Development of a Quality Metric for Judging the Suitability of Controller Designs Based on Bond Graph Technology

Introduction

When comparing the quality of different controller designs for controlling a non-linear control system, the problem arises that different controllers frequently make use of different measurement signals, sometimes they even make use of different actuator signals. Most certainly they are characterized by distinct internal structures. For this reason, it is difficult to compare their suitability in an objective fashion.

In this project, power flows through a control system from the actuator signals to the measurement signals were being analyzed. It is particularly easy to follow these power flows through the system when the system is modeled by means of bond graphs. An efficient controller should supply all parts of the control system with power, at least during an active control phase. If a part of the control system remains inactive, it is either not needed, or the controller has not been optimally designed. Using an analogy: a body part that isn't supplied with blood cannot perform any work.

The project developed methods for quantitatively determining the reachability of a system component by the power flow. Thereby metrics were created that enable the control engineer to judge:

  1. when it is time to stop with the pptimization of the parameters of a controller, and

  2. which among a series of alternative optimized controllers is most suitable for a control task at hand.


Most Important Publications

  1. McBride, R.T. and F.E. Cellier (2005), System Efficiency Measurement Through Bond Graph Modeling, Proc. ICBGM’05, 7th SCS Intl. Conf. on Bond Graph Modeling and Simulation, New Orleans, Louisiana, pp. 221-227.

  2. McBride, R.T. and F.E. Cellier (2005), Optimal Control Gain Selection Using the Power Flow Information of Bond Graph Modeling, Proc. ICBGM’05, 7th SCS Intl. Conf. on Bond Graph Modeling and Simulation, New Orleans, Louisiana, pp. 228-232.

  3. McBride, R.T. (2005), System Analysis Through Bond Graph Modeling, PhD dissertation, Dept. of Electr. & Comp. Engr., University of Arizona, Tucson, AZ.

Sponsors


Deutsche Version
Homepage


Last modified: July 13, 2005 -- © François Cellier