Thinking Out loud
Can CFD Be a Valuable Tool for Your Team
Question: If you have a system or a product with fluid flow or heat transfer involved, can you really afford not to incorporate CFD into your design process?
The right tool
I am sure there are many important tenets that are hallmark of successful companies, but I believe higher quality, lower cost, shorter development time, and meeting or exceeding customers’ expectations should be on top of the list for companies that develop products. Clearly, these principles are directly and significantly impacted by design decisions made by engineers who need efficient and robust tools to make the right decisions. With increasing product complexities, engineers need more sophisticated approaches for making the right call. Like other engineering simulation tools, computational fluid dynamics has become an invaluable tool in helping organizations to design and optimize products or troubleshoot systems in which fluids flow. Using numerical methods and algorithms, CFD provides the ability to visualize and understand the complicated flow phenomena for systems too challenging and expensive to prototype. After proper CFD simulation setup, engineers can explore large trade studies and numerous what-if scenarios. CFD lets engineers simulate extreme operating conditions, allows for hard to impossible measurements, and has the capacity for non-testable scenarios (e.g., geophysical, biological, nuclear, etc.) as well as applications that are too huge or too remote.
The virtual environment will make it easier to pursue an integrated multi-disciplinary design process and allows the engineers of various disciplines to keep pace with ongoing design changes. For example, integrating CFD in the CAD environment involves the flow simulation in the design process and will keep the engineer in a familiar environment. It can help reduce activities that require specialist knowledge and provides a more favorable cost justification for the use of CFD.
Working in the virtual environment lets the engineers to iteratively explore options for improved performance. It is much less daunting and costly using simulation to evaluate large number of design iterations to arrive at the optimum design. CFD provides an alternative to trial and error by affording ways to try out different approaches without the cost and delays associated with building the physical prototype at each step. Moreover, using CFD in the early design stages will help shape the direction of the design and promotes design changes well before all details have been set. And when problems occur in a design, engineers can use CFD results to understand the problem areas better and work faster to arrive at the solution.
Less Physical testing
As a simulation tool, CFD is already playing an essential role in the product development where physical prototyping can be time consuming and expensive. There is a number of design areas that CFD has made its mark by reducing both time and cost. With virtually no scale limitation, engineers and designers can explore large parametric studies, and majority of the design iterations can be virtually verified, which should decrease the recurrent physical prototype testing. Use of CFD will not eliminate physical testing, nor it should. However, it will change the nature of testing by establishing better product capabilities with improved CFD expertise. Therefore, physical testing time does not need to be wasted on simple verifications and validations but used judiciously in the detailed design stages.
More Comprehensive Information
generating large trade studies and exploring what if scenarios are major benefits incorporating simulation into the design process. However, what is essential is that valuable performance and manufacturability insights can be generated throughout the design progression with the simulation approach. Since CFD provides high resolution in space and time, designers can get detailed & comprehensive information from the simulations as opposed to getting global qualitative data and limited quantitative information from physical testing.
With recent advances in graphics capabilities, CFD programs are capable of delivering sophisticated fluid flow patterns and thermal gradients visuals to make it easier to see problem areas and overcome design bottlenecks. Furthermore, graphical representation and videos can undoubtedly enhance communications between team members and managers.
Significant saving in both cost and time should be expected by using CFD in the design of products that are impacted by fluid flow, heat and mass transfer, turbulence, phase changes, chemical reactions, turbulence, flow-induced noise, and other flow related phenomena. Employing CFD simulations and virtual testing in product design make it easier to pursue an integrated multi-disciplinary design process, help engineers overcome design bottlenecks and make the right call on design changes, and significantly reduce physical prototype testing.
I’ll discuss the role of CFD and tangible gains by using CFD in several design examples in an upcoming post.
Top Myths about CFD
I read an online article a few years ago about use of CFD among mechanical design engineers. It stated that only a small fraction of these engineers worldwide incorporate CFD in their design work. Why do so few design engineers use CFD? More specifically, why don’t all design engineers with fluid flowing inside or outside of their products include CFD in their design processes? There must be some real reasons and perhaps some misconceptions that are holding these engineers back. One primary reason may be that the universities don’t expose students to CFD early or enough. In many universities that I have looked into, FEA is generally taught in the junior year of the curriculum. In comparison, CFD is often placed in the senior year and perhaps as an elective course. Therefore, many graduating engineers have much smaller exposure to benefits and capabilities of CFD compared to FEA, if any at all.
Lack of exposure may partially explain why many designers don’t venture into the unknown world of CFD. However, persistent misconceptions and myths may be more of a contributing factor. There have been some past discussions about these myths, and some of them seem more significant than others, which I like to explore here.
Myth 1: CFD is too expensive
Businesses do not invest unless they can forecast a substantial return on their investment and simulation software may be a significant investment to some. To a degree, this may be a legitimate concern to any business that doesn’t have the wherewithal to invest much-needed resources in CFD. Of course, when I talk about investment here, I mean both funds for license fees as well as human capital for employees’ time to climb the software learning curve.
Recent developments in CFD market may have gone a long way to ease this investment concerns. Many software vendors now bundle their CFD software with other simulation software to create a multi-physics simulation environment, which does reduce the total cost of simulation while increasing the overall simulation capabilities. There are also vendors that have incorporated CFD simulation into their CAD environment. This approach also results in a reduction of the total software cost while allowing the designer to stay in a familiar environment (CAD), which should result in lowering the learning curve. Additionally, some CFD vendors are now offering free academic licenses to students. The free software should also expose future engineer to CFD earlier, allowing them to become competent users after graduation.
Myth 2: CFD is not practical
The misconception that CFD is not practical may be rooted in the state of CFD in a few decades ago. Back then, an organization needed CFD specialists with extensive knowledge of numerics and fluid dynamics if it wanted to employ CFD. These specialists were needed to develop, maintain, and apply their own CFD codes, which resulted in software with specific capabilities. Furthermore, CFD simulations were only possible on “supercomputers.”
The development and availability of commercial and open-source general-purpose CFD software have drastically altered CFD simulation landscape and should chip away at the myth that CFD is not practical. General-purpose CFD software can respond to a wide range of applications and flow regimes. Additionally, many commercial general-purpose CFD packages include additional capabilities such as enhanced Graphic User Interface (GUI), CAD embedding or geometry generation, automatic grid generation, and advanced post-processing features. The general-purpose codes have become attractive because organizations can choose to use a tool that is developed and maintained externally resulting in the elimination of internal expertise for code upkeep. Also, organizations can easily ramp up their CFD capabilities by hiring analysts that have already been trained on many of the general purpose CFD software.
The advances in computing hardware have also positively impacted the practicality of CFD. The increase in processor speed and the number of processors in a computer, greater availability of memory and disk storage, and advances in graphics capabilities have made “supercomputing” available on off-the-shelf desktops. Furthermore, cloud computing is providing more CFD access to individuals and organizations with limited computing resources.
Myth 3: CFD is not a design tool
The misconceptions that CFD is expensive and not practical are contributing to the myth that CFD is not truly a design tool. Another contributor to this myth is that CFD simulation used to take too much time. Traditionally, meshing has been the most time consuming and labor-intensive part of the CFD process. Furthermore, the need for a proper mesh, optimized based on the flow physics, may require a new and optimum mesh for each design change during a design process. Therefore, remeshing can make a design process more manual and less automated.
Recent advances in CFD such as automatic grid generation, grid adaption, or mesh morphing have alleviated some of the meshing challenges and can help to create a more automated mesh generation process. These technologies are especially helpful for CFD software that allow parametrization or when CFD is embedded in CAD environment. Automatic remeshing can significantly enhance the usefulness of geometric parametrization.
Incorporating CFD in the design process significantly reduces cost and time of prototype development by helping to promote early-stage design changes as well as avoid recurrent physical prototype testing. Moreover, CFD simulation can help to overcome design bottlenecks, eliminate trial-and-errors, enhance communications, and provide an environment for creativity. Lastly, CFD allows for hard to impossible measurements and provide insight on extreme operational conditions.
There are some misconceptions and myths holding back potential CFD users. However, recent advances in CFD software has started to chip away at the barriers and are making CFD a viable design tool.
What Every Design Engineer Needs to Know about CFD
The field of CFD is growing
The field of CFD is growing by leaps and bound. A quick look at the CFD software marketplace and one notices a significant number of choices and I could imagine the number would only climb. In 2008, I created a CFD group on LinkedIn by inviting 5 people to join me. This group currently boasts more than 44,000 members. These members come from different corners of the world with eclectic CFD background and expertise. Not too long ago, a typical CFD software was an ‘in-house' code usually part of a package made of several disjointed programs with scant documentations and hard to find training support. Moreover, the users of such software were required to possess extensive knowledge of mathematics and computational methodologies to develop, modify, and apply these codes for their own use. The present CFD software landscape is a vastly different scene due to the availability of commercial and open source packages.
Organizations are now taking advantage of this new norm and embedding CFD along with other CAE tools into their design and R&D processes. To capitalize on their investment in CAE tools, they are asking their engineers to be agile, versatile, and multidisciplinary in use of these packages. In most organizations with fluid dynamics analysis needs, designer engineers either work closely with CFD analysts or are asked to perform CFD analyses themselves. I regularly meet these design engineers and structural analysts in my CFD training classes.
So, what is CFD?
In simplest terms, CFD is the computer-based simulation of flow motion problems. More specifically, CFD provides approximate solutions to fluid flow problems using computers and numerical algorithms. CFD and physical testing are complementary, but CFD offers significant advantages to a design and analysis process. CFD provides detailed and comprehensive information about the flow field and helps visualize it without any intrusion. In a design process, CFD is a cost-saving alternative to trial and error practices and allows the designer to explore "what if" scenarios.
There are three major stages in a CFD simulation process: Pre-processing, Execution, and Post-processing. The Pre-processing stage of CFD simulation requires that a flow domain, which includes the geometry, to be established. The model must be simplified by excluding any geometrical features that have no significant effects on the flow field. However, it should be noted that major geometrical alterations would adversely impact the accuracy of the simulation. The discretization of the flow domain (mesh generation) also takes place in the Pre-processing stage. Traditionally, the mesh generation has been the most labor-intensive and time-consuming part of CFD analysis as analyst must strive for the optimal mesh. In a CFD simulation, the accuracy of the solution strongly depends on the number of grid points. Basically, more grid points are needed in regions of high gradient. On the other hand, the larger the grid size, the greater the computational cost (time and memory).
The Execution stage is comprised of solver setup and the number crunching. In solver setup, the analyst defines the solver setting by selecting the appropriate the physical and numerical model including material properties, domain properties, boundary conditions, initial conditions, numerical schemes, and convergence criteria. The final step in the CFD process is the Post-processing stage. The post-processing of the results allows for visualization of the flow field, extraction of the desired flow properties, as well as verification and validation of the simulation model. Documentation is also an important part of the post-processing stage.
CFD Best practices
CFD employs an iterative process to achieve a converged solution, but a converged solution does not always equate to an accurate solution. This is because errors and uncertainties are unavoidable in an CFD simulation. For example, errors can arise due to stopping the run before the converged solution has been achieved (Convergence Error) or when a less than adequate mesh is used (Discretization Error). Similarly, uncertainties are possible due to lack of definite choices during the solver setup. For instance, there is no universally accepted turbulence model that works for all flows and all regimes. Moreover, the accuracy and effectiveness of models vary depending on the specific application.
Fortunately, there are some best practices, in form of guidelines and strategies, that could help analysts produce more accurate CFD simulations. For example, a designer can use one of these best practices for building an appropriate CFD simulation model. The idea of this best practice is that simulation in the early design stages pays big dividends when not all details have been worked out. One cannot leave out important geometrical details, but an early simulation does not require exhaustive details. Furthermore, an early simulation will yield reasonable approaches in dealing with thermal-fluid issues. For complex problems, it is important that designers build a concept model first and solve the simple things first and then go on to more complex ones. More specifically, they should find actionable bite-sized places to apply CFD and build on the success of these simulations.
With advances in CFD in the past 10-20 years and the uptake in its use, CFD is primed to become a bona fide design tool. I hope to discuss the role of CFD in design and its application in various stages of design in later posts.
CFD is an essential simulation tool but fraught with many pitfalls for new users. The best practices can help these newcomers to manage the errors and uncertainties and produce credible analyses.
The old saying, “change is the only constant in life”, certainly rings true in the CFD software landscape. We continue to witness sustained change in the CFD marketplace as more, and perhaps better, software choices appear on the market for CFD users. Among these choices are multiphysics simulation packages that include other types of analysis tools (e.g., electromagnetic, structures, etc.) in addition to CFD. Other examples are integrated CFD solutions, which package CFD solvers with other complementary tools such as CAD modeling, grid generation, visualization, design optimization, etc. These examples are receiving deserved lot of attention in forums and publications. Meanwhile, there is another trend, which is getting less attention, that will have a significant impact on CFD use: the development and deployment of market-specific CFD software. Market-specific CFD simulation tools are software packages which are developed or modified for a specific market such as HVAC, data centers, turbomachinary, automotive powertrain, etc. These packages generally include tailor-made capabilities (e.g., Graphical User Interface, boundary conditions, numerical schemes, or object libraries) for use for a specific market.
A look at the past may help us to understand the emergence of these market-specific CFD tools, and how they may affect the CFD market. A few decades ago, the CFD landscape was drastically different, with most organizations developing and maintaining their own specific CFD codes. Such an ‘in-house’ CFD code was generally part of a package made of up several disjointed programs, gathered from different sources without any comprehensive documentation or training. Each of these codes had certain “special" capabilities and was developed for a particular application or flow regime. Due to this lack of generality, the users of these packages had to possess extensive knowledge of numerics as well as fluid dynamics to further develop or modify the code for their own use.
The development and availability of commercial and open-source general-purpose CFD software drastically altered the CFD simulation landscape. General-purpose CFD software can respond to a wide range of applications and flow regimes, each one with distinct requirements. Moreover, many commercial general-purpose CFD packages include additional capabilities such as enhanced Graphical User Interface (GUI), CAD embedding or geometry generation, automatic grid generation, sophisticated post-processing, etc. These general-purpose codes have become attractive because organizations can choose to use a tool that is developed and maintained externally, resulting in the elimination of a requirement for internal expertise for code upkeep. Also, organizations can amass a pool of CFD analysts trained on a single CFD software package, and use one CFD tool for most, if not all, of their applications. Hence, choosing a general-purpose CFD software can result in cost reductions whilst increasing flexibility and availability. The availability of general-purpose software has had a positive impact on the range of CFD applications available, as well as the number of new CFD users.
Obviously, many organizations continue to use specific-application software, and there are many reasons for this decision. One reason may be the long term investment that the organization has already made in the development of software, and the availability of existing expertise internally. Another reason may be the unavailability of the commercial software source code for scrutiny or further development. The cost of commercial software and the steep learning curve for open- source software may be other reasons why some organizations would choose to hang on to their in-house codes.
In the past decade, many vendors have begun to offer versions of their general-purpose software tailored for specific markets. The idea behind market-specific packages is that they provide a simultaneous increase in capability and decrease in complexity. For a field with enough demand and critical mass of users, a special edition of the software that responds to the CFD needs of that field may be more practical than the general purpose version. A specific market may require additional capabilities such as special boundary conditions, solution schemes, CAD capabilities, and object libraries for which there is no demand in other markets. Conversely, there may be options within the general purpose version that complicate the software but may never be used in that specific market. For example, a user simulating the flow in a data center may never use special capabilities such as large eddy simulation or multiphase flow modeling.
Specific-market software packages are not simply a more sophisticated version of earlier specific-application codes, and there are valid reasons for their emergence and proliferation. It may be argued that codes developed in the earlier days of CFD were often developed for specific flow regimes or for specific applications due to a lack of know how or where the science of CFD stood at the time. In contrast, specific-market software has a general-purpose solver at its core, and is capable of handling a wide spectrum of applications. However, certain capabilities are deliberately included or excluded to make the software easier to use, and easier for the vendor to market to these users. For example, it is possible to simulate airflow in an HVAC application using a general purpose code, but the task may be time consuming and labor intensive since the user has to create CAD objects from scratch or develop certain boundary conditions using user defined functions. On the other hand, an HVAC-specific CFD software package may provide only one turbulence model which works best for these types of applications.
Consequences of Market Specificity
The upward trend in the number of specific-market packages, and the increase in number of markets targeted for specific packages, can have a significant impact on the way CFD is used, and the type of users employing these packages. The availability of general-purpose software has resulted in additional CFD users. Furthermore, cloud computing is providing more CFD access to individuals and organizations with limited computing resources. On the other hand, the enhancements and simplification provided by market-specific software can translate to an increase in the variety of the new users. For instance, there may be new users who would become interested in CFD because there is CFD software that caters specifically to their needs, is focused on their type of problems, and is easy to use. Then again, simple to use software would attract users with a limited knowledge of numerics, fluid dynamics, or even engineering. It is important to take stock of the possible consequences of introducing these new users to CFD.
The use of CFD without a proper knowledge of issues and challenges inherent in CFD simulation can adversely impact its use in design and analysis. Lack of understanding of CFD issues such as orders of accuracy,error sources, model simplification, and meshing requirements can result in reworks and delays in product development. Lack or limited knowledge of fluid dynamics fundamentals can be equally as harmful, as wrong or improper results may go undetected.
Making the Impact Positive
Software vendors cannot be faulted for making their software easier to use, or for reaching for more and new users. However, they can help mitigate the adverse effects of substandard simulation by providing practical fluid dynamics, CFD, and heat transfer training in addition to their software training. Practical and fundamental training in addition to software-specific training will help vendors reap the benefits of their investment, and keep these new markets and users. Unfortunately, not all software vendors have the resources to get involved in this type of training. In contrast, colleges and universities do see a role in providing this fundamental education, and offer basic fluid dynamics and CFD courses. However, a great deal of new users may already be in the workforce and would not perceive lengthy college courses as a practical option.
CFD related engineering societies and technical associations are ideally suited to fill the training gap between software vendors and institutions of higher education. These organizations must emphasize and provide fundamental yet practical training for these new users. This training should also be general enough to encompass various forms of CFD (e.g., Navier-Stokes, Lattice Boltzmann, finite volume, finite elements, meshless, etc.). It is also important that this training includes “best practices” in CFD simulation. Engineering societies and technical associations, with help from volunteering CFD professionals, must continue to develop best practice, and publicize this through their various channels.
The CFD marketplace is certainly evolving. With an increase in the number of vendors and software, in addition to the advent of cloud computing, one can expect further and even faster growth in the CFD market. However, poor quality simulation by users with limited knowledge of fluid dynamics and CFD would eventually halt this progress. Engineering societies and technical associations can play a vital role in training new CFD users, andensuring that CFD simulation is used in the correct way, and for the right reasons, to continue the growth of CFD.
CFD education & Practice