Automated model complexity monitoring and adjustment using bond graphs

Haq, Kazi Tayubul (2008) Automated model complexity monitoring and adjustment using bond graphs. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (5MB)

Abstract

Using models of appropriate complexity is important for effective simulation-based design. Throughout a simulated event, systems can have varying inputs, or may have varying system parameters. A single model may not have the most appropriate level of complexity throughout all phases of the maneuver. Therefore using a variable-complexity model could predict the system response accurately while achieving computational savings, can be achieved. -- This thesis presents an approach for switching system model elements “on” and “off” as their importance changes, using bond graphs. Three element importance calculating methods are used to determine an element's contribution to overall system dynamics. Once the power falls below a user-defined threshold, a modified transformer element sets the output from the element to the rest of the system equal to zero. Importance of an element can still be computed as soon the element is “off” by passing the input to the element through the transformer. Again, the element can be switched back “on” if necessary. -- Three case studies are done using a half car, a quarter car and a vehicle frame model. The switching is performed according to the element importance metrics. The computational overhead of power calculations offsets any increase in processing speed due to model reduction even though the appropriate model complexity is used at all stages. Still the method is useful to determine the required model complexity at any instant without prior knowledge of input or parameter changes, and it is able to show how a sequence of systematically reduced models would perform.

Actions (login required)

View Item