Learning Forum

[Masooma_Qizilbash12]

hello friends.

I am modeling particle size distribution of PEI in PVA but when I put smaller diameters in the population balance tab it gives an error of divergence detected in AMG solver while when I use larger diameters it accepts and runs what this error means how can I get rid of it 

thank you and waiting for anybody who can help

regards,

masooma

[Rob]

If you look at the results with a larger diameter, what's going on in the flow? Divergence tends to be down to mesh, boundary conditions or solver settings. Given the model can run that then suggests something changes in the flowfield between the two solutions. 

[Masooma_Qizilbash12]

i have a particles size range from 2oo nm to 3000nm from 600 to 3000 it is ok but when i decrease the range towards smaller diameters it says that divergence detected in AMG solver also i want to ask that how to calculate  the turbulent kinetic energy and specific dissipate rate. a member of the ansys replied that i have to calculate both turbulent term by formula but in formula there is a velocity gradient i only have one velocity initially how can i dteremine the velocity gradient before starting the calculation.

thank you for the support

[Rob]

I'd not expect that to make much difference as sub 10micro (ish - very ish) particles tend to follow the carrier fluid fairly closely. Read up on relaxation time and settling velocity for more background. In the model with the larger size, how does the size distribution look? 

Where are you trying to apply the turbulent values? If they're boundary conditions then I'd favour Turbulent Intensity and Hydraulic Diameter for internal flows. Unlike the flow & thermal bc's the turbulent values need to be close to reality as the solver will rectify errors over the first few domain length scales (pipe diameter for internal flow) hence always allowing sufficient space upstream of the area of interest. You don't put the experimental probes right next to the pump so why constrain the CFD model?

[Masooma_Qizilbash12]

Hello Rob.

thank you for the reply. I am modelling distribution particle size in a polymer PVA system just like oil-water system... I don't have I let's and outlets I am only having a beaker having PVA a continuous phase stirring at 5000 rpm .. I want to see the particle distribution on the interface of rotar and stator I am achieving convergence succefully but as far as results accuracy is concerned I am not having it ... let me add one more thing that from the DSL GRAPH I have a particle size distribution jn an a shaped curve but in ansys fluent decreasing number density i observing please help me what should I do to have an accurate results

details of the model are: k-e,kw models eulerian multiphase discrete population e balance modelling 7 bins 0.1 all except 3, bin 3= 0.4 particle dia to b modelled is 1e-7 to 6e-6 from the DSL graph aggregation turbulent and luo both tried breakage luo ... time step size 9.9e-7 fluent automatically select this time step size .... at 250 number of timesteps convergence is occurring but not accuracy please help 

i shall be thankful to you.

kind regards,

masooma 

[Masooma_Qizilbash12]

These all scenarios are tried but In vein

[Rob]

How are you initialising the flow? I've seen a few cases recently with "relative" on the initial conditions so the rotating zone thinks it's spinning very quickly relative to the bulk flow, that causes high shear, and with size distributions could cause you to hit limits. If the limit was previously for a larger particle that may just prevent divergence. 

[Masooma_Qizilbash12]

I am initializing with the relative to cell zone option. there is no acceptance of smaller particle diameter yet only diameters from 6 e-7 instead of 1e-7 m is considered plus I do not have accuracy but convergence i don't know what I am doing wrong please guide me. Even after adaption, there is no sign of accuracy . 

thank you

masooma

[Rob]

What happens if you start with the working model and then adjust the size bins? Do NOT initialise again, just iterate. 

[Masooma_Qizilbash12]

thank you, Rob,

do you mean I should try to run the model with the default values just adjusting the bins?? if yes then I did every possible way, I tried several times with default values different turbulence models, etc. In the end, I can adjust the convergence only like on default values I had a divergence detected in epsilon then I adjusted the epsilon under relaxation factor and the problem was gone after that at 250 time steps by having 30 mas iterations per time steps and 9.8e-7 time step size my system got converged but the number density value was far low as compared to the experimental value... similarly tried k omega etc i am now able to get the convergence but nit the accuracy one more thing may i know the max value of the number density that is acheieved in the fluent?

thnak you 

 

[Masooma_Qizilbash12]

Rob, 

i should mentined here that i was modelling this system in transient state now i am running it in steady state as impeller is moving at constant speed of 10000 RPM... all parameters are stable in scaled residuals except the bins i defined in population balance balance modelling part. i intialize the case with 7 bins each bin having 0.1 value excep bin 3 o.4 but it ran for 200 iterations ad give the divergence detected at every bin . then i change the scheme to bin 20 and initialized the bins with the specific arrangement in such a way that the bin having mean particle size was assigning higher values than other even now solver is not initiaing and at the begining giving the divergence detected in 3 to 20 bins and not the 1st two bins please guide... i think i just need to adjust the bins arrangements because all other parameters shows more or less straight line even in monitors all parameters like vol frac and number density is stable. if you have any suggestion on arranging the bin sizes then please help

 

thank you 

[Rob]

I'd reduce the number of bins, try 5-7 and see how that behaves. The more complicated the model, or the more the result can (rapidly) change the more likely you are to have problems, reducing the complexity is then the starting point for resololving things. 

Unfortunately I can't review the case or offer specific guidance in order to remain within the legal requirements of working on the Community. 

[Masooma_Qizilbash12]

ok Rob, 

after doing some variations in the settings system ran without error but it is not converging what could be the possible reasons behind that also see the attached picture to understand the behavior of scaled residuals currently epsilon is increasing but in such a way that it is looking constant. after 1000 iteration it was on 1e-1 i don't know what should i change now should i adapt a region... the system in the picture is without patching without adaption.. results are not accurate ... also if you have any information on how to compare the DSL number graph with the simulation number density graph then please guide me ...thank you 

best regards,

masooma

[Rob]

You're running what may be a very transient system in steady state, how does it behave if you switch to transient from the current result? 

DSL? I see far too many TLAs to keep track of.... 

[Masooma_Qizilbash12]

rob, thank you for the email.

i have the results of the above pictured simulation but now the thing is i only ran the model with two bins initialy from which one bin is showing answer having 4 percent of error while the other has a significant error but the trend is same as experimental.... without adaption and patchin ...is working with only two bins authentic? i want to run the model with more bins to have the distribution curve. before i will try to run it in transient and i will inform about the results

thank you for the help 

masooma

[Rob]

I can't answer that question. In some models a fixed particle/droplet size is appropriate, in others 5-10 bins may be needed: it's situation dependent.