Is real time CFD simulation possible
 

Friends,

I am a research student working on biomass combustion simulation using CFD mainly fluent package. I would like to extend my research in the direction of real time CFD simulation. So far I have simulated the biomass combustion system, by applying the real time boundary conditions basically steady state analysis.
I am planning to take real time data from the combustion system as input to the CFD code, then based on output of CFD, i would like to control the combustion air supply System.

The reason to do this is to improve the fuel air mixture, specially in batch fuel addition type combustion systems.

So far I have used only Fluent, Can you please tell me that, is this feasible? if yes where can I found relevant material.

thanks in advance
Rakesh

http://www.cfd-online.com/Forums/att...1&d=1335359662

Short answer: I highly doubt this is feasible. I'd need more information about the simulations you want to run and on what type of machine you'd be running them. Let's just you that you probably aren't going be measuring the run time in minutes, much less real time. Post some more information on your simulations and maybe someone can make a guess of how long they'd take to run.

Rakesh,

I once read a paper on trying to do real-time CFD simulation (for smoke modeling for use by fire emergency crews supposedly). Their short coming was: do not worry with the mesh resolution and turbulence model. Using a coarse mesh is probably the only way to reduce the computational time enough to get short turn-around time.

If I was to conduct such project, I would use a different approach to achieve real-time control: use the CFD model to build/train a response network in advance, and implement the network to have intelligence in two ways (learn from the reactions of the real system, and allow for interaction with CFD for points that are outside of the initial training).

Hope this help.
Julien

Hello,

This is challenging considering the solution time. The best way is, make list of several possible input parameters and solve CFD for all these points. Make a correlation for input to output based on the several CFD results and input this as a logic to your system to control A/F ratio.

Thanks and regards,
Hari

I would say, this approach is extremely impractical.

The right way to do it would be to create a reduced order model of your system using a combination of experimental data, cfd and underlying physics.
Use this reduced order model for feedback control.

Feedback control systems are quite robust to modelling errors.

There are a few of these on Youtube:

https://www.youtube.com/watch?v=vOFcHqImXJ8

Pretty at least, but the real-time aspect is not fully explained (very slow flows in this case I assume).

I think it is feasible to run a real-time fast flow simulation with a few tens of thousands cells on a normal desktop computer. Add reaction mechanisms and that number would plummet. ;)