opm/ifem-elasticity

Elasticity applications built using IFEM

ifem-elasticity's Issues

Same model, different number of gauss points

So I create a unit square with 3x3 grid of C0-lines

This is then raised to polynomial degree (2,1)

I realize this by two different models:

Model 1: Single-patch, 3 C0-lines in the patch (duplicate knotlines in first direction)
Model 2: Multi-patch-model with topology-terms to sow together in a C0-fashion.

Now from a mathematical point of view, these models are identical, but as it turns out they produce a different gauss integration scheme:

Model 1 (single patch)

xinp-file

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>                           
                                                                                  
<simulation>                                                                      
                                                                                  
  <!-- General - geometry definitions !-->                                        
  <geometry dim="2" sets="true">                                                  
    <refine     patch="1" u="2" v="2"/>                                           
    <raiseorder patch="1" u="1" v="0"/>                                           
    <refine     patch="1" u="7" v="7"/>                                           
  </geometry>                                                                     
                                                                                  
  <!-- General - boundary conditions !-->                                         
  <boundaryconditions>                                                            
    <dirichlet set="Edge1" comp="12"/>                                            
    <neumann set="Edge2" direction="2">-1</neumann>                               
  </boundaryconditions>                                                           
                                                                                  
</simulation>                                                                     

Output

 >>> SAM model summary <<<
Number of elements    576
Number of nodes       700
Number of dofs        1400
Number of unknowns    1350
Number of quadrature points 9984 96

Pre-computing sparsity pattern for system matrix (1350x1350): nNZ = 36500

Model 2 (multi-patch)

xinp-file

<?xml version="1.0" encoding="UTF-8" standalone="yes"?>                        
                                                                               
<simulation>                                                                   
                                                                               
  <!-- General - geometry definitions !-->                                     
  <geometry dim="2">                                                           
    <patchfile>patches9.g2</patchfile>                                         
    <raiseorder lowerpatch="1" upperpatch="9" u="1" v="0"/>                    
    <refine     lowerpatch="1" upperpatch="9" u="7" v="7"/>                    
    <topology>                                                                 
      <connection master="1" medge="2" slave="4" sedge="1"/>                   
      <connection master="1" medge="4" slave="2" sedge="3"/>                   
      <connection master="2" medge="2" slave="5" sedge="1"/>                   
      <connection master="2" medge="4" slave="3" sedge="3"/>                      
      <connection master="3" medge="2" slave="6" sedge="1"/>                      
      <connection master="4" medge="2" slave="7" sedge="1"/>                      
      <connection master="4" medge="4" slave="5" sedge="3"/>                      
      <connection master="5" medge="2" slave="8" sedge="1"/>                      
      <connection master="5" medge="4" slave="6" sedge="3"/>                      
      <connection master="6" medge="2" slave="9" sedge="1"/>                      
      <connection master="7" medge="4" slave="8" sedge="3"/>                   
      <connection master="8" medge="4" slave="9" sedge="3"/>                   
    </topology>                                                                
                                                                               
    <topologysets>                                                             
      <set name="Edge1" type="edge">                                           
        <item patch="1">1</item>                                               
        <item patch="2">1</item>                                               
        <item patch="3">1</item>                                               
      </set>                                                                   
      <set name="Edge2" type="edge">                                           
        <item patch="7">2</item>                                               
        <item patch="8">2</item>                                               
        <item patch="9">2</item>                                               
      </set>                                                                   
    </topologysets>                                                            
  </geometry>                                                                  
                                                                               
  <!-- General - boundary conditions !-->                                      
  <boundaryconditions>                                                         
    <dirichlet set="Edge1" comp="12"/>                                         
    <neumann set="Edge2" direction="2">-1</neumann>                            
  </boundaryconditions>                                                        
                                                                               
</simulation>                                                                  

Bonus question: Why does not ModelGenerator work with Elasticity solvers? I try to write <geometry dim="2" nx="3" ny="3">, but it fails on refine since it can't find patch 2 or higher.

Output

 >>> SAM model summary <<<
Number of elements    576
Number of nodes       700
Number of dofs        1400
Number of unknowns    1350
Number of quadrature points 9216 96

Pre-computing sparsity pattern for system matrix (1350x1350): nNZ = 36500

Question

Why does one model produce more quadrature points than the other model?

Segfault on -ignoreSol

Seems to be caused by load having zero size

IFEM-Elasticity/Linear/main_LinEl.C

Line 520 in 4c2c30f

Vectors load(statSol ? (vizRHS ? displ.size() : 1): 0);

then referencing its first element

IFEM-Elasticity/Linear/main_LinEl.C

Line 604 in 4c2c30f

model->extractLoadVec(load.front(),0,"external load");