problem with Manzari-Dafalias material

[sanjianke]

Hi, everyone,

I tried to use Manzari-Dafalias soil model in a numerical simulation of liquefaction-induced spreading. The numerical model was built in OpenSees, and the Manzari-Dafalias constitutive law was implemented into it as explained here http://opensees.berkeley.edu/wiki/index ... s_Material. With reference to the paper "Simple Plasticity Sand Model Accounting for Fabric Change Effects", I chose the soil model input parameters as suggested by the developer (the unit of the shear modulus constant G0 of 125 is assumed to be in kPa). The parameters I specified are initial void ratio of 0.558, mass density of 2.0 t/m3 and atmospheric pressure of 101 kPa.

It was observed that the elastic gravity analysis can be done quickly and smoothly as supposed. However, after I switched the state of soil from 0 (elastic) to 1 (plastic) using updateMaterialStage command to perform plastic gravity analysis, the problem appeared. It just could't proceed even for one step. If I waited for a long enough time, a message popped up saying "divide-by-zero error coming". I tried to change my solution algorithm, integrator, numberer and convergence criteria. But the problem was still there. On the other hand, I have tried to use either quadUP or SSPquadUP element. Yet I still cannot fix the issue in this way. It is weird this problem came up because I had run a model with PressureDependMultiYield material successfully before. Then I just replaced the PressureDependMultiYield material with Manzari-Dafalias material in my model, why would I encounter this problem? Please shed some light on this problem if you can. Thank you very much for the help.

[parduino]

The Manzari Dafalias model implemented in OpenSees requires the definition of several integration parameters. These parameters have to do with the numerical implementation, not with the model response. Although these parameters provide increase capabilities and flexibility, they can not be chosen arbitrarily. A wrong set of values usually results in the model crashing. To start we suggest using intScheme = 2 and TanType = 2.

Also if the gravity analysis is done using a transient option we suggest adding damping to eliminate the shock waves that are generated when gravity is turned on. This is the most common problem observed when using the model.

It is also important to carefully choose the penalty parameter to enforce constraints. A very large value may result in poor convergence

[Ozgun]

can you try to play with alpha parameter and see if it works?

[mhm]

I have a problem in using material Manzari-Dafalias. When I compare lateral earth pressures computed by Opensees software with manually calculated values(σ_h=k_0*σ_v), computed values by Opensees are so less than manually calculated values and when I use Poisson ratio to 0.35 mentioned computed values are equal to manually calculated values. But suggested Poisson ratio by developer is 0.05. Would you please answer me what is the problem?

[Ozgun]

mhm,

I think it is not a suggested value. 0.05 is used in the generic examples but you should use the Poisson's ratio that is valid for the problem you are working on. otherwise your gravity analysis does not end up with correct initial state of stress.

[sanjianke]

Ozgun wrote:
> can you try to play with alpha parameter and see if it works?

Hi, Ozgun,

What do you mean by "alpha parameter", could you give more details? I appreciate your help.

[sanjianke]

Ozgun wrote:
> mhm,
>
> I think it is not a suggested value. 0.05 is used in the generic examples
> but you should use the Poisson's ratio that is valid for the problem you
> are working on. otherwise your gravity analysis does not end up with
> correct initial state of stress.

If I follow your thought, I should use the specific initial shear modulus G0 in my case instead of the G0=125 in the generic example. Am I right?

[sanjianke]

Ozgun wrote:
> can you try to play with alpha parameter and see if it works?

I guess you meant the alpha value for defining SSPquadUP element. I just used the formula given in the manual to compute it as alpha = 0.25*(h^2)/(den*c^2). Maybe I can try to fix the issue by increasing or decreasing the alpha value as you suggested. Thanks.

[Ozgun]

sanjianke wrote:
> Ozgun wrote:
> > mhm,
> >
> > I think it is not a suggested value. 0.05 is used in the generic examples
> > but you should use the Poisson's ratio that is valid for the problem you
> > are working on. otherwise your gravity analysis does not end up with
> > correct initial state of stress.
>
> If I follow your thought, I should use the specific initial shear modulus G0 in my
> case instead of the G0=125 in the generic example. Am I right?

Yes. You should use G0 value corresponding to your case. GO = 125 is just an example value.

[Ozgun]

sanjianke wrote:
> Ozgun wrote:
> > can you try to play with alpha parameter and see if it works?
>
> I guess you meant the alpha value for defining SSPquadUP element. I just used the
> formula given in the manual to compute it as alpha = 0.25*(h^2)/(den*c^2). Maybe I
> can try to fix the issue by increasing or decreasing the alpha value as you
> suggested. Thanks.

I do not know if playing with alpha parameter is a correct way to fix the convergence issue. But, in my case it fixed. In the manual it is said to be a stabilizing parameter.
Again, i have to admit that I do not know if that is the correct way to deal with issues. It is just my observation. Hope somebody can explain the role of alpha parameter in more detail.

[Rain]

Hi everyone,

Recently, I'm trying to simulate a undrained simple shear test under earthquake loading using the Manzari-Dafalias material. Input parametes are the same as the example gaven in the mannual. However, as the pressure almost reach to the initial vertical pressure, the convergence preblem will occor. By using other materials like the CycLiqCP material, the model runs successfully. How can I fix the issue? Thanks.

[mnt]

I have the same problem.
With quad or 9_4_quadup elements, when i update material stage from 0 to 1, i have the error "Jacobian matrix is singular".