User:EML4500.f08.JAMAMA/FE/JA

From Wikiversity
Jump to navigation Jump to search

Given: 2D Strong Form and BCs[edit]

biunit square

(identity matrix)

,

Essential boundary condition:

on

Natural boundary condition:

none

Failed to parse (Conversion error. Server ("https://wikimedia.org/api/rest_") reported: "Cannot get mml. Expected width > 0."): {\displaystyle \displaystyle }

Find: Temperature inside and error at (0,0)[edit]

Solve for until error is less then at center (x,y)=(0,0) for:

Failed to parse (Conversion error. Server ("https://wikimedia.org/api/rest_") reported: "Cannot get mml. Expected width > 0."): {\displaystyle \displaystyle }

1) Static (steady state): in

1a) 2d LIBF
1b) 2d LLEBF
1b1) Uniform mesh
1b2) Non-uniform mesh


2) Dynamic(transient):

Initial condition

Failed to parse (Conversion error. Server ("https://wikimedia.org/api/rest_") reported: "Cannot get mml. Expected width > 0."): {\displaystyle \displaystyle }

2a)

2d LIBF

Failed to parse (Conversion error. Server ("https://wikimedia.org/api/rest_") reported: "Cannot get mml. Expected width > 0."): {\displaystyle \displaystyle }

2b)

2b1) 2d LIBF
2b2) 2d LLEBF
2b2a) Uniform mesh
2b2b) non-uniform mesh

Failed to parse (Conversion error. Server ("https://wikimedia.org/api/rest_") reported: "Cannot get mml. Expected width > 0."): {\displaystyle \displaystyle }

Solution[edit]

In general by Prof. Vu Quoc's notes


[1]

, where , where represents local node.

For linear 2D case with boundary being square

Bases, are defined to be

where and corresponds to numbering of nodes on element scale.

Rect

also noting that it's a rectangular shape, such that the distance from x: 1-to-2 = 1-to-3 and node 1 = node 4 in 'x' coordiantes, y: 1-to-4 = 1-to-3 and node 1 = node 2 in 'x' coordinates, we substitute for and for and the same for other coordinates






Jacobian Matrix = BN *[mathbf{x} \ mathbf{y}]


</math>

taking the and values from the definition of parent element and substituting into Jacobian matrix we get:

From given Prof. Vu Quoc's Notes

or

With this last information we have a completely defined stiffness element matrix . Also note that since we have, 4 degrees of freedom in 1 element node, we do expect the gram matrix to be 4x4, therefore,

note, to save some space in typing the formula (else it stretched out of the page), the independent variables for were avoided


for the initial square of length of '1', we divided into 4 smaller squares each 0.5 length, the element nodes stiffness matrix came out to be:

K(:,:,1) = 
[  1/24, -1/96, -1/48, -1/96]
[ -1/96,  1/24, -1/96, -1/48]
[ -1/48, -1/96,  1/24, -1/96]
[ -1/96, -1/48, -1/96,  1/24]


K(:,:,2) = 
[  1/24, -1/96, -1/48, -1/96]
[ -1/96,  1/24, -1/96, -1/48]
[ -1/48, -1/96,  1/24, -1/96]
[ -1/96, -1/48, -1/96,  1/24]

this answer shows 2 elements stiffness matrices located along 'x' axes. Noting that because of the symmetry of each elements, the equivalent spacing of each node in elements, and because the material matrix is identity matrix, all stiffness element matrices will be the same. The only differ in element stiffness matrix will occur when we will change the number of stiffness.

,where

force[edit]

From prof. Vu Quoc's notes for elemental force expression

converting to local coordinates

F(:,:,1) = 
[  1/16,  1/16]
[ -1/16,  1/16]
[ -1/16, -1/16]
[  1/16, -1/16]


F(:,:,2) =  
[  1/16,  1/16]
[ -1/16,  1/16]
[ -1/16, -1/16]
[  1/16, -1/16]

Global Matrix creation[edit]

Then we constructed the Global stiffness matrix by the process of superimposing the local stiffness coefficients. Matlab output below shows the process of superimposing the elements to global stiffness matrix.

The matlab code for global matrix construction was written with respect to this element naming system, and then it was extended to encompass other numbers of elements

Global to Local

Original element matrices

K(:,:,1) = 

[  1/24, -1/96, -1/48, -1/96]
[ -1/96,  1/24, -1/96, -1/48]
[ -1/48, -1/96,  1/24, -1/96]
[ -1/96, -1/48, -1/96,  1/24]
K(:,:,2) = 

[  1/24, -1/96, -1/48, -1/96]
[ -1/96,  1/24, -1/96, -1/48]
[ -1/48, -1/96,  1/24, -1/96]
[ -1/96, -1/48, -1/96,  1/24]
     

Superpositions to global matrix

KGlobal =
      1/24          -1/96           0             -1/96          -1/48           0              0              0              0       
     -1/96           1/12          -1/96          -1/48          -1/48          -1/48           0              0              0       
      0             -1/96           1/24           0             -1/48          -1/96           0              0              0       
     -1/96          -1/48           0              1/12          -1/48           0             -1/96          -1/48           0       
     -1/48          -1/48          -1/48          -1/48           1/6           -1/48          -1/48          -1/48          -1/48    
      0             -1/48          -1/96           0             -1/48           1/12           0             -1/48          -1/96    
      0              0              0             -1/96          -1/48           0              1/24          -1/96           0       
      0              0              0             -1/48          -1/48          -1/48          -1/96           1/12          -1/96    
      0              0              0              0             -1/48          -1/96           0             -1/96           1/24    
FGlobal =
      1/8     
      0       
     -1/8     
      1/4     
      0       
     -1/4     
      1/8     
      0       
     -1/8     

Boundary Conditions[edit]

Noting that essential BCs were specified to be on all sides of the 2x2 square. It would mean (viewing from the above global node figure) all nodes around node 5 will be tagged as essential BC.

Global to Local1

Since we can either eliminate the nodes form Global Stiffness matrix to construct another one with Temperature ('u' function) as unknown or just '0' the node values in Global Stiffness matrix and then do inverse and multiply by Force matrix to find other non-constrained dofs.

We wrote a matlab code that automatically constrains the associate nodes and puts their contribution to force function.

Here how code applies BC's

KGlobal =
      1/24          -1/96           0             -1/96          -1/48           0              0              0              0       
     -1/96           1/12          -1/96          -1/48          -1/48          -1/48           0              0              0       
      0             -1/96           1/24           0             -1/48          -1/96           0              0              0       
     -1/96          -1/48           0              1/12          -1/48           0             -1/96          -1/48           0       
     -1/48          -1/48          -1/48          -1/48           1/6           -1/48          -1/48          -1/48          -1/48    
      0             -1/48          -1/96           0             -1/48           1/12           0             -1/48          -1/96    
      0              0              0             -1/96          -1/48           0              1/24          -1/96           0       
      0              0              0             -1/48          -1/48          -1/48          -1/96           1/12          -1/96    
      0              0              0              0             -1/48          -1/96           0             -1/96           1/24    


Kchange =
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              1/6            0              0              0              0       
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              0              0              0              0              0       
      0              0              0              0              0              0              0              0              0       


FGlobal =
     -1/16    
      0       
      1/16    
     -1/8     
      0       
      1/8     
     -1/16    
      0       
      1/16        


Fchange =
     -7/48    
     -7/48    
      0       
     -1/6     
      1/3     
      1/12    
     -7/48    
     -7/48    
      0       


Then the matrices were reduced to the ones of the importance ( Global K matrix after the BCs where applied was isolated so no 0s would be present and attributing F function value(s) where picked by code as well)

KchangeF =
      1/6     


FchangeF =
     1/3     


Finally Temperature at unknown nodes was found by using this formula

u =
     2     

Also output for 4x4 element free node values came out to be:

dcalc =
      2       
      2       
      2       
      2       

as you may see, only 1 node, in this example was present, so only 1 numerical value was generated.

FEATeam5HW7p1

Discussion[edit]

Note 1 important aspect of this output. That is, with forcing function changed to 1, the output of free degrees of freedom is still '2' as if no forcing function was present to begin with. I guess it might be explainable that because of constrained boundary conditions at all 4 ends, in order to keep the temperature constant at all 4 ends while heat was generated, the generated heat had to be instantaneously dissipated though the boundaries of the area, so that boundary temperature would stay unchanged.

As in matrix/numeral explanation, it should be noted, that because essential boundary condition is applied over all 4 sides, it has integration limits (working boundary range) as all element boundary. Therefore, with intorduction with heat source, force instead of being modified, was only scaled (remember force equation):

for f = 0 (HW 6.6)


for f= 1 (HW 7.1)

you may see that force term is a scale only term. Moreover, notice the force global term

For 4 element at f = 0:

FGlobal =
      1/8     
      0       
     -1/8     
      1/4 
      0      

     -1/4     
      1/8     
      0       
     -1/8     


Fchange =
      5/24    
      7/48    
     -1/16    
      7/24   
     -1/3     

     -5/24    
      5/24    
      7/48    
     -1/16    

For 4 element at f = 1:

FGlobal =

     -1/16    
      0       
      1/16    
     -1/8     
      0      

      1/8     
     -1/16    
      0       
      1/16    


Fchange =

     -7/48    
     -7/48    
      0       
     -1/6     
     -1/3     

      1/12    
     -7/48    
     -7/48    
      0       


For 9 element at f = 1

FGlobal =

     -1/36    
      0       
      0       
      1/36    
     -1/18    
      0 
      0     

      1/18    
     -1/18    
      0 
      0     

      1/18    
     -1/36    
      0       
      0       
      1/36    


Fchange =

     -7/108   
     -7/108   
     -7/108   
      0       
    -11/108  
      5/54    
      5/54    

      1/108   
     -5/54  
      5/54    
      5/54    

      1/54    
     -7/108   
     -7/108   
     -7/108   
      0       

, where boxed answer shows free nodes.

Note that force matrix DOES differ, but note that because of the similarity of force functions (scalar multiples of each other, namely, caused because forcing function is a constant and essential boundary condition acts on the same area as forcing function), the global force matrix, since it contained '0's on initial force function, will contain '0's on any other force function, as long as forcing function is scalar and essential B.C is the same as whole element B.C. Therefore, with same used stiffness matrix 'K' the adjusted force matrix with initial values of '0' on free nodes, always came out to be the same for specific number of elements with differing scalar value of forcing function.

As a result, for free node displacements, we observed the constant value of '2' .

--Eml5526.s11.team5.JA 20:51, 6 April 2011 (UTC)

2)[edit]

Transient Term[edit]


Overall Equation[edit]

References[edit]

  1. http://en.wikiversity.org/w/index.php?title=File:Fe1.s11.mtg30.djvu&page=5


NOTE THAT FOR LLIB CASE, THERE WAS ONLY 1 ELEMENT PRESENT (THE GLOBAL MATRIX ELEMENT) AND SINCE ALL BCs ON ELEMENT WERE CONSTRAINED BY ESSENTIAL BOUNDARIES, DUE TO THE LACK OF A FREE ELEMENT, A SOLUTION COULD NOT BE FOUND REGARDLESS OF THE NUMBER OF BASIS FUCNTIONS.

Static using 2d LLEBF uniform mesh[edit]

N=4
N=6
N=8
N=10


As you can see for this case the temperature is 2 at every node which is to be expected.

Static using 2d LLEBF non-uniform mesh[edit]

N=4
N=6
N=8
N=10

Here the answer is also 2 at every node, which agrees with the uniform mesh LLEBF. There is a slight difference between nodes which can be attributed to matlab roundoff error.

Dynamic (case 2b) using 2d LLEBF uniform mesh[edit]

N=2
N=3
N=4
N=5

Dynamic (case 2b) using 2d LLEBF non-uniform mesh[edit]


N=3
N=4
N=5


Eml5526.s11.team5.savery

Eml5526.s11.team5.smith 20:37, 21 April 2011 (UTC)

Eml5526.s11.team5.smith 21:11, 21 April 2011 (UTC)