Hi everyong, I used to run steady state runs for my model and then use the transient settings with frozen flow eq. and apply tracer experiment (to see the tracer distribution according to the time). Now, becasue of convergence problems, I want to use transient settings as my base of information (instead of steady state) for the tracer test. for this I'm using Transient Statistics option; my question is how can I use this results as an initial data for my second transient tracer run (with expert parameters: solve fluids = f , solve turbulence = f , solve volfrc = f)

Any Comments would be highly appreciated,

Best regards, houman

Hi, houman

I got the same problem. What I did was to continously input the tracer at the inlet as an additional variable. So following the procedure you described, you can get a step input tracer curve(rather than pulse input curve). I am not sure if it's theoritically correct, but it seems work.

You can also try running a steady state simulation using transient solution as initial value. If you can get a good converged results, you can use it for pulse input tracer test simulation.

Hope someone else like Robin, Jeff, etc can give better suggestions for this problem.

Regards!

JP

Dear Robin

Could you please confirm if the tracer simulation method I described is correct? It's also appreciated if you can provide some theoritical analysis for this.

Thanks!

Best Regards!

JP

Hallo JP,

Thanks for your reply, my question was specially/only about using TRANSIENT STATISTICS to run another transient setting only for the tracer distribution... (but what you said about using Additional Variable (as tracer) with a Step function, is also what I'm doing and seems to work ok...) Best regards, houman

Hi Houman,

I'm not sure what it is you're trying to do. Do you want to use the average solution from the transient run to disperse your tracer? The only problem with this is you will not get the turbulent dispersion right, although it may not be too far off. I haven't thought about this much so can't offer much more at this time.

Good luck, Robin

Hi Robin, Thanks for your reply, you are right, it will cause some problems for tracer exp. because, as you said it will destroy some of the eddy motions inside the domain... but somehow I have no other option, since I can't afford running the transient simulations for tracer dispersion for 20 minutes!!! So, I was thinking of getting averaged results from transient run and then freezing the eq. and running the simulation for the tracer...

Best regards, houman

Hi Houman,

That's fine, just as long as you are aware of the implied error. You could do one analysis both ways to compare. By the way, I envy you if a 20 minute run is long!

If you are comparing tracers from multiple inlets, why not define an AV for each inlet and track them simultaneously in a single run? As long as the diffusion coefficient of the tracer isn't a function of the tracer, and the tracer has no influence on the hydrodynamics, the transport equation for the tracer is linear and any combination of tracer boundary conditions can be built up from a linear combination of the solution for a tracer from each boundary condition.

For example, if you have 2 inlets, you could run your transient with AV1=1.0 at inlet 1 and AV2=1.0 at inlet 2. You would then have a transient solution with an average flowfield of AV1 and AV@. Later, if you require the distribution of a tracer from a combination of these inlets, say a concentration of 2.0 at inlet 1 and .5 at inlet 2, you could just open Post and create a new variable as a linear combination of the proportionate amount of the appropriate tracer; simply 2*AV1 + 0.5*AV2.

This helps if your tracers come from inlets. If they come from point sources, you would have to solve a tracer for every node or at least a large number of points, which is probably too expensive.

If it works for you, I'm sure others would be interested.

Best regards, Robin

Hallo Rabin,

Many thanks for your reply, the reason that simulation of 20 min transient run is too much for me is that I only have a desttop PC and for one minute simulation it takes almost 1 day... I'm already using AV setting for my tracer, and I use it from a point source and not from a inlet (in my freezed eq. transient run, and using the seady state results as initial values for it...), So could you please explain me more about this ... (you wrote): ---This helps if your tracers come from inlets. If they come from point sources, you would have to solve a tracer for every node or at least a large number of points, which is probably too expensive---

I mean, what is the differece if I indicate a AV at an inlet or a point source!!! ? Would appreciate your reply, Best regards, houman

Hi Houman,

I thought you meant the run took 20 minutes, not that the simulation was for a 20 minute period.

The transport equations are the same if the source is an inlet or node. The problem only relates to my suggestion. The transport or the scalar from each node will be unique, so if you want to linearly combine runs, you would have to have an existing solution of AV transport from each node. This is clearly impractical.

Regards, Robin

Hi Robin, thanks for your reply, now I see what you mean... regards, houman

Hi, everyone, I am simulating a transient problem. I found that the residence time calculated from the simulation is about 2 or 3 times lower than experiment. what could cause the big deviation?

thanks! lifeng

Hi UY,

I can only say that the Turbulance model which you are using, has an infulence on the residence time...(K-e or SST or ...) and also the dispersion force model (Lopez or Favre averaged drag force). I hope this information might help you...

best regards, houman