[gejiabin]

icoFsiElasticNonLinULSolidFoam is available in foam-extend-3.0.
After reading the papar“performance of a new partitioned procedure versus a monolithic in fluid-structure interaction”.
http://web.mit.edu/kjb/www/Principal...nteraction.pdf
I try to implement IQN-ILS algorithm in icoFsiElasticNonLinULSolidFoam according to paper.Blow is my code.

1.Add two line into createInterfaceFields.H

SLList<scalarRectangularMatrix> deltaInput;
SLList<scalarRectangularMatrix> deltaOutput;

2.The main code is in setInterfaceDisplacement.H

  else if (couplingScheme == "IQN-ILS")
    {
     Info<< "IQN-ILS iteration: "<<endl;
     //if(outerCorr>20)
     //{
     // deltaInput.removeHead();
     // deltaOutput.removeHead();
    // }
     //-------------------------------------------------------------------------
     int numColumn(0);
     int dim=2;//3
     int numDegreeOfFreedom( dim*mesh.boundaryMesh()[fluidPatchID].nPoints() );
     scalarRectangularMatrix vMatrix(numDegreeOfFreedom,deltaInput.size()+1,0);
     scalarRectangularMatrix wMatrix(numDegreeOfFreedom,deltaInput.size()+1,0);
     scalarRectangularMatrix vVector(numDegreeOfFreedom,1,0);
     scalarRectangularMatrix temp(numDegreeOfFreedom,1,0);
     
     //1.Assemble vMatrix  
     for(int id=0;id<numDegreeOfFreedom;id=id+dim)
      for(int i=0;i<dim;i++)
       temp[id+i][0]=(fsiResidualOld[id/dim][i]-fsiResidual[id/dim][i]);
     //
     forAllIter(SLList<scalarRectangularMatrix>,deltaInput,iter)
     { 
       (*iter)=(*iter)+temp;
       for(int i=0;i<numDegreeOfFreedom;i++)    
        vMatrix[i][numColumn]=(*iter)[i][0];
       numColumn+=1;
     } 
     //
     deltaInput.append(temp);
     for(int i=0;i<numDegreeOfFreedom;i++) 
      vMatrix[i][numColumn]=temp[i][0];
     
     //2.Assemble vVecotr
     for(int id=0;id<numDegreeOfFreedom;id=id+dim)
      for(int i=0;i<dim;i++)
       vVector[id+i][0]=-fsiResidual[id/dim][i];
     
     //3.Assemble wMatrix 
     numColumn=0;
     for(int id=0;id<numDegreeOfFreedom;id=id+dim)
      for(int i=0;i<dim;i++)
       temp[id+i][0]=(fluidPatchPointsDisplOld[id/dim][i]-fluidPatchPointsDispl[id/dim][i]);
     //
     forAllIter(SLList<scalarRectangularMatrix>,deltaOutput,iter)
     {  
       (*iter)=(*iter)+temp;
       for(int i=0;i<numDegreeOfFreedom;i++)    
        wMatrix[i][numColumn]=(*iter)[i][0];
       numColumn+=1;
     } 
     //
     deltaOutput.append(temp);
     for(int i=0;i<numDegreeOfFreedom;i++) 
      wMatrix[i][numColumn]=temp[i][0];
       //4.Update fluidPatchPointsDispl
     fluidPatchPointsDisplOld = fluidPatchPointsDispl;
     multiply(temp,wMatrix,SVDinv(vMatrix),vVector);
     for(int id=0;id<numDegreeOfFreedom;id=id+dim)
      for(int i=0;i<dim;i++)
       fluidPatchPointsDispl[id/dim][i]+=(temp[id+i][0]-vVector[id+i][0]) ;
     
      Info<<"end:"<<endl;
     //-------------------------------------------------------------------------
     }
}

But the new solver can't convergence for the benchmark proposed in the Hron & Turek paper.









But I change the code line from:fluidPatchPointsDispl[id/dim][i]+=(temp[id+i][0]-vVector[id+i][0]);
to:fluidPatchPointsDispl[id/dim][i]+=(temp[id+i][0]);
Things get a little better,but it still can't meet convergence criteria.
[IMG]http://b170.photo.store.qq.com/psb?/V11dkwCX1wKa4a/GjfsQ*slhvGFcYuS46SR.znc5g*L*kXjlU4qiJ.g7og!/b/dEguYGVAKAAA&bo=1wKAAgAAAAADAHI![/IMG]

[stephie]

Hey,

did you find a solution?
I work with this benchmark too and I am really interessted in implementing the new algorithm in the HronTurekFsi Case.

I would be very grateful for your help.

Best regards,
Stephie

[sharonyue]

Sorry,

But actually its quite excited to see "image can not be cited from QQ without permission", um. In my primary language.

[pi06jl6]

I am currently working with this case, Stephie, keep us posted about your findings! Are you using the FSI-PFS-2A setting? I am using M. Breuers data as reference, what data are you using? Could you describe your case setting? Best Regards /Johan

FYI: IQN-ILS is implemented in icoFsiElasticNonLinULSolidFoam.

[rasool_soofi]

Hello Dears
i am using this solver for my case.
in the tutorial case of this solver the properties of fluid and solid set to

fluid density = 1 kg/m3 kinemaic viscosity = 10e-3
solid density = 10 kg/m3

what is the meaning of this properties.?
when i set my real properties for blood and vessel

blood density = 1000 kg/m3
vessel density = 1200 kg/m3

my solution crashed by " complex eigenvalue detected for tensor ...." error.

best regards

[anandsudhi]

Hi I know this is an old post. But could your give me a link to the implementation of IQN-ILS in icoFsiElasticNonLinULSolidFoam. I am not able to find it in foam extend 3.1 as that portion is commented.

[pi06jl6]

Originally Posted by anandsudhi

Hi I know this is an old post. But could your give me a link to the implementation of IQN-ILS in icoFsiElasticNonLinULSolidFoam. I am not able to find it in foam extend 3.1 as that portion is commented.

setInterFaceDisplacement.H

I have implemented my own since the provided use explicit Q, i prefer Q free implementation using householder method.