LE2: Cylindrical shell bending patch test. Moment on shell Lip

General Questions
1

Hi Matej,

Not sure where my problem is ?
I’m appling a moment to a set of nodes on the lip of shell.
Prepomax translates the moment as a unique Cload card , direction 6 (Mz) , value 500Nm.

I’m not obtaining what I would expect so I’m considering:

1-My Prepomax Set up is wrong.
2-The moment is not properly distributed.

Each node on the lip has a different weight (I’m using second order shell elements).
Shouldn’t the moments be distributed with the same proportions that are used for example for traction?
If I distribute the moment by hand , the profile looks fine and section print check is fine too. Constant Stress across the curved shell. Same moment input and Output.
If I distribute the moment using the GUI, VM shows some peaks on the corners and Section print is wrong (Too large). Kind of problem seen when one aplies Cload to a set of nodes.

imagen
imagen1132×742 125 KB

.fctbNone{ color:#000000; } .fctbStyle0Style1{ background-color:#e6ffe6;color:#008000; } .fctbStyle1{ color:#0000ff; } ** Name: Moment-1 *Cload, Amplitude=Tabular-1 Node_Set-MZ1, 6, 500

Distributed by hand.

imagen
imagen1366×489 32.2 KB

Unique card.

imagen
imagen1089×645 121 KB

LE2: Cylindrical shell bending patch test

2

This is how ABAQUS aplies a Moment / Unit lenght to the lip nodes in this problem.

*CLOAD
NODE DOF COEF
5,6,0.041667E3 5 6 4.17E+01 1/12
10,6,0.166667E3 10 6 1.67E+02 1/3
15,6,0.083333E3 15 6 8.33E+01 1/6
20,6,0.166667E3 20 6 1.67E+02 1/3
25,6,0.041667E3 25 6 4.17E+01 1/12
5.00E+02 
*HEADING
: NAFEMS TEST LE2, Cylindrical Shell Bending Patch Test  [S8R]
*NODE
 1, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0
 5, 1.0, 30.0, 0.0, 1.0, 0.0, 0.0
 21,1.0, 0.0, 0.5, 1.0, 0.0, 0.0
 25,1.0, 30.0, 0.5, 1.0, 0.0, 0.0
 13,1.0, 20.0, 0.3, 1.0, 0.0, 0.0
*NGEN,NSET=EDGEAB
 1,21,5
*NGEN,NSET=EDGECD
 5,25,5
*NGEN,NSET=EDGEAD
 1,5
*NGEN,NSET=EDGEBC
 21,25,
*NGEN
 11,13
 13,15
 3,13,5
 13,23,5
 6,8
 8,10
 16,18
 18,20
*NSET,NSET=NALL,GEN
 1,25
*NMAP,NSET=NALL,TYPE=CYLINDRICAL
0.,0.,0.,0.,0.,1.
1.,0.,0.
*ELEMENT,TYPE=S8R,ELSET=ALLELS
 1,1,3,13,11, 2,8,12,6
*ELGEN,ELSET=ALLELS
 1,2,2,1,2,10,10
*MATERIAL,NAME=MAT
*ELASTIC,TYPE=ISOTROPIC
2.1E5,0.3
*orientation, name=ori, system=cylindrical
 0.,0.,0., 0.,0.,1.
 1,0.
*SHELL SECTION,MATERIAL=MAT,ELSET=ALLELS, orientation=ori
0.01, 
*TRANSFORM,TYPE=C,NSET=NALL
0.,0.,0.,0.,0.,1.
*BOUNDARY
 EDGEAD,ZSYMM
 EDGEBC,ZSYMM
 EDGEAB,1,6
*RESTART,WRITE
*STEP
*STATIC
*CLOAD
 5,6,0.041667E3
 10,6,0.166667E3
 15,6,0.083333E3
 20,6,0.166667E3
 25,6,0.041667E3
*EL PRINT,POS=AVERAGED AT NODES
S, 
*EL FILE,POS=AVERAGED AT NODES
S, 
*END STEP
3

if i remember properly, this only related to plot stress along thickness problem in PrePoMax, Paraview also, but it’s not for CGX. Switching to quadratic shell element with reduced integration and composite layered shell can be displaying result as expected.

4

You cannot use Moment item to distribute a load evenly. The moment item creates a concentrated load and it will only assign the same value of moment to all nodes. This works only for equally sized linear elements.

As you figured out, for parabolic elements, the weights are 1/12, 1/3, and 1/12, but don’t forget to account for the element area.

5

Do you think that a Moment /Unit Lenght or Moment /Unit area tool to distribute the moment would be more interesting?

Accounting for the element area (or element lenght in this case) on an arbitrary mesh by hand can be a nightmare.

6

I have finished the test and the result with a minimum refinement is excellent.

Reference solution

This is a test recommended by the National Agency for Finite Element Methods and Standards (U.K.): Test LE2 from NAFEMS publication TNSB, Rev. 3, “The Standard NAFEMS Benchmarks,” October 1990.
Stress: Outer surface tangential stress at point E is 60 MPa.

Stress is uniform across the shell but I had to compute and input the moments on each node by hand. I think that considering the coefficient of each node and participating area is the correct way to apply a Moment to a Set of Nodes. Such tool would be usefull.

imagen
imagen1012×668 115 KB

7

Such load would only apply to straight edges of the geometry. What happens if the edge is not straight?

8

That’s nice and enough in most cases.
We don’t have moment distribution on shells. It doesn’t work unless knots are induced.

I would say that is a similar limitation to the one exposed for *Distributing moments.

9

This strategy clearly doesn’t work for curved surfaces. Resultant moment far from the aplication point is correct. I have check it with section print, but the moment is not correctly translated into forces at the aplication point to provide a uniform stress distribution. The picture compares the expected forces distribution (left) witht the one using my proposal (right).

It is a moment about the x axis.

imagen
imagen840×510 49.6 KB

10

it’s common to assign node concentrated force and moment along an edge in local or global coordinates, units are in force/length e.g N/mm or N*mm/mm. When total of force need to distribute, length of an edge need to be known.

distribution of force at some section has different approach, required to know center of gravity of a section and nodal edge distance of lever arm. Rigid body or coupling usually in used for this task.