• 1

Parameter optimization

Most designs can be reduced to a number of essential parameters that are usually not perfectly selected in the beginning. The task of parameter optimization is therefore to find the combination of variables that best fulfills the target requirement. A classical method is usually used as the optimization algorithm. For the implementation in the FEM or CFD simulation it is important to choose the algorithm and above all in which system the parameter has been defined. For example, if the property is a material or element property, the modification can take place directly in the calculation model. However, if the parameter refers directly to the geometry, such as a radius, the step back into the CAD model is usually necessary.

However, this step back into the CAD model can cause considerable difficulties. It is quite possible that the parameter value which the optimization algorithm has selected leads to an invalid, non-regenerative geometry in the CAD system. However, the optimizer cannot distinguish whether this is a parameter combination that cannot be reproduced in reality, that is, it corresponds to an invalid solution, or whether the error was caused by an unfavorable geometry description. The designer or calculation engineer who builds the CAD model must therefore think in advance about how to create a model that is as robust as possible in the design space. But even if the geometry can be easily regenerated with the new parameters, a failure of the automated meshing is another realistic scenario that can disrupt the optimization process.

Despite all care, these situations cannot always be avoided. These cases increase especially with increasing complexity of the geometry. This applies in particular to contour descriptions based on free-form surfaces. In addition, the time required to move from a new parameter set to the finished simulation model increases. Against this background, other mesh-based methods such as morphing have become established, which allow subsequent parameterization of a mesh. In this way, the loop via the CAD system can be omitted. In this case there are other disadvantages to be taken into account. On the one hand, it is only conditionally possible to map design parameters as found in CAD 1:1, and on the other hand, an additional manual effort is required to generate the morphing model.


AdSimuTec GmbH

Warmesberg 24, 9450 Altstätten, Switzerland

+41 79 471 1612


Mo-Fri: 7:30AM–12PM, 1–6PM

We look forward to meeting you

About us

AdSimuTec GmbH is a small, innovative engineering service provider specialized in the field of numerical simulation with FEM & CFD.

You will find us in Eastern Switzerland, more precisely in the Rhine Valley of St. Gallen. We are only a stone's throw away from Vorarlberg (Austria) and Lichtenstein on the one hand and Baden-Württemberg and Bavaria (Germany) on the other.