User Tools

Site Tools


appendices:a15

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
appendices:a15 [2019/06/21 17:04]
helene [Step 3: Compute the norm of convergence CONVE]
appendices:a15 [2020/08/25 15:46] (current)
Line 51: Line 51:
 If $DNOR(1)=10^{-8}$,​ the step is converged if $UNOR(1) \leq 10^{-4}*10^{-2}=10^{-6}$ \\ If $DNOR(1)=10^{-8}$,​ the step is converged if $UNOR(1) \leq 10^{-4}*10^{-2}=10^{-6}$ \\
 If $DNOR(1)=10^{-1}$,​ the step is converged if $UNOR(1) \leq 10^{-4}*DNOR(1)=10^{-5}$ \\ If $DNOR(1)=10^{-1}$,​ the step is converged if $UNOR(1) \leq 10^{-4}*DNOR(1)=10^{-5}$ \\
 +
 +
 +
 ===== Convergence for the forces (NORME2.F) ===== ===== Convergence for the forces (NORME2.F) =====
 +A step is converged for the displacements if $CONVE=\frac{FNOR}{RNOR} \leq PRECF$
 +
 +with:
 +  * $FNOR$ = out-of-balance forces ($F_{ex}-F_{in}$) ​
 +  * $RNOR$ = reaction forces (force on the fixed DOF) 
 +The definition of the norm $CONVE$ may vary, depending on the type of norm $NTNOR$ and the absolute values for convergence COMPA(1...3) = STRAT (15...17) chosen in the execution file, as will be shown below. The algorithm for computing the norm can be decomposed in several steps.
 +
 +==== Step 1: Compute FNORMi and RNORMi for every DOF ====
 +For every degree of freedom (DOF), two quantities can be computed: $FNORM_i$ and $RNORM_i$ as shown in the table below.
 +^  NTNOR  ^  $FNORM_i$ ​ ^  $RNORM_i$ ​ ^
 +|  0  |$\displaystyle\sum_{\text{dof along i}} [\text{out-of-balance forces}]²$|$\displaystyle\sum_{\text{dof along i}} [\text{reactions}]²$|
 +|  1  |$\displaystyle\sum_{\text{dof along i}} abs(\text{out-of-balance forces})$|$\displaystyle\sum_{\text{dof along i}} abs(\text{reactions})$|
 +|  2  |$\displaystyle\max_{\text{dof along i}} abs(\text{out-of-balance forces})$|$\displaystyle\max_{\text{dof along i}} abs(\text{reactions})$|
 +In addition, two quantities are also computed: $NFOR_i$ and $NREA_i$ which contains respectively the number of out-of-balance forces (positive equation number in ID matrix) and reactions (negative equation number in ID matrix) in every direction.
 +
 +==== Step 2: Regroup $FNORM_i$ and $RNORM_i$ by dimension (i.e. force, heat flux and momentum) depending on the type of analysis ====
 +For example, for 3D analysis (NTANA = ±2), we only have one type of dimension (force):
 +\[\left \{\begin{array}{r c l}  FNOR(1) ​ & = & \displaystyle\sum_{i=1}^3 FNORM_i \\  RNOR(1) ​  & = & \displaystyle\sum_{i=1}^3 RNORM_i ​ \\ \end{array} \right . \text{and} \left \{\begin{array}{r c l}  NFO(1) ​ & = & \displaystyle\sum_{i=1}^3 NFOR_i \\  NRE(1) ​  & = & \displaystyle\sum_{i=1}^3 NREA_i ​ \\ \end{array} \right .\]
 +
 +For 3D shell analysis (NTANA = ±8), we have two types (translation and rotation):
 +\[\left \{\begin{array}{r c l}  FNOR(1) ​ & = & \displaystyle\sum_{i=1}^3 FNORM_i \\  RNOR(1) ​  & = & \displaystyle\sum_{i=1}^3 RNORM_i ​ \\ \end{array} \right . \text{ and } \left \{\begin{array}{r c l}  FNOR(3) ​ & = & \displaystyle\sum_{i=4}^6 FNORM_i \\  RNOR(3) ​  & = & \displaystyle\sum_{i=4}^6 RNORM_i ​ \\ \end{array} \right .\]
 +\[\left \{\begin{array}{r c l}  NFO(1) ​ & = & \displaystyle\sum_{i=1}^3 NFOR_i \\  NRE(1) ​  & = & \displaystyle\sum_{i=1}^3 NREA_i ​ \\ \end{array} \right . \text{ and } \left \{\begin{array}{r c l}  NFO(3) ​ & = & \displaystyle\sum_{i=4}^6 NFOR_i \\  NRE(3) ​  & = & \displaystyle\sum_{i=4}^6 NREA_i ​ \\ \end{array} \right .\]
 +
 +For 3D thermo-mechanical analysis (NTANA=6), we also have two types (translation and temperature):​
 +\[\left \{\begin{array}{r c l}  FNOR(1) ​ & = & \displaystyle\sum_{i=1}^3 FNORM_i \\  RNOR(1) ​  & = & \displaystyle\sum_{i=1}^3 RNORM_i ​ \\ \end{array} \right . \text{ and } \left \{\begin{array}{r c l}  FNOR(2) ​ & = &  FNORM_4 \\  RNOR(2) ​  & = & RNORM_4 ​ \\ \end{array} \right .\]
 +\[\left \{\begin{array}{r c l}  NFO(1) ​ & = & \displaystyle\sum_{i=1}^3 NFOR_i \\  NRE(1) ​  & = & \displaystyle\sum_{i=1}^3 NREA_i ​ \\ \end{array} \right . \text{ and } \left \{\begin{array}{r c l}  NFO(2) ​ & = &  NFOR_4 \\  NRE(2) ​  & = & NREA_4 ​ \\ \end{array} \right .\]
 +
 +==== Step 3: Compute the norm of convergence CONVE ====
 +Depending on the type of norm NTNOR and the absolute values for convergence COMPA(1...3) = STRAT (15...17), the computation of the norm CONVE vary. The norm is
 +  * relative if COMPA = 0
 +  * absolute if COMPA > 0 
 +  * either absolute or relative if COMPA < 0
 +The computation is performed as shown in the table below:
 +^  NTNOR  ^  COMPA(i) ≤ 0  ^  COMPA(i) > 0  ^
 +|  0  |\[\sqrt{\frac{1}{NDIM} \displaystyle\sum_{i=1}^{NDIM} \frac{\frac{FNOR(i)}{NFO(i)}}{\max\left(\frac{RNOR(i)}{NRE(i)},​\frac{COMPA(i)²}{NRE(i)}\right)}} \]|\[\sqrt{\frac{1}{NDIM} \displaystyle\sum_{i=1}^{NDIM} \frac{\frac{FNOR(i)}{NFO(i)}}{COMPA(i)²}} \]|
 +|  1 or 2  |\[\sqrt{\frac{1}{NDIM} \displaystyle\sum_{i=1}^{NDIM} \frac{\frac{FNOR(i)}{NFO(i)}}{\max\left(\frac{RNOR(i)}{NRE(i)},​\frac{-COMPA(i)}{NRE(i)}\right)}} \]|\[\sqrt{\frac{1}{NDIM} \displaystyle\sum_{i=1}^{NDIM} \frac{\frac{FNOR(i)}{NFO(i)}}{COMPA(i)}} \]|
 +
 +Remark: the norm can be absolute for some dimensions and relative for others, in which case the formula needs to be adapted. This will be clearer with some examples… \\
 +\\
 +For 3D analysis (NTANA = ±2) with NTNOR=0 and COMPA(1)=0:
 +\[CONVE=\sqrt{\frac{\frac{FNOR(i)}{NFO(i)}}{\frac{RNOR(i)}{NRE(i)}}}\]
 +For 3D shell analysis (NTANA = ±8) with NTNOR=0, COMPA(1)<​0 and COMPA(3)>​0:​
 +\[ CONVE=\sqrt{\frac{1}{2} \left( \frac{\frac{FNOR(1)}{NFO(1)}}{\max\left(\frac{RNOR(1)}{NRE(1)},​\frac{COMPA(1)²}{NRE(1)}\right)} +\frac{\frac{FNOR(3)}{NFO(3)}}{COMPA(3)²}\right)} \]
 +
 +==== Step 4: Check the convergence of the step ====
 +The step is converged for the displacements if $CONVE \leq PRECF$ \\
 +For example, for 3D analysis (NTANA = ±2) with NTNOR=0 and COMPA(1)=-10<​sup>​-1</​sup>​ and PRECF=10<​sup>​-2</​sup>,​ the step is converged if:
 +\[\sqrt{\frac{\frac{FNOR(1)}{NFO(1)}}{\frac{\max(RNOR(1),​COMPA(1)²)}{NRE(1)}}} \leq PRECF \\ <=> \sqrt{\frac{\frac{FNOR(1)}{NFO(1)}}{\frac{\max(RNOR(1),​10^{-2})}{NRE(1)}}} \leq 10^{-2} \\ <=> \frac{FNOR(1)}{NFO(1)} \leq 10^{-4}*\frac{\max(RNOR(1),​10^{-2})}{NRE(1)}\]
 +
 +If $RNOR(1)=10^{-8}$,​ the step is converged if $\frac{FNOR(1)}{NFO(1)} \leq 10^{-4}*\frac{10^{-2}}{NRE(1)}=\frac{10^{-6}}{NRE(1)}$ \\
 +If $RNOR(1)=10^{-1}$,​ the step is converged if $\frac{FNOR(1)}{NFO(1)} \leq 10^{-4}*\frac{RNOR(1)}{NRE(1)}=\frac{10^{-5}}{NRE(1)}$ ​ \\
  
appendices/a15.1561129446.txt.gz · Last modified: 2020/08/25 15:33 (external edit)