Multiphysics Simulation ... and barbecue party
Multiphysics simulation is often considered as a highly complex engineering field requiring a minimum of PhD degree to resolve accurately a pure validation model with no direct industrial application. Something like “coupled Rayleigh-Bénard convection of unmiscible fluid layers in gradient heated cavity” - I know that one quite well and how difficult it is to place it in a barbecue party …
A few decades ago, the industrial usage of this type of simulation was mainly limited to aerospace applications where pressure on delay to deliver results was proportionally the inverse of the full program length itself. Engineers were manually coding Fortran routines to exchange I/O between research codes; each of these codes being highly efficient in the modeling of a particular physical phenomenon (one for subsonic fluid mechanics, one for composite non-linear thermal, one for phase changing …).
Well, (some) things have changed: (some) research codes became commercial code (like when Steve MacDonald met David Gosman). Some new research projects gave birth to new codes (like SIMPACK in multi body dynamics to be compared to the “old” Adams from MDI and Dads from CADSI). In addition, specialized simulation software editors have been acquired by larger former CAD/CAM/PLM companies …
Some things have not change: the most efficient simulation codes are highly efficient in their “comfort zone” mathematically reproducing the physical phenomenon they have been historically designed for.
What for sure has changed is the coupling of all these highly specifics/efficient/skilled codes. Coupling and integration in a single platform environment. The goal is not to provide a single “general/multi-purpose” simulation code but to take the full benefit of each historical scientific richness of each code and to make a whole solution out of it.
The resulting solution is not a magic piece of artificial intelligence based software. It still requires training and methodology to ensure the quality of results that reflects the reality of the complex physics phenomenon that have to be solved. Best practices integrate the constraints of the available digital mockup and the simplification hypothesis that have been applied to optimize the results restitution time.
Below is a simple illustration of that kind of coupling: PowerFLOW is used as CFD code for what it does the best since the beginning – transient aerodynamics and transient aerothermal, and SIMPACK is used as multibody dynamics code for its unique capabilities in real-time simulation of rail/wheel contact taking into account the full kinematical environment. The coupling is bidirectional and the post-processing is achieved in a unified platform environment with unique realistic rendering capabilities.
The value for the end user is significant:
- Reduced time for technically detailed answer to Request For Quotation,
- Reduced risk for taking the right decision,
- Reduced late stage failure resulting from unpredictable load cases,
And the most important: nice topics to place in a barbecue party ;)
Stay tuned for the next multiphysics industrial applications. It won’t take a decade …
Vincent, I aggree with what you explain in your post. We see it more and more when we talk with customers. Coupling and integration between codes, for simulations including multiphysics domains as the ones shown in this example, but others as well, will be more and more key in the future. This capability and openess of simulation codes that are references in the industry is what will make the real difference when choosing what codes to use. This is already true, but will be more and more important. It will not be anymore what is the best code to solve an specific problem in a single domain, but how several domains can be included in the simulation and reproduce real physical behaviour. This is in many cases the only way to address more innovation.