Triangular shell elements not calculating XX stress correctly?

[rkiser]

I have generated a simple test case of a flat bar in tension using (20) triangular shell elements.  A cross section of 2 x 1/2 inch with a 1000 lbf force at the end which should yield a 2000 psi uniform stress in the X direction, but it (FEAPpv 5.1c) does not appear to do so.  My entire input file and a snippet of the output file are below.

Am I doing something wrong here?  Any help would be much appreciated.

**** input file *************
FEAP * * 3-D Plate Problem
18  10 1  3  6  4
MATErial,1
SHELL
ELAStic ISOTropic 30.0e6 0.29
THIC,, 0.25
! Blank termination record
COORdinates
1  0 0.0 2.0 0.0
2  0 0.0 1.0 0.0
3  0 0.0 0.0 0.0
4  0 1.0 2.0 0.0
5  0 1.0 1.0 0.0
6  0 1.0 0.0 0.0
7  0 2.0 2.0 0.0
8  0 2.0 1.0 0.0
9  0 2.0 0.0 0.0
10 0 3.0 2.0 0.0
11 0 3.0 1.0 0.0
12 0 3.0 0.0 0.0
13 0 4.0 2.0 0.0
14 0 4.0 1.0 0.0
15 0 4.0 0.0 0.0
16 0 5.0 2.0 0.0
17 0 5.0 1.0 0.0
18 0 5.0 0.0 0.0
! Blank termination record
ELEMents
1  0  1   1  2  4  4
2  0  1   4  2  5  5
3  0  1   2  3  5  5
4  0  1   5  3  6  6
5  0  1   4  5  7  7
6  0  1   7  5  8  8
7  0  1   5  6  8  8
8  0  1   8  6  9  9
9  0  1   7  8 10 10
10 0  1  10  8 11 11
11 0  1  11  8  9  9
12 0  1  11  9 12 12
13 0  1  10 11 13 13
14 0  1  13 11 14 14
15 0  1  11 12 14 14
16 0  1  14 12 15 15
17 0  1  13 14 16 16
18 0  1  16 14 17 17
19 0  1  14 15 17 17
20 0  1  17 15 18 18
! Blank termination record
BOUNdary restraints
1  0  1 0 1 1 1 1
2  0  1 1 1 1 1 1
3  0  1 0 1 1 1 1
! Blank termination record
FORCe
16  0   250.0    0.0  0.0  0.0 0.0 0.0
17  0   500.0    0.0  0.0  0.0 0.0 0.0
18  0   250.0    0.0  0.0  0.0 0.0 0.0
! Blank termination record
END
BATCH
      TANG
      FORM
      SOLVE
      DISPLACEMENT,ALL
      STRESS,ALL
      REACTION,ALL
END
STOP

**** snippet of output file *******************
 FEAP * * 3-D Plate Problem

     S h e l l   S t r e s s e s

 elmt x-coord  xx-stress  yy-stress  xy-stress   1-stress   2-stress   angle
 matl y-coord  xx-moment  yy-moment  xy-moment   1-moment   2-moment   angle
      z-coord  xx-strain  yy-strain  xy-strain  xx-curvtr  yy-curvtr  xy-curvtr
               xx-sig(t)  yy-sig(t)  xy-sig(t)   1-sig(t)   2-sig(t)   angle
               xx-sig(b)  yy-sig(b)  xy-sig(b)   1-sig(b)   2-sig(b)   angle
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

    1   0.500  7.322E+01  4.268E+02  1.768E+02  5.000E+02  1.734E-12   67.50
    1   1.750  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E+00   45.00
        0.000 -6.739E-06  5.407E-05  6.081E-05  0.000E+00  0.000E+00  0.000E+00
               2.929E+02  1.707E+03  7.071E+02  2.000E+03  6.935E-12  6.750E+01   << shouldn't XX stress be 2000 psi and YY stress near zero?
               2.929E+02  1.707E+03  7.071E+02  2.000E+03  6.935E-12  6.750E+01

    2   0.750  2.500E+02  2.500E+02 -2.500E+02  5.000E+02  3.695E-13  -45.00
    1   1.250  0.000E+00  0.000E+00  0.000E+00  0.000E+00  0.000E+00   45.00
        0.000  2.367E-05  2.367E-05 -8.600E-05  0.000E+00  0.000E+00  0.000E+00
               1.000E+03  1.000E+03 -1.000E+03  2.000E+03  1.478E-12 -4.500E+01
               1.000E+03  1.000E+03 -1.000E+03  2.000E+03  1.478E-12 -4.500E+01     

[Prof. R.L. Taylor]

The problem is to determine a unique coordinate system  for the transformation from global to the local system used for stiffness. This is performed in tran3d and uses a set of axes related to the element edges, thus, the labels for stresses are misleading, they are the local axes and this means directions vary from element to element.  So as long as all the element edges are aligned the stresses may look good, but for triangles this is not true so stresses are not usable.

If anyone can provide a way to set up coordinates better changes can be made, however, for general or closed shells it is hard.