As a consultant in the area of Computational Fluid Dynamics, it is important to review the journey of the technology over last two decades. At Tridiagonal, we have been a part of this journey since 2006 and its fascinating to see the adoption of technology, its spread across industry verticals, use of high performance computing with public and private clouds, use of open source platforms, attempts to make technology reach in the hands of operators and designers and many such developments. This article is a small attempt to highlight the progress across different stages of evolution and extend it to predict how will the technology shape over next few years…
-
Phase-1: Simple Geometry, Complex Physics
I would refer phase one as time from the year 1998 to around 2006. Prior to year 1998, the basic NS solvers were in place. CFD had already evolved from academic codes to commercial platforms. The results were validated and it was clear that one can use CFD to predict basic fluid flow pattern. Aerospace and Automotive were still the primary domains where CFD was popular. All others were emerging markets.
Phase one were early years of technology establishment. Number of available CFD codes were limited and attempts were made to simulate and analyze flow patterns over many geometric shapes. The technology was getting popular across many industry verticals. Electronics and Semiconductors, HVAC, Power, Oil& Gas, FMCG have been few early adopters. Couple of million cells was still a benchmark. Engineers were handling simple of simplified geometries with complex physics or dealing with only flow simulations on complex geometries.
-
Phase-2: Consolidation; Complex Geometry and Complex Physics
Consolidation happened in this second phase from year 2007 to around 2015. FLUENT was acquired by Ansys; Star-CD was taken over by Siemens. It was then clear that CFD will remain as an important independent technology adjacent to CAD platforms. On the technical front, Engineers had realized that simulation results are within an acceptable range of accuracy and are useful to provide good engineering judgment. It was now time to switch to complex physics along with complex geometry. Many physics-based formulations and integration with supporting physics were getting developed. This includes multiphysics models, Fluids Structure Interactions, Chemical reactions and combustion, etc… In parallel, there were advancements in mesh generation as well. Flow volume extraction, Surface Wrapping, Automated Meshing, and Use of Polyherdal Cells are few major advancements worth mentioning. By now, CFD was established across all industry verticals including the addition of Metals& Mining, Chemicals & Process, Food and even Pharmaceuticals.
The arrival of open-source tool kits is another major addition that happened during the second phase. OpenFOAM, SU2, Salome, Code Saturn/Neptune are among few that emerged and grabbed attention. Most of them happen to be well-established in
academics and are nice platforms for developers. Their presence is felt with repeat simulation work or for the development of some esoteric physics. However, extracting commercial value out of open-source codes for day today’s consulting work has been far from reality.
-
Phase-3: How soon can I get the results?
How soon will I get back the results? This is what we have been hearing for the last few years. Engineers are clear in terms of how and where technology like CFD can be useful. This is observed across all industry segments. The trust is established but now its a race against time along with complex geometry and involved physics…. Software providers have been trying to answer it in multiple ways. Three attempts are clearly visible:
High-Performance Computing:
Use of mesh files with large cell count is evident to capture all minute details and smaller features in the geometry. To keep the computation time under control, high-performance computers are used. Most of the codes offer excellent parallel scalability with additional cores. Even a small problem today, is more like 2 to 3 Million cells now and large ones easily range around 50 Million to even 100 Million cells. Combustion problems we solve at Tridiagonal easily go up to 80 – 100 Million and it is almost a norm.
Automated Workflows:
Automation is the key when it comes to repeat simulations. It helps in reducing manual errors as well as overall turnaround time. There is significant emphasis on automating the work flows and guiding the simulations. Both Ansys and Siemens seem to be focused on this aspect.
Special Purpose Tools:
Application specific tools and their usage seems to be growing as well. I see many CAD-integrated platforms, and verticalized environments with limited modeling capabilities, but they seem to be good at the application they are handling. There is a good market for that where the turn around time as well as the cost are kept under control. Off late, I see some of them offering pay per use type of model as well making the simulation more and more affordable.
Based on all of these, where do I see it is heading over next five years. Can we predict the future…
-
Phase-4: The Future…
Going forward, I see three areas where the users and CFD companies will focus heavily.
Speed of Simulation:
Reducing the run time and making the simulations even faster will certainly be on the top list for all OEMs. This means not just CPUs, but use of GPU based formulations will come up. The presence of new solutions methods like LBM, SPH is felt even today; But I feel it will get deepened over next few years. Seamless integration with additional technologies such as Structural calculations, Particle transport via DEM, etc.. will also make the simulations more advanced and realistic.
Web based Pay per use platforms:
Going forward, I clearly see two types of code/solver options available for the user. General purpose solvers will prevail for all complex simulations. However, web based application specific tools (mostly with pay per use type of business options) will be attractive for many new users. Availability of compute cores over AWS or Azure are going to make it more and more evident. This will also help to expand the market little more around designers/plat operators, etc…
Handling of Simulation Data:
Lot of simulations have been performed over past several years. In parallel, we clearly see that Data Science has emerged and have started influencing the decisions in plants and operations. Use of simulation data with AI/ML to impact business decisions is just round the corner. It will also help to align R&D and Plant think along same direction for higher business impacts.
About the Author:
Ashish Kulkarni is a CFD professional and technology enthusiast since 2000. He is one of the founding members of Tridiagonal Solutions. Over last several years, Ashish has played many different roles right from establishing consulting teams to even leading development of application specific tools. At present, Ashish is VP for consulting business unit at Tridiagonal with focus on growing CFD consulting practice across the globe.
For More Details: www.tridiagonal.com
