Trends and Challenges in CFD

This is a brief thought experiment about the latest trends and challenges in CFD (Computational Fluid Dynamics).

I am sure most of you know very well what is CFD. But still: CFD is a quite young discipline, where the fluid dynamics is being simulated on a computer. CFD helps the engineers with designing and testing products to be for example more efficient, more durable or cheaper to manufacture.CFD is widely used in the industry. In the automotive, turbomachinery, aerospace or civil engineering. And of course, the industry has a major impact on CFD development.

The CFD market is now a fully global business. And as many other engineering fields, also CFD is undergoing a big transformation. Here is my favorite quote from American philosopher Eric Hoffer, who said: “In a time of drastic change, it is the learners who inherit the future.” The CFD users’ needs are changing. And we should learn from them and deal with them right now - to succeed in the future.

At CFD Support, a CFD consultancy company I work for, we receive new and new requests on CFD projects every day. Based on these requirements, we have picked the most significant trends and challenges going on in CFD. These are some other useful references: [1] [2] [3] [4] [5] [6] [7] [8].

Democratization

The first one would be the Democratization. More and more people are getting involved in CFD. In the past, CFD used to be, by its nature, a complicated science for the scientists, for the developers, for the analysts - for the experts. But in the last few decades, CFD has become a tool. A tool, which is being used by a wide range of people, the engineers, designers – the non-experts. Who use CFD as a tool. The users want to deal with their real value added work. Their goal is designing their product - not to deal with a boundary layer or turbulence models. For this reason, there is a big challenge for CFD codes to be ease of use, robust and accurate at the same time.

Workflow Automation

The next big trend is the Automation. The CFD workflow is changing rapidly. Ten years ago, the CFD workflow was a set of individual jobs. Someone had to clean the CAD model, someone had to create the mesh, someone had to set-up and run the simulation, someone had to evaluate the results. This is the past, that never comes back. In the future, all the standard CFD workflows are going to be fully automated. For standard CFD jobs, CFD will work as a calculator. The engineer puts the data in and after a while he picks the results. Automation brings a huge increase in productivity as well as it diminishes a risk of a human error. The automation is extremely effective.

Code Integration

The Integration. As you know, CFD is a piece of the puzzle of a bigger picture called CAE (Computer Aided Engineering) or even bigger picture called Virtual Prototyping, or PLM (Product Lifecycle Management). Even CAE is facing the effect of democratization and automation. The simulations are not only CFD (at all). CAE includes also structural simulations and multiphysics. Users need to solve complex problems. For this reason, in the future, all the CFD codes will have to communicate with the other CAE codes. The users will need to combine the codes. Codes should have strong input and output interfaces. A code can run another code. Or to be run by another code. Perhaps some codes will be acquired and merged.

Optimization

The optimization is a real gamechanger in CFD. In the past, CFD used to give us one answer to one question. Or better said, CFD was used in a way of a cycle of questions and answers. For example, an engineer asked: “What is this car body drag coefficient?” CFD answered: “0.5” Engineer asked again: “What is this enhanced car body drag coefficient?” CFD answered: “0.4” Engineer asked again, CFD answered again … etc. etc. But when the optimization is employed into the CFD workflow (and nobody says it is an easy task), the engineer asks onetime: “What is the best drag coefficient this car body can have?” CFD answers onetime: “0.345” Do you see the difference? The optimization is a big principal shift in using CFD.

Focused vs. General Purpose Codes

This one is the most speculative one. In the past, most of the CFD codes, if not all of them, were General purpose codes - a single CFD code for everything. They were supposed to be used for any CFD job. The opposite to the General purpose code is a focused code - a code specialized in a particular CFD job. I think, in the future, the Focused CFD codes are going to win this competition with General purpose codes. CFD code users, the engineers are not going to be very loyal in the future. They simply want to have their job done. No matter which code they use. They will choose the one that helps. BTW: Have you ever met an engineer who wants to simulate everything? Or anything? No. Not me. Sure, the general purpose codes will not die, they have to focus and create great templates for specialized particular CFD jobs – and that is not easy.

Cloud Computing

This one is the best known. What a big shift from owning the resources to pay per use. Everybody agrees the cloud computing is the future. And the only question is when and how fast is it gonna happen. What is strange to me is, that everybody agrees, but nothing is really happening about it. So far, the transition is much slower than expected. Some people say it is for an IP security reason. I think the users and their bosses are just conservative. Cloud computing brings changes. Big changes. In thinking, in processes, in planning. And most of the organizations do not seem to be ready yet. 

Time & Costs, Scalability

Time always will be critical in CFD. Not speaking about the costs. They are critical pretty everywhere. There is a strong pressure on CFD codes - to be quick and costs efficient. And they should scale. A typical question. If I pay double money or use double the resources, do I get my results two times faster? We can expect a constant pressure on convergence speed, costs, and scalability.

GPU (Graphical Processor Unit)

GPU makes CFD even more complicated. CFD methods are implemented for CPU. Due to the different hardware architecture, at GPU, there is an extra effort to be done (a lot of work). Then the ROI is questionable. In my opinion, but I am quite pessimistic about GPU.

CAD Embedding

What sounds good is adding a CFD code in form of a module into a CAD platform. Which brings CFD back to the issue of General purpose code. And usually, a half solution gives half results. As far as I know, the CAD Embedded CFD is not used for any real CFD projects.

Mobile Devices

Yes. Definitely. Not for simulations of course. But yes. Users like comfort.

Artificial Intelligence

I am a fan of this. Artificial intelligence is a huge trend everywhere and also in CFD. It is in a very early stage now, but very progressive and very promising. The wheels are already in spin. For standard CFD jobs, many results can be precomputed and stored in large databases. AI can play with them and learn.

Open-Source Software

The best for the last. This trend is a matter of my heart. Hard to explain eight years of my life in a nutshell. In the past, CFD in the industry was mainly a field of commercial codes (licensed software). In the last decade, in the same way as in the other software fields, also in CFD appeared open-source codes as an alternative to established commercial codes. At first sight, the greatest benefit of the open-source codes is they are for free. But they are not. The bill has to be paid. Either by time spent (we are speaking about man-years here) or by a consultancy with professionals. The real benefit of the open-sources is they can scale the simulations. The most prominent representative of open-sources in CFD is OpenFOAM. Many companies failed when tried adopting OpenFOAM and didn't know how to use it. Others succeeded. Here are some more thoughts on this topic: [2]. For sure the open-source changed the CFD roadmap for good.

I finish with a point that all the above trends are closely related to each other. Look, the more and more people are involved in CFD. They need a tool that supports them in their real value added work. They are not experts, they need automated solutions. They want the whole picture of integrated solutions. They want to know what is their best-optimized option. They want a focused application that helps. They want to choose the resources. They want to scale their simulations. And they want to do all these in a comfort and easy way.

And finally, yeah, please take this post easy. It is just a summary. Each above topic desires its own post and broad discussion. Let's spot on them next time.