1 Topic by Alireza (edited by Alireza 04-08-2010 21:48:58)

  • From: Vancouver-Canada
  • Registered: 19-05-2007
  • Posts: 60

Topic: Orthotropic NonLocal

Dear Borek,

I'm implementing my Orthotropic Averaging scheme in OOFEM.
To do so, I have made new nonlocal interface and structural nonlocal interface classes to deal with an array of weights not an scalar weight.

The problem is in the StructuralElement class, the "StructuralNonlocalMaterialExtensionInterface" class is called to handle the preparations for non-local averaging and this is not the interface that my non-local scheme belongs to.

I was wondering what's your suggestion.
Shall I modify the existing non-local interfaces and make them capable of dealing with both isotropic and orthotropic averaging?

Thanks
Alireza

  • From: Vancouver-Canada
  • Registered: 19-05-2007
  • Posts: 60

Re: Orthotropic NonLocal

Basically all the structural classes look at "StructuralNonlocalMaterialExtensionInterface".  so my "StructuralNonlocalOrthoMateiralExtensionInterface" is not recognized and called by teh structuralElement and cross section classes in the code.
then as the second option I tried to modify the NonlocalMaterialExtensionInterface class to make it capable of performing a non-uniform averaging. But the problem is that the whole class is written exclusively to deal with a uniform isoptropic averaging.
For example:
- The structure called "localIntegrationRecord" has a double variable as weight whereas I need a FloatArray;
- The "buildNonlocalPointtable" member is written to deal with scalar averagign,
etc.

Considering all this, what would be the best way to implement a non-uniform non-local scheme with a different wieght structure?

Thanks,
Alireza

3 Reply by bp

  • bp
  • Administrator
  • Offline
  • From: Prague, Czech Rep.
  • Registered: 02-03-2007
  • Posts: 1,047

Re: Orthotropic NonLocal

Dear Alireza,

the principal question here is whether you expect that one material motel can work with different averaging schemes (isotropic, orthotropic, etc). If this is the case, then you must generalize the design of existing implementation. This will include generalization of localIntegrationRecord, update of localIntegrationRecord, etc, as you have suggested.

On the other hand, if you want to develop a new model designed to work only with orthotropic sheme, then you can follow ( and I would recomment it) the idea of introducing StructuralNonlocalOrthoMateiralExtensionInterface, and design appropriate localOrthoIntegrationRecord, etc. The point  is that your StructuralNonlocalOrthoMateiralExtensionInterface should be derived from StructuralNonlocalMaterialExtensionInterface and your material should return this when requested for NonlocalMaterialExtensionInterfaceType. The NonlocalMaterialExtensionInterfaceType does not declare any service specific or limited to isotropic averaging.

Hope that it will help,
Borek

bp's Website

  • From: Vancouver-Canada
  • Registered: 19-05-2007
  • Posts: 60

Re: Orthotropic NonLocal

Dear Boerk,

During this period, I worked on the first idea and applied the modifications to the existing non-local interface that allows orthotropic averaging scheme.
In future if I get the chance, I'll work on your second suggestion.

Cheers
Alireza