November 1, 2014
By Eric Stamper
It’s now common for users solving finite element (FE) problems to use machines with multiple CPUs. With these more powerful computers, tasks such as creating a mesh and solving can be performed in parallel. Yet, as our FE models continue to grow in size by including more components in the total assembly, the limiting factor in the overall analysis process tends to become how quickly an engineer can set up the simulation.
The same idea of parallelization can also be applied to the modeling effort; however, instead of multiple CPUs, we use multiple engineers. This can be a tricky task to achieve in an efficient manner. But if the following five steps are followed, it can be an effective way to decrease the lead time for obtaining the overall result. This process can even evolve into a standard simulation plan for a company that regularly models large assemblies.
1. Divide up the assembly into modular components.
The goal of having multiple engineers building one FE model is to arrive at the finished assembly model faster. Therefore, each individual needs to work on one component or a subset of components in the assembly and follow the same setup process as everyone else. The end result will be to combine everyone’s model (or database) to create the full assembly.
2. Create component alignment.
Each engineer should build his or her component in the correct “modeling space” so that all the parts, once assembled, will align properly. A master file relating all the positions of all the components could also be used when combining the models, but this can easily be avoided in the upfront planning.
3. Tag attributes.
A tagged attribute is simply a way to identify certain features in the model that can be used in the setup when combining all the models into the full assembly. These attributes will aid in this process when it’s time to assign the properties—material properties, boundary conditions, loads and contact pairs, for example. Depending on the software, it might be easier to assign materials, for example, within each database. In other finite element analysis (FEA) software, it might be easier to use the named attributes within each database to assign the materials once combined into the full assembly. Whichever approach is used, these properties, saved into each individual database, must transfer into the main assembly and will be used to define their respective conditions.
4. Implement assembly contact.
The individual databases will eventually be assembled and connected together in the full model. The use of bonded contact will be of great importance in assembly modeling, as it allows for all the components to be meshed independently by each engineer, without having to know what the mesh on all the other components will look like. Some FEA software, like ANSYS Mechanical, has the built-in capability to automatically recognize mating surfaces and create a bonded contact pair. If this capability is not available when combining the full assembly, this is when the consistent naming of surfaces (tagged attributes) facilitates manually setting up the contact between components. For example, a toy figure is divided into four components (“body,” “left arm,” “right arm” and “legs”) in individual databases. Each component was built in the correct position in the modeling space, and the surfaces where the parts will “stick together” are tagged with the names. In the “left arm” database, some surfaces are tagged with “bond_left_arm_to_body;” in the “body” database, some surfaces are tagged with “bond_body_to_left_arm.” Listing all the contacts that need to be bonded together in a master file will make it easy to automate the contact creation with a short script.
5. Follow guidelines and document, document, document.
While this step is listed last, it needs to be established at the very beginning before each engineer starts working on his or her model—and be continuously updated throughout the entire setup process. Documentation will allow the team to keep track of what’s in each model. It ensures that everyone is building their model in a consistent way, naming all their attributes with a similar convention and notating how all the components will come together in the full assembly. Clear and consistent communication among all the engineers is critically important.
This modular approach to large assembly modeling helps decrease the overall lead time on setting up the analysis. Another added benefit of using this technique is that it makes it easier to swap new components in and out without affecting the rest of the assembly. This leads right into applying this modular simulation approach in a design of experiments to perform quick design changes on the component level and determine the full system’s response to the new design.
Eric Stamper is engineering manager at CAE Associates Inc. He has more than 10 years of experience in engineering simulation and analysis. Contact him via CAEAI.com/fea-services.