CFD Gets Faster Without All the Mesh
A 30-year-old approach to computational fluid dynamics has reemerged, making the transition from CAD to CFD analysis faster and requiring fewer compute resources.
“Nobody wants to spend days or weeks creating a mesh and prepping a computer-aided design [CAD] model for computational fluid dynamics [CFD] simulation,” says Alberto Griffa, head of North America sales and operations for Karalit (karalit.com). “Instead, analysts want to skip the time-consuming prep and go quickly and directly from CAD changes to highly accurate CFD analysis. They want to generate more iterations in less time so they can better optimize their designs.”
Karalit’s “direct CFD” approach makes that happen in “no time,” continues Griffa. “Pre-processing as we know it is gone for good.”
Some history about CFD
In the 1990s, the Holy Grail for CFD was finding the perfect grid-generation technique, explains Griffa. “All types of unstructured and adaptive mesh techniques were developed, including tetrahedral, hexahedra, prisms, and hybrid mesh generation. No method was found superior to others, leaving no perfect solution. Unstructured mesh generation became popular because it was fast, but its quality was often poor.” Analysts found that small amounts of cell distortion were inevitable when constructing body-fitted grids in complex geometries. These distortions inevitably introduced some degradation in the algorithms used, leading to distortions and other inaccuracies in the final results. And throughout, generating a body-conforming mesh was a difficult, iterative, time-consuming process. The immersed boundary (IB) method, first developed in the 1970s, gets rid of the complicated mesh and all the work that goes with it (creation, analysis, and iterative re-creation). IB replaces the conventional body-fitted mesh with a projection of the geometry under analysis against a background of regular, rectangular squares—a simple Cartesian grid. (Imagine projecting the geometry on a window screen.) There’s no body-fitted mesh that has to adapt or conform to, say, the shape of a car. Moreover, even though the shape of the car may change because it’s moving or deformed by something else (a solid object, wind, water, etc.), there’s no need to remesh the new shape. Instead, the grid stays the same; only the geometry has to be recomputed as the object moves around. The recomputation is completely transparent to the user.
Flash forward. The IB method was rediscovered a decade ago as computing power erased previous problems in processing structured multi-block meshes. Around that same time, research began proving IB’s ease-of-use, accuracy, and effectiveness for fluid problems ranging from external flows to automotive applications to internal and biological flows.
IB in use
Recently announced Karalit CFD 3D is based on the IB method. To use it, the analyst first imports 3D geometry in standard STL format. (Other formats might become available depending on customer demand.) Then the analyst uses a Karalit “app,” basically a problem-specific template, to set up the simulations. After entering the simulation parameters, Karalit does the rest. “Set-up time is a few minutes, compared to hours or even days of preprocessing with traditional CFD software,” says Griffa. Currently there are Karalit apps for valves, aerospace, automotive, and architectural/engineering/construction simulations.
Once the geometry is immersed in the grid and the problem defined in an app, Karalit CFD 3D proceeds as with conventional CFD. (That is, the software analyzes finite elements à la CFD. Standard physics as described by Navier-Stokes equations applies equally to grids as with meshes.)
Because the IB method alone might not the best solution for turbulent simulations, Karalit has developed the immersed mesh method. IM automatically generates a body-fitted grid in the regions close to the body surface of the object. This is essentially a body-fitted mesh immersed into a Cartesian grid the same way a body surface is immersed into a Cartesian grid as in the IB method. “This combination takes advantage of both the immersed boundary and the body-fitted approaches,” explains Griffa. “The result is much faster migration from CAD to CFD, speedier computation, and increased accuracy.”
The IB approach affords several advantages. There’s no meshing and remeshing the computational domain—all that prepping and post-processing common in conventional CFD programs. In addition to the time savings, no remeshing also eliminates a complicated morphing processes that can deteriorate the existing mesh. Next, basing CFD on Cartesian grids increases the accuracy of the results because the grids are not subject to the cell distortion that occurs in body-fitted meshes. Plus, the grid’s consistency makes finite element calculations easier than accounting for a variety of irregular finite elements (such as polyhedral that need to be distorted to fit a curve). All this leads to IT-lifecycle costs being smaller for IB and IM, compared to conventional CFD. IB-based analysis takes less computational time and uses less computer memory. The software can run on commonplace and evolving CPU technology—no special or exorbitant hardware required. Because Karalit CFD 3D is already a multicore software system, it can be run on as many CPU cores as are available—without additional software licensing costs.
Automotive manufacturers are meeting CAFE fuel-efficiency standards through lightweighting, which requires simulation software for design engineers.
Systems engineering in increasingly being recognized as a valuable approach to vehicle development - both in design and production. Siemens posits that PLM is the right software system for systems engineering.
Although the term “continuous improvement” is generally associated with another company, Honda is certainly pursuing that approach, as is evidenced by the Accord, which is now in its ninth generation.