Automatic Differentiation used in Simulation Systems for Industrial Applications
by
Wolfram Klein
SIEMENS AG, Corporate Technology

Simulation has a growing importance in industrial production since many years. In the meantime, simulation is used for planing of products or systems as well as for the operation, the reengineering or the service of such systems. Moreover, corresponding to faster changing product cycles, new and more complicated simulation models have to be developed in a shorter time.

The modeling and realization of the functional behaviour of system components is the essential task concerning the development of simulation models. The interaction between different components as well as the nonlinear behaviour of these components leads in many cases to the need of gradient information. Here, the application of systems for automatic differentiation (AD) to different real world simulation systems will be proposed.

In this talk, results concerning several AD-systems like ADIFOR, ODYSSÉE, ADOL-F/C and FADBAD will be discussed. These systems are applied to transistor models coming from circuit simulation, e.g.. A comparison of numerical and automatic differentiation as well as a comparison of the FADBAD and ADOL-C systems will be shown for the simulation of a cooling system. Nevertheless, the application of AD-systems to a large and complex, object oriented simulation system for power steam stations is still a real challenge for the technique of automatic differentiation. Here, difficulties arising in the application of the AD-systems FADBAD and ADOL-C will be discussed.

(Invited speaker at application day)

Date received: February 28, 2000

Copyright © 2000 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cads-94.