The Five Myths of Computational Fluid Dynamics

Putting forward some views and opinions that you may or not agree with, Ivo Weinhold from Mentor Graphics suggest that recent advances can explode the myths of CFD

By Ivo Weinhold Mentor Graphics, Mechanical Analysis Division Computational fluid dynamics (CFD) simulation has long had the reputation of being too difficult, slow and expensive to be performed as part of the mainstream design process. A decade ago, these ideas were grounded in reality. But CFD has changed a lot over the past decade and today has become much easier, faster and less expensive to use. However, the previous disadvantages of CFD persist as myths that are standing in the way of greater use of CFD in the early phases of mechanical design. These myths help explain why only about 30,000 out of over 1 million mechanical design engineers worldwide use CFD to simulate the flow of fluids inside and around their products. This article will explain the impact of the latest advancements in CFD software that are helping companies use CFD to improve product performance, reduce time to market, and lower engineering costs.
Myth #1: CFD is too difficult to be used in the design process In a recent survey of mechanical design engineers, 58.1% said that most engineers dont have the necessary expertise and knowledge to use a CFD code. This is most likely because all CFD codes of a decade ago and many of today require the user to have a deep understanding of the computational aspects of fluid dynamics in order to be certain of obtaining accurate results. For example, users need to know how to translate their computer aided design model into the CFD environment, then reverse the model so that empty flow space rather than the solid product is modeled, create a mesh with the right properties, determine boundary conditions, select the right physical models, tweak solver settings to ensure convergence, as well as other tasks. Previous generations of CFD software also required a substantial amount of tuning and tweaking such as manually modifying cells to improve the mesh quality and adjusting solver controls such as the relaxation factors in an effort to get it to converge to a solution. These tasks provided another reason why CFD required the skills of a specialist and could not be performed as part of the design process. But in the last few years a new generation of CFD software has been introduced that addresses all the major reasons for the relative lack of use of CFD software. Its use of native 3D CAD data, automatic gridding of the flow space, and managing the flow parameters as object-based features eliminates the need for engineers to understand the computational part of CFD and instead enables them to focus on the fluid dynamics of the product which is already their responsibility to understand and master. The newest generation of CFD software contains sophisticated automatic control functions that ensure convergence in almost every application without the need for manual tuning. Perhaps the most important function controls the quality of the mesh to avoid one of the biggest reasons for run divergence. Virtually the only situation where the software will not converge is in a problem with a naturally unsteady flow. This 1

type of problem can usually be solved simply by switching to transient analysis. The skills required to operate the CFD software are simply knowledge of the CAD system and the physics of the product, both of which the vast majority of design engineers already possess. The engineer is thus able to focus his or her time and attention on optimizing the performance of the product as opposed to operating the software.

Every product which interacts with a fluid or gas and every production process involving fluids or gases is a potential subject of analysis.
Myth #2: CFD takes too long to use during the design process CFD has long been primarily used as a validation step after the design process has been essentially completed. The major reason is that in the past CFD has taken so long that if a design iteration were to be analyzed during the design process it would have been superseded by other iterations before the analysis results were available. The greatest time sink is the meshing process. This used to require a time-consuming manual step of translating the geometry from the CAD software to the CFD package. From this point, another tedious step was required to extract the cavities from the CAD model and then mesh them. While automatic meshers have long been available, a considerable amount of manual intervention was still needed in order to maintain the quality of the mesh by eliminating gaps and overlaps, and maintaining the required skewness, aspect ratio, warpage and volume of individual cells. This manual process had to be repeated for every design change. The new generation of CFD software greatly reduces the time required for analysis by automating all of these steps. In the new approach, native 3D CAD data is used directly for fluid flow simulations without the need for 2

translations or copies. All ancillary data required for flow simulation, such as material properties and boundary conditions is associatively linked to the CAD model and carried along with all design changes. The software analyzes the CAD model and automatically identifies fluid and solid regions without user interaction. Then the automatic mesher creates the mesh while maintaining the skewness, aspect ratio, warpage and volume to high levels of quality, completely eliminating the need for any manual intervention. The result is that new parts and design changes can be meshed in a matter of minutes, dramatically reducing the time required for analysis. Myth #3: CFD is too expensive to be used by mechanical designers Another factor limiting the use of CFD in the past has been the high cost of software. Traditional CFD codes cost in the region of $25,000 to lease for one year. An even greater cost for most companies is the need to hire analytical experts to run the codes. Analytical experts were required in the past for all CFD codes and are still required for most codes because the software requires the user to have a deep understanding of the mathematical aspects of computational fluid dynamics. These analysts need to spend a considerable amount of time training in order to keep up with the latest changes in the code. In the recent past CFD codes also required specialized hardware such as a supercomputer in order to provide results within a reasonable period of time. The latest generation of CFD code intended for use during the mainstream design process substantially reduces the cost of performing CFD. The cost of the software has fallen to something in the order of $25,000 for a perpetual license, with the only ongoing cost being a maintenance fee on the order of 18% per year. This software eliminates the need for analytical specialists because it can be used by mechanical design engineers with minimal training. The software runs inside the same CAD software environment that design engineers are already familiar with,

and there is no need for the user to understand the mathematical aspects of CFD. Finally, the latest generation of CFD software runs on personal computers and even laptops which have the power possessed by a supercomputer only a decade ago but cost just a few thousand dollars. Myth #4: You cant directly use your CAD model to do CFD analysis In the past it was necessary to copy or translate the CAD model to a different program and then modify it substantially in order to create the CFD model. The main reason was that the tools available to translate the CAD model to the CFD program required a considerable amount of manual intervention. Typically the translation process might work for 80% of the geometry but the rest had to be re-created or simplified by hand. Many users found it more reliable to start from scratch by creating the geometry in the CFD program, although this also involved a considerable expenditure of time. The new generation of CFD software uses the native 3D CAD data directly for flow simulations without the need for translations or copies. All design changes based on simulation results are carried out directly in the CAD system using familiar solid modeling functions. There is no need to create phantom objects in the feature tree to represent the fluid spaces. Flow conditions are defined directly on the CAD model and organized similarly to other design data in the feature tree. As a result, the original CAD model can be used without modification for CFD analysis.

The newest generation of CFD software contains sophisticated automatic control functions that ensure convergence in almost every application without the need for manual tuning.
Myth #5: Most products dont need CFD analysis In the past, the use of CFD has been concentrated in a relatively small number of industries such as automotive, aerospace and power generation producing high-value products in which fluid flow has an obvious and major impact on performance. This has led to the belief among many that CFD only adds value to these high-value products. Actually, fluid flow has a major impact on the performance of a very wide range of products - performance which can be significantly improved by analysis. Every product which interacts with a fluid or gas and every production process involving fluids or gases is a potential subject of analysis. This helps explain why today CFD is being used to improve products as diverse as swimming pools, toilets, lawn sprinklers, gas meters, production printing systems, disk drives, and oil filters. Contact Ivo Weinhold, Mentor Graphics [email protected]

Conclusion Many engineers still think of CFD software as being difficult, time-consuming and expensive to use when actually a new generation of CFD software has addressed all these concerns. The new softwares ability to integrate CFD into the design process allows design engineers to focus on the physics of the problem, which in most cases they already understand well, and generate results much more quickly. This makes it possible for design engineers to get reliable answers gain insight, and deliver tangible results that lead to improvements in product performance and reductions in time to market.
3