Advanced Computing in the Age of AI | Thursday, March 28, 2024

CFD For the Rest of Us 

<div><img style="float: left;" src="http://media2.hpcwire.com/dmr/cutplane_zipp_fvstamped.png" alt="" width="97" height="45" />Recent advances in the way users access and work with computational fluid dynamics (CFD) are making this complex technology more available to small and medium sized manufacturers. The result – increased creativity, more innovative product design, and a boost in competitiveness.</div>

Scott Plasma racing bicycle frame with Zipp Firecrest 808 wheels.  DES simulation performed with STAR-CCM+ (CD-adapco), 3 cutplanes showing relative air velocity shown. (FieldView image courtesy of Intelligent Light)

You would think a technology developed almost 50 years ago would have made it into the mainstream by now.  

But computational fluid dynamics (CFD), although a staple at large manufacturers in industries ranging from aerospace and automotive to sports equipment and turbomachinery, has yet to make significant inroads into the “missing middle,” those small- to medium sized manufacturers (SMMs) that have been slow to adopt advanced digital manufacturing technologies.

The problem has to do with money, time and talent – the lack of all three.  But there are forces at work that may help bring CFD capabilities within the grasp of smaller players in the manufacturing field.  

Intelligent Light is one of the companies helping to pave the way. The company provides CFD post-processing and workflow software including FieldView, the most widely used software for CFD post-processing by the engineering and research community.

In recognition of the challenges associated with using CFD and to experience firsthand the hurdles faced by their customers, Intelligent Light embarked on a multi-year CFD workflow study.  The CFD simulations were performed on aerodynamic bicycle racing wheels.  Intelligent Light used the studies to improve its own ability to help its customer become dramatically more productive with CFD.  

In recognition of its leadership, Intelligent Light received the prestigious HPC Innovation Excellence Award from International Data Corporation at SC12.  The award showcases the advantages that the innovative use of HPC can bring to small- to medium-sized businesses (SMBs).

Steve M. Legensky, founder and general manager of Intelligent Light, says that the innovation award is an indication that the technology is becoming more accessible to smaller manufacturers.  For example, the company’s efforts caught the attention of Zipp Speed Weaponry, a SMM that is one of the leaders in manufacturing cutting-edge high tech bicycle racing wheels.

Commenting on the award, Intelligent Light’s Dr. Matthew N. Godo, a CFD practitioner himself, said, “By applying state of the art CFD tools and practices along with commercially available technology, we achieved new insights into the aerodynamic performance of these wheels.”  

Scott Plasma racing bicycle frame with Zipp Firecrest 808 wheels.  DES simulation performed with STAR-CCM+ (CD-adapco), 3 cutplanes showing relative air velocity shown. (FieldView image courtesy of Intelligent Light and Zipp Speed Weaponry)

Godo developed a highly automated CFD workflow and used cloud-based HPC resources from R Systems and Dell Cloud Service to quickly amass enough data to isolate effects that are not directly observable during physical testing.  For example, vortices shedding off the wheel are seen as a single force in wind tunnel testing, but CFD simulations revealed new details that have major stability and control implications.  Also, results were available in days, rather than weeks due to the use of the Intelligent Light post-processing and workflow automation tools.

Comments Roger R. Rintala, head of strategic relations for Intelligent Light, “It is this highly-productive and repeatable workflow that is a key enabler for SMMs who aren’t looking to invest in on-staff CFD experts or outside CFD practitioners who lack their product design and domain knowledge.”

Intelligent Light and Zipp have embarked on a strategic partnership to design and deploy CFD with the design engineers at the company for ongoing use, embedding their knowledge and best practices and providing a basis for their own ongoing process development.  This, says Rintala, recognizes that this confluence of technology and services expertise can help SMMs  affordably build a sustainable competitive advantage using CFD.

Zipp combined the CFD results with its own substantial design capabilities and came up with Firecrest, the fastest bicycle racing wheel every built and a best seller.  To meet the demand, the company has added 120 new manufacturing jobs in Indiana.

CFD for the Rest of US
Anywhere that fluid flows or gases swirl, CFD can bring new insights to the manufacturing process and help manufacturers sharpen their competitive edge.

CFD results are obtained using partial differential equations.  The equations are typically solved by breaking up the physical domain to be investigated into small volumes called cells.  The collection of cells that make up the domain is called the mesh.  Software, known as the solver code, reads the mesh and setup equations and produces results files that contain the predicted flow for each cell.  Since flow is dynamic, the software runs repeatedly for many small time steps.

The process encompasses three steps:
•    Pre-processing to build the mesh and set operating parameters
•    Solving by running the solver code
•    Post-processing the results files to obtain actionable information

Post-processing is numerically intensive and can require more wall-time that solver runs which are more scalable.  Many organizations rely heavily on batch post-processing on HPC servers (where the solvers ran) and/or dedicated visualization machines co-located with the HPC servers.  Post-processing allows the engineers to deal with these large, complex results files and gain an understanding of critical flow features that will impact the design of the finished product.  

According to Rintala, “Interactive post-processing and visualization is a luxury reserved for the times when people need to explore the data to increase understanding and identify specific flow characteristics of interest.  From there, key design criteria can be identified that can be reported upon automatically with numeric, plots, images, animations and the like (figures/numbers of merit).  The also enhances the ability of the process to rapidly compare multiple design alternatives and design-points.”

A major problem is the fact that many companies only need to invoke these CFD capabilities to try out new ideas and evaluate design decisions at certain points in the development cycle.  They have neither the need nor the financial resources available to create a full-fledged in-house CFD capability.  And it is here, says Rintala, that the new, cloud-based approach to CFD is making inroads.

Zipp, for example, is using CD-adapco’s STAR-CCM+ /Power-on-Demand which allows users to access on-demand cloud computing services such as the HPC capabilities provided by R Systems.  The Zipp engineers can access and work with the CFD results data in the cloud whenever needed.  R Systems also provides technical support.

“Being able to leave all that data on the server is critical,” says Rintala. He points out that because of the huge file sizes, previously the only way to get the data back and forth to the users was to ship hard drives or spend days transferring files across the Internet.  Now, using Intelligent Light’s FieldView post-processing software, users can automate the repeated tasks, interact directly with the data in the cloud, and extract only what is needed for interactive post-processing and archiving on the local client.  

“The user has the option of working in client/server mode or running a batch job in the cloud and then bring down subsets of the data in the form of FieldView XDB files,” he explains. “This is the kind of workflow we are now seeing at leading edge customers – from Formula 1 to major aerospace, to SMMs.  It isn’t size that determines success, it is defining needs and developing these productive workflows.

Center of pressure analysis on aerodynamic frame and wheel combination. (FieldView image courtesy of Intelligent Light and Zipp Speed Weaponry)

Passport to Creativity
Legensky reinforces and adds to Rintala’s comments: “For smaller businesses like Zipp wheels to effectively use CFD you need three things. The first is readily available inexpensive, on-demand HPC computing.”  He adds that CFD breaks down well for parallel processing operations on HPC systems.  

The second requirement is being handled very well by the solver companies – they are supporting automatic meshing as well as offering licensing schemes that allows SMMs access to the kind of core counts needed to properly conduct a CFD analyses on these large data sets.

Finally, products like Intelligent Light’s FieldView are restructuring workflows to mitigate the post-processing bottleneck.  

“For example, once we set Zipp up with a comprehensive workflow and the computing power in the cloud to support it, meshing and post-processing became automatic, freeing up their engineers to be highly creative,” Legensky says. “Wind tunnels are great when you are down to one or two designs.  But, by being able to manipulate CFD in a virtual world, you get a chance to try more designs and learn things that you would not have uncovered in the wind tunnel.  For example, at Zipp when they saw those vortices coming off the wheel during an unsteady simulation, the Zipp engineers got it right way and were able to guide the design to get the wheel they wanted.”

Concludes Legensky, “What’s exciting is that this new way of working is not about IT and CFD, it’s all about giving these engineers the creative tools that allow them to make better decisions and better products. It’s no longer just an expert’s game.”

EnterpriseAI