mor4ansys (version 1.8): Compact Behavioral Models from ANSYS by Means of Model Order Reduction E. B. Rudnyi, J. G. Korvink, IMTEK, Freiburg University, Germany, {rudnyi,korvink}@imtek.uni-freiburg.de http://www.imtek.uni-freiburg/simulation/mor4ansys/ 21.02.2005 What is new in 1.8 All information from the FULL file can be read into mor4ansys. This simplifies the transfer of infor- mation from ANSYS. As result, options -D and -T are removed. Two additional methods to deal with the second order systems: the transformation to the first order system and the SOAR algorithm. The code is rewritten. Now the computational routines make the mor4ansys library. What is new in 1.6 Numerous bugs are fixed. Model with constraint equations (CE command family) can be handled by mor4ansys. The implemen- tation has not been optimized yet but seems to work. Introduction The goal of the manual is to describe the software mor4ansys (pronounced “more for ANSYS”) that generates compact models directly from ANSYS models by means of model order reduction (see [1] for an overview). The advantages of this approach are as follows: • ANSYS models are quite accurate. They can be created for a realistic CAD geometry of the device by taking into account small details that are important to make an accurate model. • The software allows us to reduce the dimension of an ANSYS model significantly. From our experi- ence [2], for thermal and mechanical problems the dimension of a reduced model of less than 30 allows us to have an accurate representation of original ANSYS models up to a dimension of 500 000. • In order to run the software, only minimal knowledge about the background theory is sufficient to pro- ceed. The generation of a compact model is almost automatic. • The computational cost of model reduction is relatively small. Usually, the model reduction time is comparable to the solution of an according stationary problem in ANSYS. The analysis of the compu- tational complexity is in Ref [3]. Let us explain what we mean by almost automatic. This means that there are a few important decisions to be taken by the engineer responsible for the simulation, namely: • The software expects the ANSYS model to be linear. This means that if material properties and depend on temperature, one has to choose an appropriate operation point to linearize the model. Note that the nonlinearity in the input function, however, is allowed. • Currently, it is not possible to choose the best order of the reduced system a priory. A user should choose the order of the reduced system manually. We believe that these two conditions are quite reasonable considering the huge reduction in model dimension that you can achieve by using mor4ansys. Note that we have research results (see publications at the mor4ansys site) that will allow us to