Interface_Material Mode and Virtual Thickness

Dear Nitha,

The custom option is for you to be able to directly control the stiffness and strength of the interface behaviour. You can therefore select a dedicated material to better control the interface behaviour. This gives you the option to avoid using the Rinter property which affects both stiffness and strength (as explained in the Reference manual).

Regarding the virtual thickness, this purely affects the numerical part of the analysis. As explained, the virtual thickness does not exist in reality but is the way to handle the presence of the interface in a finite element method world. 
The difference in results can be explained as mentioned in the Reference manual as a consequence of using a larger elastic zone to be considered, which should not be the case.

I would recommend though to try it in a simpler example and not involving water, diaphragm wall and change in soil loading to see in a simpler way how it behaves.

If you have further questions about it please submit a service request. Then, one of our support engineers can help you in detail:
https://apps.bentley.com/srmanager/ProductSupport

Dear Mr. 

@Stefanos Papavasileiou

if we provide the interface elements, we can obtain the interface stresses and we can take that that amount of stresses is transferred to the structures from the soil? and why elastic deformation is ignored as mentioned in the following paragraph?

further i have some questions

01. Virtual thickness of interface=> as mentioned in the reference manual , the deformation depends on the virtual thickness, but based on the equation virtual thickness only affects when we use standard option, if kn/ks is used no effect on it, am i correct? 

Update: I did a simple analysis using cantilever sheetpile with 1.5m depth excavation and results shows if kn/ks method used there is no effect by virtual thickness factor as shown below.

So here i put another question, how do we validate the provided virtual thickness factor is correct or not for a analysis? and how to find the virtual thickness?

02. What is the reason using vi=0.45? any specific reasons or rules and Gi = (R^2)Gsoil <= how it's came?

03. when we reduce the soil structure friction angle we used (2/3)xphi then taking tan function but here it is mentioned as Rinter x tan (phi), why not  tan( Rinter x tan(phi))?

04. The stresses (particularly on the shear stress)  in the soil and the on interfaces varies. My question is, this variation is occurred due to the change of stress angle due to the interface or any other reasons?

Thanks.

Any help please

Dear Nitha,

Interfaces are to model the interaction zone between structure and soil. This zone can be disturbed (for instance due to installation of the structural elements) in which case we may want to give the interface lower stiffness and/or strength than the undisturbed soil.
It doesn't say that elastic deformations are ignored, it only says that the stiffness of the interface is chosen such that the elastic deformation is small. In the interaction zone the majority of the deformation will be plastic due to reduced strength, not elastic. Hence, the spring stiffnesses in the interface are chosen such that the elastic deformation of the interface is not suddenly very large compared to the elastic deformation of the soil.

01)
Indeed, virtual thickness is not used when the user directly inputs the kn and ks values.
How do you define correct? We can't .... we have typically no measurements of displacement of wall and soil so you can't validate if you don't have data. Hence, it's a question that can't be answered. If you do have data, it's a matter of calibrating....

02) vi and Gi area chosen based on experience to give the best numerical stability for a wide range of soil parameters. 

03) In PLAXIS not the friction angle is reduced, but the tangent of the friction angle is reduced. So indeed tan(phi_i) = Rinter*tan(phi_soil). 
Or, if you prefer : phi_i = arctan(Rinter(tan(phi_soil)) 

04) The stresses in the interface are really calculated in the interface. The stresses in the soil come from a cross section for which soil stresses are extrapolated from soil stress points to the soil element nodes and then from the nodes they are interpolated to the cross section. As you can imagine, the extrapolation and interpolation can cause inaccuracy, and if the stress gradient is quite high (which is usually  the case close to a structural element) the inaccuracy is quite high. This is the most probable cause for the differences. The stresses from the interface are the most accurate here.

With kind regards,

Dennis Waterman

Dear Nitha,

Interfaces are to model the interaction zone between structure and soil. This zone can be disturbed (for instance due to installation of the structural elements) in which case we may want to give the interface lower stiffness and/or strength than the undisturbed soil.
It doesn't say that elastic deformations are ignored, it only says that the stiffness of the interface is chosen such that the elastic deformation is small. In the interaction zone the majority of the deformation will be plastic due to reduced strength, not elastic. Hence, the spring stiffnesses in the interface are chosen such that the elastic deformation of the interface is not suddenly very large compared to the elastic deformation of the soil.

01)
Indeed, virtual thickness is not used when the user directly inputs the kn and ks values.
How do you define correct? We can't .... we have typically no measurements of displacement of wall and soil so you can't validate if you don't have data. Hence, it's a question that can't be answered. If you do have data, it's a matter of calibrating....

02) vi and Gi area chosen based on experience to give the best numerical stability for a wide range of soil parameters. 

03) In PLAXIS not the friction angle is reduced, but the tangent of the friction angle is reduced. So indeed tan(phi_i) = Rinter*tan(phi_soil). 
Or, if you prefer : phi_i = arctan(Rinter(tan(phi_soil)) 

04) The stresses in the interface are really calculated in the interface. The stresses in the soil come from a cross section for which soil stresses are extrapolated from soil stress points to the soil element nodes and then from the nodes they are interpolated to the cross section. As you can imagine, the extrapolation and interpolation can cause inaccuracy, and if the stress gradient is quite high (which is usually  the case close to a structural element) the inaccuracy is quite high. This is the most probable cause for the differences. The stresses from the interface are the most accurate here.

With kind regards,

Dennis Waterman

Geez, I've been using Plaxis for years and I never know there was a virtual interface thickness that the user could change. Out of interest I loaded up the Plaxis 2D to see where the thickness parameter. I found it but it's grayed out. I tested a few different settings but it remains greyed out (see below). For what condition would it be editable?

I also find the last sentence of the reference Nina posted(see below) explains some strange Plaxis overlapping results when I tried setting rinter to 0.1 to model the bearing of a smooth frictionless flexible foundation. Indeed I had to use the manual interface setting with very high Kn and very low Ks to get an accurate result. 

Dear Mr. Dennis, 

Thanks for the reply

Dear Mr. Martin, 

What is the equation or correlation you have used to define kn and ks? 

Hi Martin,

I'm going to add to the confusion I'm afraid ....

The parameter you're now referring to is the "real interface thickness", this is available when using HS or HSsmall and the dilatancy cut-off option activated. Dilatancy cut-off is based on void ratio and void ratio is calculated from volume strains. So in order to have a dilatancy cut-off in an interface we need to calculate the volume strain in the interface. Since interfaces don't have a thickness by default, the user should specify this "real interface thickness". Usually one chooses a few centimeters...

The "virtual thickness factor" that Nitha refers to is not a material parameter, it's a property of the interface itself. This virtual thickness factor is used to determine ks and kn when the standard stiffness determination is chosen. 

With kind regards,

Dennis Waterman

Nitha,

In my case I was validating the FEA against a closed form solution which did not consider interface friction, so I wanted an interface between the bottom of the plate and the soil that was as close to frictionless as numerically possible. I didn't want any interface compression because the interface was just a boundary and I wanted all the deformations to be only from the plate and soil. I set the Kn high as I could. 10x9 kN/m^3 caused numerical convergence problems so I used 10x6/m^3, and set the kn as low as possible which was 0.01 kN/m^3. This worked great, and fixed the issues I was having using a low rinter (0.1) which caused the plate to overlap with the soil, which is only possible in the numerical modeling world, and nit not physically possible in the real world.

FYI, using this method, I got near perfect correlation between the Plaxis results and the closed for solution (back calculated rotational stiffness of a rigid, smooth circular footing on an elastic half-space).

I almost always treat the interface as nothing more than a convenient numerical method to cause relative slip between two parts of the model. The Plaxis default setting for the interface usually always work great for me, the one exception is when the rinter is low because, as I just learnt from this thread that the normal stress becomes very low and results in overlapping.

Martin