CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   polyflow (http://www.cfd-online.com/Forums/fluent/91307-polyflow.html)

hamid1 August 6, 2011 17:31

polyflow
 
does anybody know how can i sue polyflow for the viscoelastic materials with "complex mudulus" properties?
thank you

nclemeur September 7, 2011 22:14

Quote:

Originally Posted by hamid1 (Post 319149)
does anybody know how can i sue polyflow for the viscoelastic materials with "complex mudulus" properties?
thank you

You will usually need more information than simply complex moduli properties. You can use the moduli (storage and loss moduli) to fit the linear properties of your material (relaxation time and viscosity parameter) and then you need to use a non-linear property to fit the others parameters. You can use steady shear viscosity or elongationnal viscosity for that.

I have written a software called RheoChart (google for it) that can help you in this procedure.

Amir September 8, 2011 04:05

Quote:

Originally Posted by nclemeur (Post 323349)
You will usually need more information than simply complex moduli properties. You can use the moduli (storage and loss moduli) to fit the linear properties of your material (relaxation time and viscosity parameter) and then you need to use a non-linear property to fit the others parameters. You can use steady shear viscosity or elongationnal viscosity for that.

I have written a software called RheoChart (google for it) that can help you in this procedure.

Dear Nicolas,

Do you have any experience in using Cox-Merz rule for this purpose? I have some graphs for dynamic modulus in linear regime but I don't have any data in non-linear. As I know, we cannot use linear properties such as dynamic modules to find non-linear variables but you said that you used both linear and non linear properties simultaneously! would you elaborate it more? (is it physically correct?) I guess you're using linear and non-linear spectra as two parallel elements; right?

PS: Thanks for your great software; when do you intend to release newer version? (is there any chance that creep properties are also included?)

Bests,

nclemeur September 8, 2011 04:16

Quote:

Originally Posted by Amir (Post 323367)
Dear Nicolas,
Do you have any experience in using Cox-Merz rule for this purpose?

Not really, but that might be better than nothing anyway. If you don't have data, you have to make some assumptions and that's propably a reasonnable one if your flow is dominated by shear effects. Now, you'll find that in many case, some non-linear parameters are not really sensitive to shear properties (ie they don't have a big influence on the shear behaviour, so if you have only shear data, these parameters can't be determined really easily). It is usually, much better to have elongational data, but these data are also usually much more difficult to obtain.

Quote:

Originally Posted by Amir (Post 323367)
but you said that you used both linear and non linear properties simultaneously! would you elaborate it more? (is it physically correct?)

I am not sure I have said that. I said that in the first step you fit the parameter with the linear data and then you use non-linear data to fit non-linear parameter...

Amir September 8, 2011 04:59

Dear Nicolas,

I apologize if I couldn't understand the point; I'm really new in viscoelastic modelling. As you know, I want to model a gel-like material in non-linear regime; because I don't have non linear data, I want to use cox-merz rule to obtain steady shear viscosity and finally as you've suggested before, I'll use one GNF mode and some non linear mode such as giesekus and I'll try to find variables of these modes with your software. Upon your experience, is this procedure correct? As you see, I don't use linear data directly. would you please elaborate your procedure in this example?

PS: Thanks for your great software; when do you intend to release newer version? (is there any chance that creep properties are also included?)

Bests,

nclemeur September 8, 2011 07:55

Quote:

Originally Posted by Amir (Post 323379)
I apologize if I couldn't understand the point; I'm really new in viscoelastic modelling. As you know, I want to model a gel-like material in non-linear regime; because I don't have non linear data, I want to use cox-merz rule to obtain steady shear viscosity and finally as you've suggested before, I'll use one GNF mode and some non linear mode such as giesekus and I'll try to find variables of these modes with your software.
Upon your experience, is this procedure correct? As you see, I don't use linear data directly. would you please elaborate your procedure in this example?

It is definetely possible to use a Generalised Newtonian with some Giesekus mode within RheoChart, but it might not be really necessary to include the purely viscous component. How many viscoelastic modes are you intending to use? I am a bit lost when you say you don't use linear data directly. How do you determine the relaxation time and viscosity parameter? One procedure would be to select your relaxation time based on the frequency/shear rate that you want to cover. For example, just distribute them in a logarithmic way on your range (ie 0.01, 0.1, 1.0,...) and then find the viscosity (or modulus) by fitting your storage and loss moduli curve. For that, you need to keep the relaxation time and your non-linear parameters constant. After this step, you have determined your linear parameter (relaxation time and modulus). Now, in the next step, you keep the linear parameters constants and you determine the non linear parameter(s) based on the some non linear property.

Quote:

Originally Posted by Amir (Post 323379)
PS: Thanks for your great software; when do you intend to release newer version? (is there any chance that creep properties are also included?)

Thank you! There should be a release reasonnably soon (I can't really commit to a date, but I think one or 2 months is probably reasonnable.) I am actually working at implementing the creep kinematic in shear flow.

Amir September 8, 2011 09:20

Quote:

Originally Posted by nclemeur (Post 323390)
It is definetely possible to use a Generalised Newtonian with some Giesekus mode within RheoChart, but it might not be really necessary to include the purely viscous component.

Yes I know but in OpenFOAM, which is a solver I'm going to use, I have to set solvent viscosity specifically; so I should use a GNF mode.
Quote:

Originally Posted by nclemeur (Post 323390)
How many viscoelastic modes are you intending to use? I am a bit lost when you say you don't use linear data directly. How do you determine the relaxation time and viscosity parameter? One procedure would be to select your relaxation time based on the frequency/shear rate that you want to cover. For example, just distribute them in a logarithmic way on your range (ie 0.01, 0.1, 1.0,...) and then find the viscosity (or modulus) by fitting your storage and loss moduli curve. For that, you need to keep the relaxation time and your non-linear parameters constant. After this step, you have determined your linear parameter (relaxation time and modulus). Now, in the next step, you keep the linear parameters constants and you determine the non linear parameter(s) based on the some non linear property.

Thank you very much Nicolas; your explanation was really perfect. I think my understanding of the procedures which your software works was wrong. I thought that by providing non linear data for set of non linear modes such as giesekus, I could obtain three parameters.(G,\lambda_{b},\alpha)
Now it's clear that G and \lambda_{b} should be set by linear data and \alpha with non linear one.
I think that in student license there shouldn't be any restriction over number of modes, so I will increase mode numbers gradually in order to achieve best fitting.

Thanks again,

nclemeur September 8, 2011 09:30

Quote:

Originally Posted by Amir (Post 323403)
Yes I know but in OpenFOAM, which is a solver I'm going to use, I have to set solvent viscosity specifically; so I should use a GNF mode.

Is a solvent viscosity really necessary? I would suspect this is only optional, but I obviously don't know OpenFOAM. In Polyflow, the solvent viscosity is optional.

Quote:

Originally Posted by Amir (Post 323403)
I think that in student license there shouldn't be any restriction over number of modes, so I will increase mode numbers gradually in order to achieve best fitting.
Thanks again,

You are welcome. Yes you are right, you have multi-mode in the student version. But this is becoming really specific to RheoChart, so maybe we should move the discussion to a new thread in the RheoChart forum.
Cheers

Nicolas

Amir September 8, 2011 09:45

Quote:

Originally Posted by nclemeur (Post 323406)
Is a solvent viscosity really necessary? I would suspect this is only optional, but I obviously don't know OpenFOAM. In Polyflow, the solvent viscosity is optional.

Unfortunately, it's necessary; additionally, you have to introduce solvent viscosity as number of as polymer modes! as a first guess, we can set solvent viscosity of each mode as (solvent viscosity)/(mode number); in other word, each mode has it's own solvent viscosity. we have to use this procedure because of DEVSS method of solving momentum equation.
Quote:

Originally Posted by nclemeur (Post 323406)
You are welcome. Yes you are right, you have multi-mode in the student version. But this is becoming really specific to RheoChart, so maybe we should move the discussion to a new thread in the RheoChart forum.

Sorry for inconvenience

Bests,

nclemeur September 8, 2011 21:07

Quote:

Originally Posted by Amir (Post 323409)
In other word, each mode has it's own solvent viscosity. we have to use this procedure because of DEVSS method of solving momentum equation.

Ok that might be right, it's been a long time I have not used these techniques. But it might also be worth checking as I have seen a presentation where the solvent viscosity was not mentionned on every simulation. It might be an omission though. Check for example the parameters for the multimode DCPP results in
http://powerlab.fsb.hr/ped/kturbo/Op...cFluidFoam.pdf

There is no mention of a solvent viscosity there... Anyway, I am not the best person to answer that question.

Cheers

Nicolas

Amir September 9, 2011 03:12

Quote:

Originally Posted by nclemeur (Post 323481)
There is no mention of a solvent viscosity there... Anyway, I am not the best person to answer that question.

Dear Nicolas,

Thanks for your time. About DCPP model you have mentioned, I don't agree with you; in page 16, polymer shear stress is computed by usiig parameters of this model and after summing these stresses, DEVSS method is used in page 18; note that solvent viscosity is in left hand side of this equation:
-(\eta_{s}+k)\nabla . (\nabla U)
It's worth to note that virtual function is used for evaluating shear stress in OpenFOAM, so the procedure of computing stress is same for all models, just stress formulations change.
Here, because I know that the solvent is water, I'll introduce one constant viscosity mode and hold its value during fitting in your software.

Bests,

nclemeur September 9, 2011 03:24

Quote:

Originally Posted by Amir (Post 323496)
Dear Nicolas,

Thanks for your time. About DCPP model you have mentioned, I don't agree with you; in page 16, polymer shear stress is computed by usiig parameters of this model and after summing these stresses, DEVSS method is used in page 18; note that solvent viscosity is in left hand side of this equation:
-(\eta_{s}+k)\nabla . (\nabla U)
It's worth to note that virtual function is used for evaluating shear stress in OpenFOAM, so the procedure of computing stress is same for all models, just stress formulations change.
Here, because I know that the solvent is water, I'll introduce one constant viscosity mode and hold its value during fitting in your software.
Bests,

No where it is said that \eta_{s} can't be set to 0. It is certainly more stable numerically to have a non zero value, but it can be set to 0. Which is what was done in the multimode DCPP or Giesekus examples in the presentation above. Anyway it is probably better to ask the author of that presentation.

Amir September 9, 2011 03:44

Quote:

Originally Posted by nclemeur (Post 323497)
No where it is said that \eta_{s} can't be set to 0. It is certainly more stable numerically to have a non zero value, but it can be set to 0. Which is what was done in the multimode DCPP or Giesekus examples in the presentation above. Anyway it is probably better to ask the author of that presentation.

As you know: \nabla .\tau = \nabla . (\sum \tau_{p}) + \nabla . \tau_{s}
So it seems reasonable that we consider solvent viscosity as one or more separate modes and the formulation won't change. It's a great flexibility of your software which can handle both.(with or without solvent mode)
As you said, I think also that this separation in considered to improve stability. But in many tutorials, solvent viscosity is considered.

Bests,


All times are GMT -4. The time now is 09:41.