Morphing of a body in white
In our article on mesh morphing, we argue that this is a very effective and powerful method for making geometry adjustments to an existing FEM or CFD model. We are trying to show here, using a simplified body-in-white as an example, that there is something to this statement. Even this greatly reduced model is already based on many complicated formed sheet metal parts, which all have to be changed several times for an extensive design study.
However, before we go into the example in more detail, we will briefly compare the current workflows for a design iteration.
In the CAD-driven workflow, many CAD models and even more design elements in the respective parts history would have to be adapted for a geometry variation of our bodyshell. Since complex geometries and assemblies can rarely be fully parameterized in CAD across all parts, this ultimately means a lot of manual work for the designer or CAD team. However, this is not the only additional effort that can arise.
When transferring the changed geometry into the preprocessor, additional difficulties may occur. This is even the case when using direct interfaces to the CAD system. Often small surface patches disappear or are replaced by new ones. This has nothing to do with the interface itself, but is simply due to the geometry description. For example, if surfaces are added or dropped, the reference to load and/or boundary conditions is quickly lost. In this way, a previously contiguous area for a contact definition can have openings at once. The mere possibility that such things happen demands an additional careful model check. This also requires interaction with the user. A completely autonomous design study is therefore not always feasible for complex structures.
These disadvantages can be avoided with a different workflow. In combined or hybrid workflows, CAD and preprocessor are responsible for the design changes to the same extent, but at different points. The CAD system initially delivers the basic design for the calculation model as usual. After the first simulation and evaluation of the results, however, the desired changes are made to the existing model in the preprocessor. Depending on the method with which the geometry is to be changed, an additional model definition is required once. In the case of box morphing, this would be the generation and parameterization of the control volumes. Once a design has been found that meets the requirements, the geometry is returned to the CAD. As a rule, however, this cannot be used directly, but serves the designer as a "guideline" for a new CAD design, which can finally be checked again in the simulation. In this way, the desired design study can be carried out autonomously, at least from the first model creation up to the return of the geometry to the CAD system.
Application on the example of the body-in-white
In our example, the following changes to the body-in-white are to be examined. The position and inclination of the B-pillar, the width and height of the roof and the position of the connection between the windscreen and the beginning of the roof. Box morphing is a suitable method for processing the task. For this purpose, the areas affected by the changes must first be enclosed with the corresponding control volumes. Due to the symmetry the morphing boxes can be mirrored at the symmetry plane for simplification. In addition, the symmetrical control volumes can be linked together so that all changes on one half of the model automatically affect the opposite side. In a second step the design parameters have to be defined and finally the model has to be checked if the desired geometry changes are sufficiently well mapped. The images below show the results of the changes, once isolated for each parameter and once in any random combination.
Effects of individual design parameters on the model
Effects of any combination of design parameters on the model
In summary, the following can be said. The widely used CAD-driven workflow has two major weaknesses. On the one hand, it requires personnel-intensive and therefore expensive work steps, especially for complex models, and on the other hand it carries a not inconsiderable risk of errors being introduced into the simulation model. If only limited resources are available, both ultimately lead to a poorer design, since fewer variants can be assessed within the given time frame and sometimes the informative value suffers due to poor model quality.
If, however, the desired changes are made in the preprocessor through mesh morphing, these disadvantages can be avoided and the effort for individual design iterations is kept significantly lower. The time-consuming step back into CAD is only necessary at the end after a satisfactory design has been found. This also gives the designer a clear design guideline, thus reducing misunderstandings in communication.
The one-time effort for creating and parameterizing the morphing model is usually small compared to the savings achieved. Nevertheless, this should not hide the fact that this step can also be a little more time-consuming. Also, one should not expect that the geometry changes by morphing have the same precision as in CAD. However, this is also not the claim of the method. The goal is rather to find out in a short time which parameters are dominant, which interactions exist and where an optimum can lie.
In addition, the merging of both processes is a possible and in certain cases an appropriate scenario. From our point of view, flexibility in the workflow is an important aspect on the way to better products.
Many roads lead not only to Rome, but also to a better design.
We are happy to be at your side when it comes to exploring new ways!