will allow us to judge the ability of the new MPC and determine its capabilities and functionality for real industrial processes.   

Dr. E.P. Gatzke’s Homepage

http://opus.che.sc.edu/~gatzke/index.html

Process Controls

http://www.theorem.net/control.html

http://www.manufacturing.net/ctl/

Using MATLAB in Control Theory

http://www.eece.maine.edu/mm/matweb.html#as-h3-1932210

MPC Links

http://www.che.utexas.edu/~qin/cpcv/cpcv14.html

http://www.ee.ethz.ch/research/control/predict.en.html

http://www.dii.unisi.it/~bemporad/mpc.html