Model Predictive Control (MPC) utilizes a model of the system under control to simulate performance over a time horizon into the future, and then optimizes the current control input data to achieve a future goal. This necessitates that the algorithm run much faster than real time. Although MPC has been applied extensively in industrial process control where the processes tend to have relatively long time constants, its application to aerospace systems has been limited because the large computational effort required precludes real-time operation. The time necessary to calculate a solution is a function of several factors including the time horizons selected, the number of state variables, and the data that are input to the system. The NASA Glenn Research Center, in collaboration with the Cleveland State University, has utilized a modified approach to enable a real-time implementation of MPC to aerospace systems while maintaining closed-loop performance close to that obtained by using the original MPC implementation.
The control application considered for this study is a large commercial turbofan engine simulation, capable of running faster than real time. A non-real-time implementation of MPC incorporating this simulation had demonstrated the capability to achieve complex tradeoffs in engine control, such as minimizing turbine temperature to extend part life while still maintaining acceptable transient performance during takeoff, and minimizing fuel consumption during cruise. This original MPC algorithm was not able to run in real time because of the number of variables that had to be optimized at each control interval. Specifically, the simulation has three fast-acting control input variables: fuel flow, variable guide vanes, and variable stator vanes. To simplify the model and thus reduce the computational load, the researchers used a multiplexed actuation approach. This multiplexed MPC (MMPC) technique computes a change in a single actuator at each control interval, holding the others constant. The algorithm rotates through each of the actuators over three time steps, thus significantly reducing the computational effort required.
MMPC performance compared with MPC performance, and comparison of the computation time required per control interval.
Long description of figure.
The closed-loop system performance achieved with the multiplexed approach was similar to the original case where all actuators were updated together, whereas the computation time was reduced to about one-eighth. In addition, no convergence problems were observed with the optimization algorithms. These results indicate that the MMPC approach can effectively enhance the control of complex aerospace systems. For the propulsion system used for this study, the faster-than-real-time execution of MMPC allows users to extend the control time horizon or to reduce the control interval, thus potentially further reducing the small performance gap between the original MPC and the MMPC.
This work was performed under the grant “Computationally Efficient Predictive Control Strategies for Real-Time Implementation” with the Cleveland State University, with Prof. Hanz Richter as the principal investigator. This work was supported by the Propulsion 21 project under the Fundamental Aeronautics program.
Find out more about the research of Glenn’s Controls & Dynamics Branch: http://www.grc.nasa.gov/WWW/cdtb/
U.S. Army Research Laboratory at Glenn contact:
Jonathan S. Litt, 216-433-3748, Jonathan.S.Litt@nasa.gov
Cleveland State University contact:
Dr. Hanz Richter, 216-687-5232, h.richter@csuohio.edu
Author:
Jonathan S. Litt
Headquarters program office:
Fundamental Aeronautics
Programs/projects:
Propulsion 21
Last updated: December 14, 2007
Responsible NASA Official:
Gynelle.C.Steele@nasa.gov
216-433-8258
Point of contact for NASA Glenn's Research & Technology reports:
Cynthia.L.Dreibelbis@nasa.gov
216-433-2912
SGT, Inc.
Web page curator:
Nancy.L.Obryan@nasa.gov
216-433-5793
Wyle Information Systems, LLC
