Authors: Akio Takamura, Chief Engineer, Honda R&D and Benoit Mallol, Head of Marine Products & Applications Group, NUMECA International
It is common knowledge that 80% of the throughput time of a CFD simulation is spent on preparation and set-up. Typical challenges of this preparation process are poor quality of the available CAD and high complexity
of the geometries. Dirty geometries containing holes and cavities and complex structures normally require a lot of manual input to result in a good quality mesh. It is a laborious process, making it difficult to create various design alternatives for iterative testing.
Many techniques exist to tackle these issues: from capping the holes and removing geometry features to wrapping the geometry, etc. These solutions are offered by many CFD providers, including NUMECA, and they all have their advantages and drawbacks. But none of them really respond to the full specification list that certain users are looking to fulfill: a solution that works for both CAD and STL files, that does not lose any detail of the geometry in the process and that does it as fast and as automated as possible. Honda had that exact same wishlist and found the answer in the NUMECA AutoSeal technology. One of the typical reasons that CAD cleaning in the native CAD system can take days or even weeks depending on the complexity of the geometry, is the mere fact that the objectives of the CFD engineer differ from those of the CAD designer. Think for example about details like welds, seals, nuts, bolts, etc... All these have no added value per se for a simulation, but they are of key importance for manufacturing purposes, so the CFD engineer needs to deal with them. This is where AutoSeal combined with OMNIS™/ Hexpress can make a big difference, and Honda was one of its early adopters.
Honda was already a long-term user of OMNIS™/ Hexpress, and the AutoSeal technology was a logical next step for making their design process more efficient. They chose to apply it for the first time for the design of cabin interiors, and they were very positive about the results. Mr. Akio Takamura, Chief Engineer at Honda Automotive, reported that where a skilled engineer typically needed one full week to close all the holes of the cabin space before, now this whole process was brought down to just about one hour with AutoSeal. And the robustness of the technology was definitely proven: They had a 100% success rate on 10 different vehicle models.
"We have observed speed up improvements for each version of OMNIS™/ Hexpress for the past years, but this one is a major breakthrough"
Mr. Akio Takamura, Chief Engineer at Honda Automotive
Here is how it works: Based on successive octree refinements, AutoSeal is able to detect which cells of a dirty geometry are actually inside or outside of it. This allows the system to progressively identify missing surfaces under a given user-defined threshold. AutoSeal then automatically closes the surfaces of these holes and cavities.
An additional advantage of the process is that it can all be done by one person and within the same environment: Passing to mesh set-up straight after AutoSeal has run, without the need to go back and forth with a design department for this. In order to test whether these meshes return the exact same results as the CAD models would, Honda decided to make a comparison of the aerodynamic performance calculations. They looked at aerodynamic coefficients and passing wind speed of the heat exchanger on over 10 different models and found little to no difference between the models, stated Mr. Takamura.
To speed up the process further once the CAD is prepared, Honda used OMNIS™/ Hexpress’ unique capability to mesh in parallel on distributed memory. This technology allows for meshes to be generated in parallel on several computers or cluster modes at the same time. As a result tens of millions of cells can be created in a matter of minutes.
For example a mesh of 360 million cells that they made previously for the Honda CR-V model took just under 13 hours to create on 16 cores. With parallel meshing today this same mesh would take only 46 minutes on 1200 cores of a cluster, or the amount of time reflected in the table to the right depending on the amount of cores available.
Together with a number of clients in the automotive industry, NUMECA did the math and estimated that they can save up to 50.000 USD for each CFD application they run.