FEAP - - A Finite Element Analysis Program

Add to my manuals
77 Pages

advertisement

FEAP - - A Finite Element Analysis Program | Manualzz
FEAP - - A Finite Element Analysis Program
Version 8.1 Programmer Manual
Robert L. Taylor
Department of Civil and Environmental Engineering
University of California at Berkeley
Berkeley, California 94720-1710
E-Mail: [email protected]
January 2007
Contents
1 Introduction
1.1 Setting Program Options . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 Uses of Common and Include Statements . . . . . . . . . . . . . . . . .
1
1
3
2 Data Input and Output
2.1 Parameters and Expressions . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Array Outputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
5
7
3 Allocating Arrays
8
4 User Functions
4.1 Mesh Input Functions - UMESHn. . . . . .
4.1.1 Command line data . . . . . . . .
4.2 User Material Models . . . . . . . . . . .
4.2.1 The UMATIn Module . . . . . . .
4.2.2 The UMATLn Module . . . . . . .
4.2.3 Auto time step control . . . . . .
4.3 Mesh Manipulation Functions - UMANIn.
4.4 Solution Command Functions - UMACRn. .
4.5 Plot Command Functions - UPLOTn. . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5 Adding Elements
5.1 Material property storage . . . . . . . . . . . . . . .
5.2 Non-linear Transient Solution Forms . . . . . . . . .
5.3 Example: 2-Node Truss Element . . . . . . . . . . . .
5.3.1 Theory for a Truss . . . . . . . . . . . . . . .
5.4 Additional Options in Elements . . . . . . . . . . . .
5.4.1 Task 1 Options . . . . . . . . . . . . . . . . .
5.4.2 Task 6 Options . . . . . . . . . . . . . . . . .
5.5 Projection of element variables to nodes . . . . . . .
5.6 Elements with History Variables . . . . . . . . . . . .
5.6.1 Assigning amount of storage for each element
5.6.2 Accessing history data for each element . . . .
i
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
14
14
15
16
17
17
19
21
22
23
.
.
.
.
.
.
.
.
.
.
.
26
30
30
41
42
49
49
49
50
51
53
53
CONTENTS
5.7
5.8
ii
Energy Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A Non-linear Theory for a Truss . . . . . . . . . . . . . . . . . . . . . .
6 Utility routines
6.1 Numerical quadrature routines . . . . . . .
6.1.1 One dimensional quadrature . . . .
6.1.2 Two dimensional quadrature . . . .
6.1.3 Three dimensional quadrature . . .
6.2 Shape function subprograms . . . . . . . .
6.2.1 Shape functions in one-dimension .
6.2.2 Shape functions in two-dimensions
6.2.3 Shape functions in three-dimensions
6.3 Eigenvalues for 3 × 3 matrix . . . . . . . .
6.4 Plot routines . . . . . . . . . . . . . . . .
6.4.1 Mesh plots . . . . . . . . . . . . . .
6.4.2 Element data plots . . . . . . . . .
6.4.3 Other user plots . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
54
55
58
58
58
59
60
61
61
63
64
64
65
65
67
69
List of Figures
4.1 Sample UMATI1 module . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Sample UMATLn module . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
20
5.1
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
26
27
31
32
43
46
47
48
52
FEAP Element Subprogram. Part 1. . . . . . .
FEAP Element Subprogram. Part 2. . . . . . .
FEAP Element Common Blocks. . . . . . . . .
FEAP Element Common Blocks using Includes.
2-Node Truss Element . . . . . . . . . . . . . .
Element residual for two node truss . . . . . . .
Truss Tangent Matrix. Part 1 . . . . . . . . . .
Truss Tangent Matrix. Part 2 . . . . . . . . . .
Element variable projection routine . . . . . . .
iii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
List of Tables
3.1 Mesh Array Names, Numbers and Sizes . . . . . . . .
3.2 Solution Array Names, Numbers and Sized . . . . . .
3.3 Element Array Names, Numbers and Sizes . . . . . .
3.4 Element connection array IX use for element e . . . .
3.5 Element control array IE use for material number ma
.
.
.
.
.
11
11
11
12
12
4.1
Color Table for Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
5.10
5.11
Arguments of FEAP Element Subprogram. .
Task Options for FEAP Element Subprogram.
FEAP common block definitions. . . . . . . .
FEAP common block definitions. . . . . . . .
Material Parameters . . . . . . . . . . . . . .
Material Parameters . . . . . . . . . . . . . .
Material Parameters . . . . . . . . . . . . . .
Material Parameters . . . . . . . . . . . . . .
Tangent Parameters . . . . . . . . . . . . . .
Element Plot Definition Subprograms . . . . .
Momenta and Energy Assignments . . . . . .
6.1 Quadrature for triangles . .
6.2 Quadrature for tetrahedra .
6.3 Color pallet for FEAP plots
6.4 Values for control of plots .
.
.
.
.
.
.
.
.
.
.
.
.
iv
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
. .
. .
. .
. .
. .
. .
. .
. .
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
28
29
33
34
35
36
37
38
41
50
55
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
60
61
68
68
.
.
.
.
Chapter 1
INTRODUCTION
In this part of the FEAP manual some of the options to extend the capabilities of
the program are described. We begin by describing the utilities provided in FEAP for
use in data input. Options to add user commands for mesh and command language
extensions is then described and finally the method to add an element to the program
is described.
1.1
Setting Program Options
The size of problems which may be solved by FEAP depends on the amount of memory
available in the computer, as well as, solution options used. Memory for the main
arrays used to solve problems is dynamically allocated during the solution. Arrays are
allocated and deallocated using a system subprogram PALLOC or, for user developed
modules using subprogram UALLOC. Further information on use of these routines is
given in Section 3.
The IPR parameter in the FEAP81 module controls the specification of the size of REAL
variables. For typical UNIX and PC systems all real variables should be double precision and IPR is set to 2. For systems in which REAL*8 variables are single precision with
the same working length as integer variables the IPR parameter is set to 1. Any error
in setting this parameter may lead to incorrect behavior of the program, consequently,
do not reset the parameter to single precision unless a careful assessment of compiler
behavior for REAL*8 variables has been made.
By placing an alphanumeric version of each manual page in a separate file which has
the name of the command and a .t extender (e.g. coor.t for the mesh coordinate
input command) it is possible to read each page during execution using the HELP,name
command (where name is the command name whose manual page is to be read). For
this option to work properly it is necessary to define the path name to each manual
1
CHAPTER 1. INTRODUCTION
2
page in the FEAP81 module. For example:
file(1) = ’c:\Feap\Manual\Mesh/’
file(2) = ’c:\Feap\Manual\Macr/’
file(3) = ’c:\Feap\Manual\Plot/’
defines a typical path for a PC system. Each system requires a proper path definition.
FEAP will add the requested command name to each of the above paths to find mesh,
solution, or plot commands.
Normally FEAP reads each input data line as text data and checks each character for
the presence of parameters, expressions, and constants. For very large data sets this
parsing of each instruction can consume several seconds of compute time. If all data
is normally provided as numerical data, without use of any parameters or expressions,
the input time may be reduced by setting the value of the logical variable COFLG in
FEAP81 to false. FEAP will automatically switch to parsing mode if any record contains
non-numerical data item. It is also possible to use the PARSe and NOPArse commands
to set the appropriate mode of data input.
During the input of plot commands FEAP has the option to either set input options
automatically (DEFAult mode) or to read the values or range of contours to plot.
The default mode of operation may be assigned in the FEAP81 module by setting the
variables DEFALT and PROMPT. Setting DEFALT to true indicates that all default options
are to be set automatically. If DEFALT is set false, a prompt for contour intervals may
be requested by setting PROMPT to true.
FEAP has options to produce encapsulated PostScript output files in either gray scale
of in color. The default mode may be established by setting the variable PSCOLR and
PSREVS. Setting PSCOLR true indicates the PostScript files will be in color (unless set
otherwise by the plot COLOr data command. The PSREVS variable reverses the color
sequence.
Arrays in FEAP are dynamically allocated during execution. Thus, it is possible to
define and destroy arrays as well as to increase or decrease the size of an array. A parameter is provided to control when an array is to be decreased in size. The parameter
is INCRED and an array is decreased in size only when the new size is less than the old
size by the assigned value.
The last parameter which may be set in the FEAP81 module is the level for displaying
available commands when the HELP command is used while in mesh, solution, or plot
mode. FEAP contains a large number of commands which are not commonly used by
many users. To control the default number of commands displayed to users the commands have been separated into four levels: (0) Basic; (1) Intermediate; (2) Advanced;
and (3) Expert. The level to be displayed when using the HELP command is given may
be set in the integer variable HLPLEV. That is, setting:
CHAPTER 1. INTRODUCTION
hlplev = 1
3
! Intermediate
results in commands up to the intermediate level being displayed. It is possible to raise
or lower the level during execution using the command MANUal,,level where level is
the numerical value desired.
When developing program modules it is often desirable to have output of specific
quantities available (e.g. tracking the change in some parameters during successive
iterations. FEAP provides for a switch to make the outputs active or inactive during
an execution. The switch is named debug and placed in
integer
ndebug
logical
debug
common /debugs/ ndebug,debug
The value of the debug is set true by the solution command DEBUg and false by the
command DEBUg,OFF. Thus, placing code fragments into modules as
c
if(debug) then
write(iow,*) ’LABEL’,list ...
and/or
write( *,*) ’LABEL’,list ...
endif ! debug
! writes to output file
! writes to screen
This device supplements use of available debuggers on the computer.
1.2
Uses of Common and Include Statements
FEAP contains many COMMON statements which are used to pass parameters and small
array values between subprograms. For example, access to the debugging parameter
debug is facilitated through common /debugs/. Users may either place the common
statement (as well as data typing statements) directly in the routine or may use an
include statement. For debugging the statement would be
include
’debugs.h’
which during compilation would direct the precompiler to load the current common
statement from this file. In FEAP all include files have the same name as the common
with an added extender .h. The only exception is a dummy blank common which uses
the file name comblk.h which is defined as
real*8 hr
integer
mr
common hr(1),mr(1)
CHAPTER 1. INTRODUCTION
4
The dummy arrays hr(1) and mr(1) serve to pass all dynamically allocated arrays
between subprograms using a pointer array contained in the common np(*) and located
in the include file pointer.h. See Section 3 for more details on use of pointers. All
include files are located in the directory include.
It is highly recommended that users use include files rather than giving equivalent
common statements directly. If later releases of the FEAP program revise contents in
a common block, it will only be necessary to recompile the user routine rather than
change all the common statement definitions.
Chapter 2
DATA INPUT AND OUTPUT
FEAP includes utilities to perform input and to output small arrays of data. Users
are strongly encouraged to use the input utilities but often may wish to use their own
utilities to output data.
2.1
Parameters and Expressions
The subroutines PINPUT and TINPUT are input subprograms used by FEAP to input
each data record. They permit the data to be in a free form format with up to 255
characters on each record, as well as to employ expressions, parameters, and numerical
representations for each data item. These routines also should be used to input data in
any new program module developed. The PINPUT routine returns data to the calling
subprogram in a double precision array. The following statements may be included as
part of the routine performing the input.
subroutine xxx(.....)
logical errck, pinput
integer
ior,iow,ilg
common /iofile /ior,iow,ilg
real*8 td(5)
1
if(ior.lt.0) write(*,3000)
errck = pinput(td, 5)
if(errck) go to 1
The parameters defined in the common block are:
ior
- input
file unit number (if negative, input
5
CHAPTER 2. DATA INPUT AND OUTPUT
iow
ilg
6
from keyboard)
- output file unit number
- solution log file unit number
If an error occurs during input from the keyboard FEAP returns a value of true for the
function and a user may reinput the record if the implied loop shown above is used.
For inputs from a file, the program will stop and an error message indicating the type
of error occurring and the location in an input file is written to the output file.
The input routines return data in a real*8 array td(*). If any td(i) is to be used as an
integer or real*4 quantity, it must be cast to the correct type. That is, the following
operations should be used to properly cast the variable type:
real*4
real*8
integer
logical
t
td(5)
j
errck, pinput
errck = pinput (td, 5)
j = nint( td(1))
t = float(td(2))
! Integer assignment
! Real*4 assignment
PINPUT may be used to input up to 16 individual expressions on one input record
(each input record is, however, limited to 255 characters).
The routine TINPUT differs from PINPUT by permitting text data to also be input.
It is useful for writing user commands or to input data described by character arrays.
The routine is used as
logical
integer
character
real*8
errck, tinput
nt, nn
text(16)*16
td(16)
errck = tinput(text,nt,td,nn)
The parameter nt specifies the number of text values to input and the nn specifies the
number of real data values to input. The value for parameter nt or nn may be zero.
Thus the use of
errck = tinput(text,0,td,nn)
is equivalent to
errck = pinput(td,nn)
CHAPTER 2. DATA INPUT AND OUTPUT
7
Text variables may be converted to numerical (REAL*8) form using the subroutine call
call setval(text,nc,td)
where text is a string with nc characters and td a REAL*8 variable. The text string
can contain any parameters, expressions or numerical constants which evaluate to a
single value.
2.2
Array Outputs
Two subprograms exist to output arrays of integer and real (double precision) data.
The routine MPRINT is used to output real data and is accessed by the statement:
call mprint( array, nrow, ncol, ndim, label)
where array is the name of the array to print, nrow and ncol are the number of rows and
columns to output, ndim is the first dimension on the array, and label is a character
label which is added to the output. For example the statements:
real*8 aa(8,6)
. . .
call mprint( aa(2,4), 2, 3, 8, ’AA’)
outputs a 2 × 3 submatrix from the array aa starting with the entry aa(2,4). The
output entries will be ordered as the terms:
aa(2,4)
aa(3,4)
aa(2,5)
aa(3,5)
aa(2,6)
aa(3,6)
The MPRINT routine adds row and column labels as well as the character label.
The routine IPRINT is used to output integer data and is accessed by the statement:
call iprint( array, nrow, ncol, ndim, label)
where all parameters are identical to those for MPRINT except the array must be of type
integer.
Chapter 3
ALLOCATING ARRAYS
Dynamic data allocation is accomplished in FEAP by defining addresses in pointers
contained in the common block defined in pointer.h. This common block contains
pointers np for standard program arrays and up for user defined arrays and has the
form
integer
parameter
num_nps
, num_ups
(num_nps = 400 , num_ups = 200)
integer
np
common /pointer/ np(num_nps)
, up
, up(num_ups)
Each pointer is an offset relative to the address of a REAL*8 array hr(1) or an INTEGER
array mr(1) defined in a blank common
real*8
integer
common
hr
mr
hr(1),mr(1)
which is placed in the file comblk.h in the include/integer4 or include/integer8
directory. The pointers in the integer4 subdirectory have 32-bit lengths and should
be used for cases where addressing is within a 4 GByte range. The pointers in the
integer8 subdirectory have 64-bit lengths and should be used for larger index ranges.
The arrays ’hr’ and ’mr’ are used to establish addresses only and not to physically store
data. This mechanism permits references to elements in arrays which have positions
relative to hr or mr that may be after or before 1. Thus, FEAP must be compiled
without strict array bound checking. Size of problems is limited only by the available
memory in the computer used.
When using 64-bit pointers users must be careful to always define the address of an
array in a calling statement to also be 64-bits in length. For example use of
8
CHAPTER 3. ALLOCATING ARRAYS
9
integer
ioff
...
ioff = np(111) + numnp
call submat( hr(ioff), ...)
would cause an error since the pointer ioff is only 32 bits in length. To avoid this
problem it is necessary to either declare ioff to be 64-bits long as
integer*8 ioff
or use one of the FEAP include files p int.h (defining the integer type array fp(10))
or p point.h (defining the integer type scalar point). A type of correct length is
controlled at compile time by which subdirectory is used (i.e., integer4 or integer8).
Using this scheme permits direct reference to either real*8 or integer arrays in program modules without need to pass arrays through arguments of subprograms. A
subprogram PALLOC controls the allocation of all standard arrays in FEAP defined by
the np pointers and a subprogram UALLOC permits users to add allocation for their
own arrays defined by the pointers up. The basic use of the routines is provided by an
instruction
setvar = palloc(number,’NAME’,length,precision)
or
setvar = ualloc(number,’NAME’,length,precision)
where setvar, palloc and ualloc are logical types, number is an integer number of
the array, NAME is a 5 character name of the array, length is the number of words
of storage needed for the array, and precision is the type of array to allocate (1 for
integer and 2 for real*8 types). Upon initial assignment of any array its values
are set to zero. Thus, if the array is to be used only once it need not be set to zero
before accumulating additional values. If the array is to be reused or resized (see
below) it must be reinitialized prior to accumulating any additional values. Use of
these subprograms controls the assignment of memory space for all arrays such that no
conflicts occur between hr and mr referenced arrays. Each routine which makes direct
reference to an allocated array using a pointer (e.g., hr(np(43)) or mr(up(1))) must
contain include files as
include
include
’pointer.h’
’comblk.h’
CHAPTER 3. ALLOCATING ARRAYS
10
As an example for the use of the above allocation scheme consider a case where it is
desired to allocate a real (double precision array) with length NUMNP (number of nodes
in mesh) and an integer array with length NUMEL (number of elements in mesh). The
parameters NUMNP and NUMEL are contained in COMMON /CDATA/ and available using the
include file cdata.h. The new arrays aredefined using the temporary names TEMP1 and
TEMP2 which have numerical locations ‘111’ and ‘112’, respectively.1 The two arrays
are allocated using the statements
setvar = palloc( 111, ’TEMP1’, numnp, 2 )
setvar = palloc( 112, ’TEMP2’, numel, 1 )
where the last entry indicates whether the array is REAL*8 (2) or INTEGER (1). These
arrays are now available in any subprogram by specifying the pointer.h and comblk.h
include files and referencing the arrays using their pointers, e.g., in a subroutine call
as:
include ’pointer.h’
include ’comblk.h’
...
call subname ( hr(np(111)) , mr(np(112)) .... )
Note the use of hr(*) and mr(*) for the double precision and integer references,
respectively. Also, the use of the pointers avoids a need to include the array reference
until it is needed in a computation.
A short list of the mesh arrays available in FEAP is given in Table 3.1, for solution
arrays in Table 3.2, and for element arrays in Table 3.3. The names of all active arrays
in any analysis may be obtained using the SHOW,DICTionary solution command. Tables
3.4 and 3.5 describe the use of individual entries in the arrays IX and IE, respectively.
The subprograms PALLOC and UALLOC may also be used to destroy a previously defined
array. This is achieved when the length of the array is specified as zero (0). For
example, to destroy the arrays defined as TEMP1 and TEMP2 the statements
setvar = palloc( 111, ’TEMP1’, 0, 2 )
setvar = palloc( 112, ’TEMP2’, 0, 1 )
are given. Use of these statements results in the pointers np(111) and np(112) being
set to zero and the space used by the arrays being released for use by other allocations
at a later point in the program.
A call to PALLOC or UALLOC for any previously defined array but with a different nonzero length causes the size of the array to be either increased or decreased.
1
See the subprogram palloc.f in the program directory for the names and numbers of existing
arrays.
CHAPTER 3. ALLOCATING ARRAYS
NAME
ANG
D
F
ID
IE
IX
T
U
VEL
X
11
Num. dim 1
dim 2 dim 3
45 numnp
25
ndd nummat
27
ndf
numnp
2
31
ndf
numnp
2
32
nie nummat
33
nen1
numel
38 numnp
40
ndf
numnp
3
42
ndf
numnp
nt
43
ndm
numnp
-
Description
Angle
Material parameters
Force and Displacement
Equation nos.
Element control, dofs, etc.
Element connections
Temperature
Solution array
Solution rate array
Coordinates
Table 3.1: Mesh Array Names, Numbers and Sizes
NAME
CMASn
DAMPn
JPn
LMASn
TANGn
UTANn
Num.
n+8
n+16
n+20
n+12
n
n+4
dim 1 dim 2
compro
compro
neq
neq
maxpro
maxpro
-
dim 3
-
Description
Consistent Mass
Damping
Profile pointer
Lump Mass
Symmetric tangent
Unsymmetric tangent
Table 3.2: Solution Array Names, Numbers and Sized
NAME
ANGL
LD
P
S
TL
UL
XL
Num. dim 1 dim 2 dim 3
46
nen
35
nst
35
nst
36
nst
nst
39
nen
41
ndf
nen
6
44
ndm
nen
-
Description
Angle
Assembly nos.
Element vector
Element matrix
Temperature
Solution array
Coordinates
Table 3.3: Element Array Names, Numbers and Sizes
CHAPTER 3. ALLOCATING ARRAYS
NAME
IX( 1 ,e)
···
IX(nen ,e)
IX(nen+1,e)
IX(nen+2,e)
IX(nen+3,e)
IX(nen1 ,e)
IX(nen1-1,e)
IX(nen1-2,e)
IX(nen1-3,e)
Description
Global node 1
to
Global node nen
H1 history data pointer
H2 history data pointer
H3 history data pointer
Element material type number
Element region number (default = 0)
Active/deactive indicator
Active/deactive start
Table 3.4: Element connection array IX use for element e
NAME
IE(1,ma)
···
IE(ndf,ma)
IE(nie ,ma)
IE(nie-1,ma)
IE(nie-2,ma)
IE(nie-3,ma)
IE(nie-4,ma)
IE(nie-5,ma)
IE(nie-6,ma)
IE(nie-7,ma)
IE(nie-8,ma)
Description
Global DOF-1
to
Global DOF-ndf
Number history variables/element (NH1 and NH2)
Element material type number (ELMT01 = 1, etc.)
Element material type identifier (default = ma)
Offset to NH1/2 history variables (default = 0)
Offset to NH3 history variables (default = 0)
Number history variables/element (NH3)
Finite rotation update number (for UROTxx)
Get tangent from element if 0; if > 0 numerically differentiate residual to obtain tangent.
Equation number for element Lagrange multiplier
Table 3.5: Element control array IE use for material number ma
12
CHAPTER 3. ALLOCATING ARRAYS
13
For user defined arrays specified in UALLOC care should be exercised in selecting the
alphanumeric NAME parameter, which is limited to 5 characters, so that conflicts are not
created with existing names (use of the SHOW,DICT command is one way to investigate
names of arrays used in an analysis) or check the names already contained in the
subprogram PALLOC.
The subroutine PGETD also may be used to retrieve internal data arrays by NAME for use
in user developed modules. For example, if a development requires the nodal coordinate
data the call
integer xpoint, xlen, xpre
logical flag
....
call pgetd (’X ’,xpoint,xlen,xpre,flag)
will return the first word address in memory for the coordinates as xpoint, the length
of the array as xlen, and the precision of the array as xpre. If the retrieval is successful
flag is returned as true, whereas if the array is not found it is false. The precision
will be either one (1) or two (2) for INTEGER or double precision (REAL*8) quantities,
respectively. Thus, the above coordinate call will return xpre as 2 and xlen will be
the product of the space dimension of the mesh and the total number of nodes in the
mesh. The first coordinate, x1 , may be given as
x1 = hr(xpoint)
any other coordinates at nodes may also be recovered by a correct positioning in later
words of hr. For example y1 is located at hr(xpoint+1). The use of pgetd can lead
to errors for situations in which the length of arrays changes during execution, since in
these cases the value of the pointer xpoint can change. For such cases a call to pgetd
must be made prior to each reference involving xpoint. On the other hand, reference
using the pointers defined in arrays NP or UP are adjusted each time an array changes
size. However, users must ensure that a calling sequence is not sensitive to a change
in pointer. One way pointer changes can still lead to errors is through a program
call subname ( hr(np(111)), mr(np(112)), ....)
and then change the length of the array number ‘111’ or ‘112’ in the subroutine.
Chapter 4
USER FUNCTIONS
Users may add their own procedures to facilitate additional mesh input features, to perform transformations or manipulations on mesh data, to add new solution commands,
or to add new plot capabilities.
4.1
Mesh Input Functions - UMESHn.
To add a mesh input command a subprogram with the name UMESHn, where n has a
value between 0 and 9 must be written, compiled, and linked with the program. The
basic structure of the routine UMESH1 is:
subroutine umesh1(uprt)
c
User defined routine to input mesh data to FEAP
implicit
include
logical
c
none
’umac1.h’
uprt
! Contains UCT variable
Set name ’mes1’ to user defined
if(pcomp(uct,’mes1’,4)) then
uct = ’xxxx’
! Set user defined command name
else
c
User execution function statements follow
end if
end
14
CHAPTER 4. USER FUNCTIONS
15
The parameter UPRT is a logical parameter which is set to false when the NOPRint
mesh command is given and to true when the PRINt command is used (default is
true). The common block UMAC1 transfers the character variable UCT for the name of
the command. The default names are MESn where n is the same as the routine name
number. Assignment of a unique character name (which must not conflict with names
already assigned for mesh input commands) should be used to replace the xxxx shown.
When FEAP begins execution it scans all of the UMESHn routines and replaces the
command names mes1, etc., by the user furnished names. Thus, when the command
HELP is issued while in interactive MESH mode, the user name will appear in the
list instead of the default name (note, FEAP does not always display all available
commands. To see all commands issue the command MANUal,3 and then the HELP
command).
The ability to get array names as shown in Chapter 3 can be used to develop user
routines for input of coordinates, element connections, etc. With this facility it is
possible to develop an ability to directly input data prepared by other programs which
may be in a format which is not compatible with the requirements of standard FEAP
mesh commands.
4.1.1
Command line data
It is possible to include extra data on the command line for user functions. If the data
is not passed as an argument (e.g., as in the UMESHn functions) the data is recoverd
from a character string yyy*255 contained in the labeled common statement chdata.
The string is formatted into words with a field widths of 15 characters. Text words
are left justified and numerical values are right justified. The data must be recovered
in the function before any additional input statements are processed. For example if a
user input function has the command line:
GETData VALUes 35
is developed in the user function UMESH1 the first argument GETData is assigned to the
variable uct as the name of the function (see above), however, by default the other
two words are ignored. To recover their values the statement
include ’chdata.h’
character lct*15
real*8
ctl
integer
itl
is added to the user function and the items recovered in the else option of the function
using the statments:
CHAPTER 4. USER FUNCTIONS
16
lct = yyy(16:30)
call setval(yyy(31:45),15, ctl)
If FEAP parameters or expressions are used to define a numerical entry the subprogram
setval will make the appropriate evaluation. If necessary, the real value for ctl can
be cast into an integer using
itl = nint(ctl)
Note that the total number of added words must be 15 items or less (this is imposed
by the total of 16 items on any FEAP input record).
4.2
User Material Models
Users may add material models to elements by appending subprograms UMATIn and
UMATLn (where n have values from 0 to 9) to the FEAP system. The subprogram
UMATIn defines parameters used by the model and the subprogram UMATLn is called by
the element for each computation point (i.e., the quadrature point), receives the value
of a deformation measure as input and must return the value of stress and tangent
moduli as output.
To activate a user material model the input data for the mesh MATErial command
must include a statement with UCON as the first field. For example in a solid element
the command sequence can be
MATErial ma
SOLId
UCONstitutive xxxx v1 v2 ...
The role of the xxxx and vi data will be described in Section 4.2.1.
It is possible to use standard input parameters defined in Tables 5.5 to 5.8, as well
as by preceding the UCON command with a normal input sequence. For example, if
isotropic elastic properties are needed they may be included in the input sequence as
MATErial ma
SOLId
ELAStic ISOTropic e nu
UCONstitutive xxxx v1 v2 ...
No standard commands should follow the UCON command.
Alternatively, users may input elastic properties as part of their UMATIn module. If
the user routine does input additional data records (after the UCON record) and these
CHAPTER 4. USER FUNCTIONS
17
are terminated by a blank record, a second blank record will be needed to discontinue
material data input for this set. In all cases at least one blank record is always needed
to terminate the input of standard options for the material set.
4.2.1
The UMATIn Module
A sample module for a user constitutive model is shown in Fig. 4.1. As shown in this
figure, the UMATIn module has 5 arguments. The name of the constitutive equation
to be described is passed in the first parameter type. The second parameter passes
an array (vv(*)) which may be used to define up to 5 parameters for the material
model. The example shown above for the UCON includes the type data as xxxx and
the array vv(*) values as v1 v2 .... Users may also provide additional input within
the UMATIn module using the routines PINPUT or TINPUT described in Sect. 2.1. The
values of user parameters must be saved in the array ud(*) (the third argument of
UMATIn). In the current version there are 50 words of double precision values available
by default. Additional values may be allocated by assigning a larger value on the
control record (first record after the FEAP title record). Each material model is assigned
a user material number to the return parameter umat. This number must be a positive
integer. Finally, the number of history parameters to be assigned to each computation
(quadrature) point must be returned in the parameter n1. Currently, the parameter n3
may be set but is not available to the user material model. Thus, all history variables
must be retained in the n1 list. Use of history variables is described later as part of
the UMODEL module.
4.2.2
The UMATLn Module
A sample for the UMATL1 module is shown in Fig. 4.2. This subprogram will be called
by many of the elements included within FEAP if a user model has been specified
as part of the MATE mesh data (see previous subsection). The user model will not be
called for truss, frame, plate, and shell elements which use resultant models to describe
behavior. Also, any form which requires a one-dimensional model will not use a UMATLn
module. The module is designed to compute three-dimensional constitutive models in
which the stress and strain are stored as 6-component vectors and the tangent moduli
as a 6 × 6 matrix. Strains are passed to UMATLn in the argument array eps(6) and
stored in the order
T
= 11 22 33 γ12 γ23 γ31
where γij = 2 ij is the engineering shearing strain. Stress and moduli are to be
associated with the same ordering and returned in the argument arrays dimensioned
as sig(6) and dd(6,6), respectively. All values are to double precision (i.e., REAL*8).
When UMATLn is called the model n will be that which is defined in the module UMATIn.
Current values of the strains are, as mentioned above, passed in the array eps(6) and
CHAPTER 4. USER FUNCTIONS
18
subroutine umati1(type,vv, ud,n1,n3)
c-----[--.----+----.----+----.-------------------------------------]
c
Purpose: User material model interface
c
c
c
Inputs:
type
vv(*)
- Name of constitutive model (character variable)
- Parameters: user parameters from command line
c
Outputs:
c
n1
- Number history terms: nh1,nh2
c
n3
- Number history terms: nh3
c
ud(*)
- User material parameters
c-----[--.----+----.----+----.-------------------------------------]
implicit none
include ’iofile.h’
logical
pcomp
character type*15
integer
n1,n3
real*8
vv(5), ud(*)
c
Specify type of user model
if(pcomp(type,’mat1’,4)) then
type = ’E-1d’
c
c
! Specify new name for model
Input/output user data and save in ud(*) array
Set values of ’n1’ if required
n1
= ...
write(iow,*) ’ User Constitutive Inputs: E = ’,vv(1)
ud(1) = vv(1)
endif
end
Figure 4.1: Sample UMATI1 module
CHAPTER 4. USER FUNCTIONS
19
the trace of the strain in the parameter theta. Thus,
θ = ii = 11 + 22 + 33 .
In addition, if thermal problems are being solved the current value for the temperature
is passed as td. All material parameters for the current model are passed in the arrays
d(*) and ud(*). The array d(*) contains parameters assigned by standard FEAP
commands as described in Tables 5.5 to 5.8 and the array ud(*) contains values as
assigned in the user module UMATIn.
For constitutive equations with additional (internal) variables that evolve in time, users
must define entries for the h1(*) array. The number of entries available in the array
for each evaluation (i.e., each quadrature point) is nh. The value for nh is defined by
the parameter n1 in module UMATIn (see Fig. 4.1). Values from the previous time step
are passed back to the module in the array hn(*) (which also contains nh entries).
Users should never modify entries in the hn(*) array. Finally, the values of the
element operation switch is passed as the parameter isw (See Chapter 5 for operations
performed during different values of isw).
Using the above information users must compute values for the stress and the associated
tangent matrix. These are returned to the element in the arrays sig(6) and dd(6,6).
In addition, updates for any of the history parameters must be assigned in the array
h1(*) and returned to the element. Values of history variables returned are not used
for all values of isw (e.g., when reporting or projecting stresses under isw = 4 and isw
= 8 they are not saved). Values retained in the h1(*) array are copied to the hn(*)
array each time the command statement TIME is issued in a solution.
4.2.3
Auto time step control
The solution command:
AUTO MATErial rvalu(1) rvalu(2) rvalu(3)
initiates an attempt to control the solution process by a variable time stepping algorithm based on a user set value in the material constitution. The value to be set is
named rmeas which is passed between constitution and solution modules in the labeled
common
real*8
rmeas,rvalu
logical
aratfl
common /elauto/ rmeas,rvalu(3),aratfl
The three parameters may be used in defining an acceptable value for rmeas. The
algorithm coded monitors the solution during a standard iteration process set by, for
example:
CHAPTER 4. USER FUNCTIONS
20
subroutine umatl1(eps,theta,td,d,ud,hn,h1,nh,istrt,sig,dd,isw)
c-----[--.----+----.----+----.-------------------------------------]
c
Purpose: User Constitutive Model
c
c
c
c
c
c
c
c
c
c
c
c
Input:
eps(*)
c
c
c
Output:
sig(6) dd(6,6) -
theta
td
d(*)
ud(*)
hn(nh)
h1(nh)
nh
isw
-
Current strains at point
(small deformation)
Deformation gradient at point(finite deformation)
Trace of strain at point
Determinant of deformation gradient
Temperature change
Program material parameters
User material parameters
History terms at point: t_n
History terms at point: t_n+1
Number of history terms
Solution option from element
Stresses at point.
Current material tangent moduli
c-----[--.----+----.----+----.-------------------------------------]
implicit none
integer
real*8
real*8
c
umat,nh,isw, i
td
eps(*),theta(*),d(*),ud(*),hn(nh),h1(nh), sig(6),dd(6,6)
Dummy model:
sig = ud(1)*eps
do i = 1,6
dd(i,i) = ud(1)
sig(i) = ud(1)*eps(i)
end do
endif
end
Figure 4.2: Sample UMATLn module
CHAPTER 4. USER FUNCTIONS
21
LOOP,,n
TANG,,1
NEXT
If during any iteration up to n the value of rmeas exceeds a value of 2 (rmeas = 0 at
the start of the loop) a new value of ∆t is immediately set to
∆tnew = 0.85 ∆t/rmeas
and the iteration process is started over. On the other hand if convergence occurs
during the time step and the value of rmeas is smaller than 1.25, the time step is
adjusted according to
∆tnew = 1.50 ∆t
; rmeas ≤ 0.5
∆tnew = 1.25 ∆t
; 0.5 < rmeas ≤ 0.8
∆tnew = ∆t/rmeas ; 0.8 < rmeas
Finally, if convergence does not occur with in the n steps, then the time step is reset
according to
∆tnew = 0.85 ∆t/rmeas ; 1.25 < rmeas
∆tnew = ∆t/3
; otherwise.
After any of the above adjustments the value of rmeas is reset to zero (0).
An optimal value of rmeas is 1.25 – which leaves the step unchanged. The above
algorithm was proposed by Weber et al. [1].
4.3
Mesh Manipulation Functions - UMANIn.
The UMANIn modules, where n ranges from 0 to 9, may be used to perform transformations or manipulations on previously prescribed data. These commands appear
between the mesh input END command and the first INTEractive or BATCh solution command. To add a mesh manipulation command a subprogram with the name UMANIn,
where n has a value between 0 and 9 must be written, compiled, and linked with the
program. The basic structure of the routine UMANI1 is:
subroutine umani‘(prtu)
c
User defined routine to manipulate mesh data for FEAP
implicit
include
logical
none
’umac1.h’
prtu
! Contains UCT variable
CHAPTER 4. USER FUNCTIONS
c
22
Set name ’man1’ to user defined
if(pcomp(uct,’man1’,4)) then
uct = ’xxxx’
! Set user defined command name
else
c
User execution function statements follow
end if
end
The parameter PRTU is a logical parameter which is set to false when the NOPRint
mesh command is given and to true when the PRINt command is used (default is
true). The common block UMAC1 transfers the character variable UCT for the name of
the command. The default names are MANn where n is the same as the routine name
number. Assignment of a unique character name (which must not conflict with names
already assigned for mesh input commands) should be used to replace the xxxx shown.
After FEAP completes the input of mesh data it scans all of the UMANIn routines and
replaces the command names man1, etc., by the user furnished names.
The ability to get array names as shown in Chapter 3 can be used to develop user
routines for manipulation of the mesh data. For example, if a user has added the
specification of information by coordinates it may later be necessary to associate the
data with specific node numbers. This can be accomplished using a manipulation
command which searches for the node number whose coordinates are closest to the
specified location.
4.4
Solution Command Functions - UMACRn.
In a similar manner, users may add solution commands to the program by adding a
routine with the name UMACRn where n ranges from 0 to 9.
subroutine umacr0(lct,ctl,uprt)
c
User solution command function
implicit
include
none
’umac1.h’
logical
uprt
character lct*15
real*8
ctl(3)
! Contains the variable UCT
CHAPTER 4. USER FUNCTIONS
c
23
Set command word
if(pcomp(uct,’mac0’,4)) then
uct = ’xxxx’
else
c
User command statements are placed here
endif
end
The parameters LCT and CTL are used to pass the second word of a solution command
and the three parameter values read, respectively. Again the name xxxx should be
selected to not conflict with existing solution command names and will appear whenever
HELP is issued.
4.5
Plot Command Functions - UPLOTn.
In a similar manner, users may add new plot commands to the program by adding a
routine with the name UPLOTn where n ranges from 0 to 9.
subroutine uplot0(ctl,uprt)
c
User plot command function
implicit
c
none
include
’umac1.h’
logical
real*8
uprt
ctl(3)
! Contains the variable UCT
Set command word
if(pcomp(uct,’plt0’,4)) then
uct = ’xxxx’
else
c
User plot command statements are placed here
endif
end
The parameters CTL(3) are used to pass the three parameter values read, respectively.
Again the name xxxx should be selected to not conflict with existing plot command
CHAPTER 4. USER FUNCTIONS
24
names and will appear whenever HELP is issued.
Two plot utilities are available for placing lines on the screen. These are named DPLOT
and PLOTL. The calling form for DPLOT is given as
call dplot(s1,s2,ipen}
where s1, s2 are screen coordinates ranging from 0 to 1. Similarly, the calling sequence
for PLOTL is
call plotl(x1,x2,x3,ipen}
where x1, x2, x3 are coordinates values of the mesh. The value of ipen ranges from
1 to 3: 1 starts a filled panel; 2 draws a line from the current previous point to the
new point; 3 moves to the new point without drawing a line. If a filled panel is started
it must be closed by inserting the statement
call clpan()
Lines are drawn or panels filled in the current color. A color is set using the statement
call pppcol(color, switch)
where color is an integer defining the color number and switch should be zero. The
color values are given in Table 4.1.
CHAPTER 4. USER FUNCTIONS
Number
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
25
Color
Black
White
Red
Green
Blue
Yellow
Cyan
Magenta
Orange
Coral
Green-Yellow
Wheat
Royal Blue
Purple
Aquamarine
Violet-Red
Dark Slate Blue
Gray
Light Gray
Table 4.1: Color Table for Plots
Chapter 5
ADDING ELEMENTS
FEAP permits users to add their own element modules to the program by writing a
single subprogram called
subroutine elmtnn(d,ul,xl,ix,tl,s,r,ndf,ndm,nst,isw)
where nn may have values between 01 and 50. Each element subprogram must be
added before loading the FEAP library since dummy subprograms are included in the
library to avoid unsatisfied externals. The basic structure for an element routine is
shown in Figures 5.1 and 5.1.
Information is provided to the element subprogram through data passed as arguments
and data passed in common blocks. The data passed as arguments consists of eleven
subroutine elmtnn(d,ul,xl,ix,tl,s,r,ndf,ndm,nst,isw)
c
Prototype FEAP Element Routine:
nn = 01 to 50
implicit none
c
c
Common blocks: See Figure 5.2.
integer ndf,ndm,nst,isw
integer ix(nen1,1)
real*8 d(*),ul(ndf,*),xl(ndm,*),tl(*),s(nst,nst),r(nst)
if(isw.eq.0 .and. ior.lt.0) then
Output element description
write(*,*) ’
Elmt 1: Element description’
Figure 5.1: FEAP Element Subprogram. Part 1.
26
CHAPTER 5. ADDING ELEMENTS
27
c
c
c
c
elseif(isw.eq.1) then
Input/output of property data after command: ’mate’
d(*) stores information for each material set
Return: nh1 = number of nh1/nh2 words/element
Return: nh3 = number of nh3
words/element
c
elseif(isw.eq.2) then
Check element for errors.
c
c
c
c
c
c
elseif(isw.eq.3) then
Return: Element coefficient matrix and residual
s(nst,nst) element coefficient matrix
r(ndf,nen) element residual
hr(nh1)
history data base: previous time step
hr(nh2)
history data base: current time step
hr(nh3)
history data base: time independent
Negative jacobian, etc.
c
c
c
elseif(isw.eq.4) then
Output element quantities (e.g., stresses)
elseif(isw.eq.5) then
Return: Element mass matrix
s(nst,nst) consistent matrix
r(ndf,nen) diagonal matrix
c
c
elseif(isw.eq.6) then
Compute residual only
r(ndf,nen) element residual
c
c
c
elseif(isw.eq.7) then
Return: Surface loading for element
s(nst,nst) coefficient matrix
r(ndf,nst) nodal forces
c
c
c
c
elseif(isw.eq.8) then
Compute stress projections to
p(nen)
projection weight:
s(nen,*) projection values:
(default: project
endif
end
c
nodes (diagonal)
wt(nen)
st(nen,*)
8 quantities)
Figure 5.1: FEAP Element Subprogram. Part 2.
(11) items which are briefly described in Table 5.11 .
1
Note in Table 5.1 that FEAP transfers the values for most of the solution parameters in array
UL(NDF,NEN,*) at time tn+a , where a denotes a value between 0 and 1. The value of a is 1 (i.e.,
values are reported for time tn+1 ) unless generalized midpoint integration methods are used. For the
present we will assume a is 1.
CHAPTER 5. ADDING ELEMENTS
28
FEAP carries out tasks according to the value of the parameter ISW passed as the
eleventh parameter of the ELMTnn subprogram. A short description of the task carried
out by each value, as currently implemented, is shown in Table 5.2.
To use the basic features available in FEAP it is necessary to program tasks 0 to 6, 8,
and 10 shown above. If elements have local variables that need to be retained between
subsequent time steps history variables may be defined as described in Section 5.6. In
this case it is necessary to code task 12 and if any of the parameters have non-zero
initial values task 14 is used to set these values (zero values are set by default). Finally,
if special plotting options are desired it may be necessary to program task 20 (note
that contours for element variables such as stress, strain, etc. are developed from task
8).
It is not necessary to implement all other tasks in an element, however, for those
tasks that are not implemented it is important that the element routine not perform
any calculations. Thus if the form of the branch is programmed as an IF-THEN-ELSE
Parameter
d(*)
Description
Element data parameters
(Moduli, body loads, etc.)
ul(ndf,nen,j) Element nodal solution parameters
nen is number of nodes on an element (max)
(k)
j = 1: Displacement un+a
(k)
j = 2: Increment un+a − un
(k)
(k−1)
j = 3: Increment un+1 − un+1
(k)
j = 4: Rate vn+a
(k)
j = 5: Rate an+a
j = 6: Rate vn
xl(ndm,nen) Element nodal reference coordinates
ix(nen)
Element global node numbers
tl(nen)
Element nodal temperature values
s(nst,nst)
Element matrix (e.g., stiffness, mass)
r(ndf,nen)
Element vector (e.g., residual, mass)
may also be used as r(nst)
ndf
Number unknowns (max) per node
ndm
Space dimension of mesh
nst
Size of element arrays S and R
N.B. Normally nst = ndf*nen
isw
Task parameter to control computation
See prototype element in Figure 5.1
Table 5.1: Arguments of FEAP Element Subprogram.
CHAPTER 5. ADDING ELEMENTS
29
construct as shown in Fig. 5.1 then the ELSE should not carry out any operations
unless all options for ISW are programmed. Similarly if the element is programmed
using a SELECT-CASE form as
c
select case (isw)
case(1)
Input material parameters
...
case default
isw task
0
1
2
3
Description
Output label
Input d(*) parameters
Check elements
Compute tangent/residual
Store in S/r
4 Output element variables
5 Compute cons/lump mass
Store in S/r
6 Compute residual
7 Surface load/tangents
8 Nodal projections
9
10
11
12
13
14
15
16
17
18
19
20
23
25
26
Damping
Augmented Lagrangian update
Error estimator
History update
Energy/momentum
Initialize history
Body force
J integrals
Set after activation
Set after deactivation
NOT AVAILABLE: used in modal/base
Uses isw = 5 in element
Element plotting
Compute element loads only
Zienkiewicz-Zhu projection
Used to compute mesh boundary
Access Command
SHOW,ELEM
Mesh:MATE,n
Soln:CHECk
Soln:TANG
UTAN
Soln:STRE
Soln:MASS
MASS,LUMP
Soln:FORM,REAC
Plot:REAC
Mesh:SLOAd
Soln:STRE,NODE
Plot:STRE,PSTR
Soln:DAMP
Soln:AUGM
Soln:ERRO
Soln:TIME
Soln:TPLO,ENER
BATCh,INTEr
Mesh:BODY
Soln: JINT
Soln:ACTI
Soln:DEAC
BASE
Plot:PELM
ARCL
Soln:ZZHU
Called by default.
Table 5.2: Task Options for FEAP Element Subprogram.
CHAPTER 5. ADDING ELEMENTS
30
...
end select
the CASE DEFAULT should not perform any operations unless all options are programmed.
Finally, if the form
go to (1,2,..... ), isw
return
is used the RETURN statement should always be included as shown. This prevents any
unexpected execution of a statement that appears after the GO TO one.
Some of the options for additional data passed through common blocks is shown in
Figure 5.2 with each variable defined in Table 5.3. Also, in Figure 5.3 the reference
to common blocks using include statements is shown. In the prototype routine the
number of nodes on an element (nen) which is used to dimension ul is passed in the
labeled common /cdata/. Additional discussion is given below on use of some of the
other data passed through the common blocks.
5.1
Material property storage
The material parameters to be stored in the array D with pointer np(25) may be input
using the subprogram INPT2D. This subroutine is accessed by the statement:
CALL INPT2D(D,TDOF, NEV, TYPE)
where D is the array storing the material parameters; TDOF is returned as the parameter
to access temperature; NEV is the number of element history variables to allocate to
NH1; and TYPE is the element type. This routine inputs the commands as described in
the user manual and stores the data for each material set into the D array elements as
described in the following tables.
5.2
Non-linear Transient Solution Forms
Before describing the steps in developing an element we summarize first the basic
structure of the algorithms employed by FEAP to solve problems. Each problem to
be solved using an ELMTnn routine is established in a standard finite element form
as described in standard references (e.g., The Finite Element Method, 4th ed., by
O.C. Zienkiewicz and R.L. Taylor, McGraw-Hill, London, 1989 (vol 1), 1991 (vol 2)).
Here it is assumed this step leads to a set of non-linear ordinary differential equations
CHAPTER 5. ADDING ELEMENTS
character*4
common /bdata/
o,head
o,head(20)
integer
common /cdata/
numnp,numel,nummat,nen,neq,ipr
numnp,numel,nummat,nen,neq,ipr
31
integer
nstep,niter,naugm, titer,taugm,tform
common /counts/ nstep,niter,naugm, titer,taugm,tform
integer
iaugm, iform,intvc,iautl, nstepa
common /counts/ iaugm, iform,intvc,iautl, nstepa
real*8
dm
integer
n,ma,mct,iel,nel
common /eldata/ dm,n,ma,mct,iel,nel
real*8
tt
common /elplot/ tt(200)
real*8
bpr,
ctan,
psil
common /eltran/ bpr(3),ctan(3),psil
real*8
ut
common /eluser/ ut(200)
integer
common /hdata/
nh1,nh2,nh3,nlm
nh1,nh2,nh3,nlm
integer
ior,iow,ilg
common /iofile/ ior,iow,ilg
integer
nph
real*8
erav,j_int
,jshft
common /prstrs/ nph,ner,erav,j_int(3),jshft
integer
common /sdata/
ndf,ndm,nen1,nst,nneq,ndl,nnlm,nadd
ndf,ndm,nen1,nst,nneq,ndl,nnlm,nadd
real*8
common /tdata/
ttim,dt,c1,c2,c3,c4,c5, chi
ttim,dt,c1,c2,c3,c4,c5, chi
real*8
integer
common /
hr
mr
/ hr(1),mr(1000)
Figure 5.2: FEAP Element Common Blocks.
expressed in terms of nodal displacements, velocities, and accelerations given by ui (t),
u̇i (t), and üi (t), respectively. We denote the differential equation for node-i as the
CHAPTER 5. ADDING ELEMENTS
32
residual equation:
Ri (ui (t), u̇i (t), üi (t), t) = 0 .
To solve for the nodal displacements,velocities and accelerations it is necessary to
introduce an algorithm to integrate the nodal quantities in time, specify a constitutive
relation, and develop an algorithm to solve a (possibly) non-linear problem.
In FEAP, the integration method for nodal quantities is taken as a one step algorithm
with each quantity defined only at discrete times tn . Accordingly, we have displacements ui (tn ) with velocities and accelerations denoted as
u̇i (tn ) ≈ vi (tn )
and
üi (tn ) ≈ ai (tn )
A typical example for an integration algorithm for these discrete quantities is Newmark’s method where
ui ( tn+1 ) = ui ( tn ) + ∆t vi ( tn ) + ∆t2 [(
1
− β ) ai ( tn ) + β ai ( tn+1 )]
2
and
vi ( tn+1 ) = vi ( tn ) + ∆t [( 1 − γ ) ai ( tn ) + γ ai ( tn+1 )]
with u, v, and a being the set of displacements, velocities, and accelerations at node-i,
respectively.
A Newton method is commonly adopted to solve a non-linear (or linear) problem. To
implement a Newton method it is necessary to linearize the residual equation. For
FEAP, the Newton equation may be written as
(k+1)
Ri
include
include
include
include
include
include
include
include
include
include
include
include
(k)
= Ri
+
∂Ri (k)
(k)
| dαj = 0
∂αj
’bdata.h’
’cdata.h’
’counts.h’
’eldata.h’
’elplot.h’
’eltran.h’
’hdata.h’
’iofile.h’
’prstrs.h’
’tdata.h’
’pointer.h’
’comblk.h’
Figure 5.3: FEAP Element Common Blocks using Includes.
CHAPTER 5. ADDING ELEMENTS
Variable
o
head
numnp
numel
nummat
nen
neq
ipr
nstep
niter
naugm
titer
taubm
iaugm
iform
dm
n
ma
mct
iel
nel
tt
bpr
ctan
ut
33
Definition
Page eject option
Title record
Number of mesh nodes
Number of mesh elements
Number of material sets
Maximum nodes/element
Number active equations
Real variable precision
Total number of time steps
Number of iterations current step
Number of augments current step
Total iterations
Total augments
Augmenting counter
Number residuals in line search
Element proportional load
Current element number
Current element material set
Print counter
User element number
Number nodes on current element
Element stress values for TPLOt
Principal stretch
Element multipliers
Element user values for TPLOt
Table 5.3: FEAP common block definitions.
where αj is one of the variables at time tn+1 (e.g., uj ( tn+1 )). We define
(k)
Sij
and solve
(k)
= −
(k)
Sij dαj
∂Ri (k)
|
∂αj
(k)
= Ri
.
The solution is updated using
(k+1)
αj
(k)
= αj
(k)
+ dαj
.
In the above (k) is the iteration number for the Newton algorithm. To start the solution
for each step, FEAP sets
(0)
αj ( tn+1 ) = αj ( tn )
CHAPTER 5. ADDING ELEMENTS
Variable
nh1
nh2
nh3
ior
iow
nph
ner
erav
j-int
ndf
ndm
nen1
nst
nneq
ttim
dt
ci
hr
mr
34
Definition
Pointer to tn history data
Pointer to tn+1 history data
Pointer to element history
Current input logical unit
Current output logical unit
Pointer to global projection arrays
Pointer to global error indicator
Element error value
J integral values
Maximum dof/node
Mesh space dimension
Dimension 1 on IX array
Size of element matrix
Total dof in problem
Current time
Current time increment
Integration parameters
Real array data
Integer array data
Table 5.4: FEAP common block definitions.
where a quantity without the (k) superscript represents a converged value. For a linear
problem, Newton’s method converges in one iteration. Computing the residual after
one iteration must yield a zero value to within the roundoff of the computer used.
For non-linear problems, a properly implemented Newton’s method must exhibit a
quadratic asymptotic rate of convergence. Failure of the above performance for linear
and non-linear cases implies a programming error in an implementation or lack of a
consistently linearized algorithm (i.e., Sij is not an exact derivative of the residual).
In a non-linear problem, Newmark’s method may be parameterized in terms of increments of displacement, velocity, or acceleration. From the Newmark formulas, the
relations
dui = β ∆t2 dai
and
dvi = γ ∆t dai
define the relationships between the increments. Note that only scalar multipliers
involving β, γ, and ∆t are involved between the different measures.
The tangent matrix for the transient problem using Newmark’s method may be expressed in terms of the incremental displacement, velocity, or acceleration. As an
CHAPTER 5. ADDING ELEMENTS
Parameter
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Name
E
ν
α
ρ
a
q
T0
κ
b1
b2
b3
h
nh1
stype
Description
Young’s modulus
Poisson ratio
Thermal expansion coefficient
Mass density
Quadrature order for arrays
Quadrature order for outputs
Mass interpolator (a = 0: Diagonal; a = 1: Consistent
Loading intensity (plates/shells)
Stress free reference temperature
Shear factor (plates/shells/beams)
Body force/volume in 1-directions
Body force/volume in 2-directions
Body force/volume in 3-directions
Thickness (plates/shells)
History variable counter
Two dimensional type: 1 - plane stress; 2 - plane strain;
3 - axisymmetric2
17 etype Element formulation: 1 - displ; 2 - mixed; 3 - enhanced
18 dtype Deformation type: <: finite; > small
19 tdof
Thermal degree-of-freedom
20 imat
Non-linear elastic material type
21 d11
Material moduli
22 d22
Material moduli
23 d33
Material moduli
24 d12
Material moduli
25 d23
Material moduli
26 d31
Material moduli
27 g12
Material moduli
28 g23
Material moduli
29 g31
Material moduli
30 htype Heat flag
Table 5.5: Material Parameters
35
CHAPTER 5. ADDING ELEMENTS
Parameter
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Name
ψ
A
I11
I22
I12
J
κ1
κ2
Y0
Y∞
β
Hiso
Hkin
β1
β2
β3
ν1
τ1
ν2
τ2
ν3
τ3
nvis
k
Description
Orthotropic angle x1 principal axis 1
Area cross section (beam/truss)
Inertia cross section (beam/truss)
Inertia cross section (beam/truss)
Inertia cross section (beam/truss)
Polar inertia cross section (beam/truss)
Shear factor plate
Shear factor plate
Non-linear flag (beam/truss)
Inelastic material model type
Initial yield stress (Mises)
Final yield stress (Mises)
Exponential hardening rate
Isotropic hardening modulus (linear)
Kinematic hardening modulus (linear)
Yield flag
Orthotropic thermal stress
Orthotropic thermal stress
Orthotropic thermal stress
Error estimator parameter
Viscoelastic shear parameter
Viscoelastic relaxation time
Viscoelastic shear parameter
Viscoelastic relaxation time
Viscoelastic shear parameter
Viscoelastic relaxation time
Number of viscoelastic terms (1-3)
Damage limit
Damage rate
Penalty parameter
Table 5.6: Material Parameters
36
CHAPTER 5. ADDING ELEMENTS
Parameter
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
Name
K1
K2
K3
c
ω
Q
g1
g2
g3
p1
p2
p3
a0
a1
-
d31
d32
α3
Description
Fourier thermal conductivity
Fourier thermal conductivity
Fourier thermal conductivity
Fourier specific heat
Angular velocity
Body heat
Heat constitution added indicator
Follower loading indicator
Rotational mass factor
Damping factor
Ground acceleration factor
Ground acceleration factor
Ground acceleration factor
Ground acceleration proportional load number
Ground acceleration proportional load number
Ground acceleration proportional load number
Rayleigh damping mass ratio
Rayleigh damping stiffness ratio
Plate/Shell/Rod shear activation flag
Method: Type 1
Method: Type 2
Truss/Rod quadrature number
Axial loading value
Constitutive start indicator
Polar angle indicator
Polar angle coord 1
Polar angle coord 2
Polar angle coord 3
Constitution transient type
Plane stress recovery
Plane stress recovery
Plane stress recovery
Table 5.7: Material Parameters
37
CHAPTER 5. ADDING ELEMENTS
Parameter
93
94
95
96
97
98
99
100
101-126
127
128
129
130
131-148
149
150
151-159
160
161-166
167
170
171
172
173
174
175
176
177
178
179
180-181
182
183
184
185
186-200
Name
sref
y1
y2
lref
n1
n2
n3
nseg
σij
C
a1
a2
a3
a4
a5
a6
a7
a8
a9
βm
c
-
Description
Shear center type
Shear center coordinate
Shear center coordinate
Reference vector type
Reference vector parameter
Reference vector parameter
Reference vector parameter
Cross section shape type: 1 = rectangles; 2 = tube;
3 = Wide flange; 4 = Channel; 5 = Angle; 5 = Circle
Shape data
Surface convection (h)
Free-stream temperature (T∞ )
Reference absolute temperature
Number of hardening segments
Segment data sets ep Yiso Hkin
Total variables on frame section
Piezoelectric flag
Piezoelectric data
Initial stress flag
Initial stresses (constant)
Tension/compression only indicator
Fung pseudo elastic model modulus
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Fung model energy parameter
Viscoplastic rate parameters
Nodal quadrature parameters
ML − MC mass scaling factor
Estimate on maximum wave speed
Augmentation switch
Unused
Table 5.8: Material Parameters
38
CHAPTER 5. ADDING ELEMENTS
39
example, consider the case where the solution is parameterized in terms of increments
of the displacements (i.e., αj is the displacement vector uj ). For this case, the tangent
matrix is (we do not show dependence on the iteration (k) for simplicity of notation)
Sij duj
= −
∂Ri
∂Ri ∂vk
∂Ri ∂ak
duj −
duj −
duj .
∂uj
∂vk ∂uj
∂ak ∂uj
Note that from the Newmark formulas
∂ak
∂uj
1
δ kj
β ∆t2
=
;
∂vk
∂uj
=
∂vk ∂al
∂al ∂uj
=
γ
δ kj
β ∆t
in which δ kj is the Kronnecker delta identity matrix for the k,j nodal pair . From the
residual we observe that
Kij
= −
∂Ri
∂uj
;
Cij
= −
∂Ri
∂vj
;
Mij
= −
∂Ri
∂aj
define the tangent stiffness, damping, and mass, respectively. Thus, for the Newmark
algorithm the total tangent matrix in terms of the incremental displacements is
Sij
= Kij +
γ
1
Cij +
Mij .
β ∆t
β ∆t2
For other choices of increments, the tangent may be written in the general form
Sij
= c1 Kij + c2 Cij + c3 Mij
where the ci are scalar quantities involving the integration parameters of the method
selected and ∆t. Thus, any one step integrator may be considered and will affect only
the specification of the constants in the tangent.
In FEAP the element tangent matrix, Sij , is stored as a two dimensional array which
is dimensioned as s(nst,nst), where nst is the product of ndf and nen, with ndf
the maximum number of degree-of-freedoms at any node in the problem and nen the
maximum number of nodes on any element. The ordering of the unknowns into nst
must be carefully aligned in order for FEAP to properly assemble each element matrix
into the global tangent. The ordering is such that sub-matrices are defined for each
node attached to the element. Thus


S11 S12 S13 ··
S21 S22 S23 ··

S = 
S31 S32 S33 ··
··
··
·· ··
where Sij is the sub-matrix for nodal pairs i, j. Each of the sub-matrices is a square
matrix of the size of the maximum number of degree-of-freedoms in the problem which
CHAPTER 5. ADDING ELEMENTS
40
is passed to the subprogram as ndf. Thus,

 ij
ij
ij
S11 S12
S13
··
ij
ij
ij
S21
·· 
S23
S22

Sij = 
ij
ij
ij
S31 S32 S33
·· 
ij
··
··
·· Sndf,ndf
ij
in which Sab
is an array coefficient for nodal pair i, j for the degree-of-freedom pair a, b.
In FEAP, the element residual may be stored as one dimensional array which is dimensioned r(nst) with entries stored in the same order as the rows of the element tangent
matrix or as a two dimensional array which is dimensioned as r(ndf,nen). The one
dimensional form of the residual is given as
 
R1
R2 
 
R = R 
 3
..
.
where the entries in each submatrix are given as
 i 
R1
 Ri 
 2 
 i 
Ri =  R 3  .
 .. 
 . 
i
Rndf
The two dimensional form places the entries Ri as columns. Accordingly,
R = R1 R2 R3 · · · .
The two forms for defining the residual r are equivalent based on the Fortran ordering
of information into double subscript arrays.
If ndf is larger than needed for the element and residual the unused positions need not
be defined (the tangent array s and the residual r are set to zero before each element
routine is called).
The arrays xl(i,j), ul(i,j,1), ul(i,j,4) and ul(i,j,5) (described in Table 5.1)
are used to obtain the nodal coordinates, displacements,velocities and accelerations,
respectively. When FEAP solves a problem without transient loading (e.g., inertial
loading as mass times acceleration) the velocities and accelerations are set to zero
prior to calling the element subroutine. Consequently, in programming the steps to
compute the residual r the inertia terms have no effect for static or quasi-static problems and may be included (generally there are very few additional operations involved
CHAPTER 5. ADDING ELEMENTS
41
Parameter Description
ctan(1)
c1 : Multiplier of s matrix for ul(i,j,1) terms
(e.g., stiffness matrix multiplier)
ctan(2)
c2 : Multiplier of s matrix for ul(i,j,4) terms
(e.g., damping matrix multiplier)
ctan(3)
c3 : Multiplier of s matrix for ul(i,j,5) terms
(e.g., mass matrix multiplier)
Table 5.9: Tangent Parameters
to add these terms). The programming of the tangent array, however, must distinguish
between cases in which transient (e.g., inertial) loads are present and those in which
they are omitted. The different cases are implemented in FEAP by making appropriate assignments to the ci parameters. To facilitate the programming of the tangent
array returned in s for the various cases, a parameter array ctan(3) is passed to the
subprogram in labeled common eltran. When the task parameter isw is 3, the values
in the ctan array are interpreted according to Table 5.9.
Thus, in solid mechanics applications the tangent matrix is defined in an element
routine as
S = ctan(1) K + ctan(2) C + ctan(3) M
where K is the stiffness matrix, C is the damping matrix, and M is the mass matrix.
For non-linear applications these matrices normally are computed with respect to the
current values of the available solution parameters. The values provided in the ctan
array are set by FEAP according to the active transient solution option. For a static
option both ctan(2) and ctan(3) are zero. For options integrating first order differential equations in time only ctan(3) will be zero. For options integrating second order
differential equations in time all the parameters are non-zero.
5.3
Example: 2-Node Truss Element
An element routine carries out tasks according to the value assigned to the parameter
isw as indicated in Table 5.2 To describe basic steps to program the various tasks
defined by isw, we consider next the problem of a 2-node, linear elastic truss element
for small deformation applications. The element is described in sufficient generality to
permit solution of both two and three dimensional truss problems.
CHAPTER 5. ADDING ELEMENTS
5.3.1
42
Theory for a Truss
The governing equations for a typical truss member element, shown in Figure 5.4, are
the balance of momentum equation:
∂(Aσss )
+ A bs = ρ A üs
∂s
the strain-displacement equation for small deformations:
∂us
∂s
and a constitutive equation. For example, considering a linear elastic material the
constitutive equation may be written as
ss =
σss = E ss .
Boundary and initial conditions must also be specified to obtain a well posed problem;
however, our emphasis here is the derivation of the element arrays associated with the
above differential equations. In the above:
• s is the coordinate along the truss member axis,
• bs is a loading in direction s per unit length,
• A is the truss cross-section area,
• ρ is the mass density per unit volume,
• us is a displacement in direction s,
• v̇s is an acceleration in direction s (v = u̇),
• ss is a strain along the truss member axis, and
• σss is the stress on a truss cross section.
The equations may also be deduced from the variational equation
Z
δΠ =
δss σss A ds +
L
d Z
X
i=1
L
δui ρ A v̇i ds −
d Z
X
i=1
δui bi ds + δΠext
L
where δΠext contains the boundary and loading terms not associated with an element.
Where, in addition to previously defined quantities, we define:
CHAPTER 5. ADDING ELEMENTS
43
2
u2
2
u
2
u3
2
1
1
u2
x2
u
1
1
3
u
1
1
x1
x3
Figure 5.4: 2-Node Truss Element
• d is the spatial dimension of the truss (1, 2, or 3),
• xi are the Cartesian coordinates in the d directions.
• L is the length of the truss member,
• δui is a virtual displacement in direction xi ,
• v̇i is an acceleration in direction xi (v = u̇),
• bi is a loading in direction xi per unit length, and
• δss is a virtual strain along the truss axis.
For a straight truss member the displacement along the axis, us may be expressed in
terms of the components in the directions xi as
us = l · u( s , t ) =
d
X
li ui ( s , t )
i=1
where t is time, u is the displacement vector with components ui , l is a unit vector
along the axis of the member with direction cosines li defined by
li =
2
L =
∂xi
xi2 − xi1
=
∂s
L
d
X
i=1
( xi2 − xi1 )2
CHAPTER 5. ADDING ELEMENTS
44
and xi1 , xi2 are the coordinates of nodes 1 and 2, respectively. The displacement
components are interpolated on the 2-node truss member as
ui ( s , t ) = ( 1 − ξ ) ui1 ( t ) + ξ ui2 ( t ) ; ξ =
s
L
in which ui1 , ui2 are the displacements at nodes 1 and 2. The virtual displacements
are obtained from the above by replacing ui by δui , etc. The truss strain is
d
ss =
∂us X ∂ui
=
.
li
∂s
∂s
i=1
Using the interpolations for the displacement components yields
d
1 X
ss = 2
∆xi ∆ui
L i=1
where
∆xi = xi2 − xi1 = li L
and
∆ui = ui2 − ui1 .
Thus, in matrix form the strain is
d
ui1
1 X
− ∆xi ∆xi
ss = 2
ui2
L i=1
Using the above displacement interpolations, the variational equation for the truss may
be expressed in matrix form as
Z
Z 1 −∆xi
ü
1−ξ
δΠ = δui1 δui2
σss Ads +
ρA 1 − ξ ξ ds i1
2
∆xi
ξ
üi2
L
L
L
Z 1−ξ
−
bi ds + δΠext .
ξ
L
FEAP constructs the finite element arrays from the element residuals which are obtained from the negative of the terms multiplying the nodal displacements. Accordingly,
Z Ri1
1−ξ
Ri =
=
bi ds
ξ
Ri2
L
Z Z
1 − ∆xi
1−ξ
ü
σss A ds −
ρ A 1 − ξ ξ ds i1
−
2
∆x
ξ
ü
i
i2
L L
L
CHAPTER 5. ADDING ELEMENTS
45
is the residual for the i-coordinate direction. For constant properties and loading over
an element length (note that for this case the stress will also be constant since strains
are constant on the element), the above may be integrated to yield
1
σss A − ∆xi
ρ A L 2 1 üi1
Ri1
1
Ri =
= bi L
−
−
.
Ri2
1
∆xi
1 2 üi2
2
L
6
(5.1)
For the present we assume the material model is a linear elastic in which the stress is
related to strain through
σss = E ss
where E is the Young’s modulus.
Based on a linear elastic material, the term in the residual involving σss may be written
as
d
uj1
σss A − ∆xi
E A − ∆xi X − ∆xj ∆xj
.
=
uj2
∆xi
∆xi
L
L3
j=1
For the linear elastic material, a stiffness matrix may be expressed as
E A − ∆xi kij −kij
− ∆xj ∆xj =
Kij =
∆xi
−kij kij
L3
where
EA
∆xi ∆xj .
L3
The residual may now be written using a stiffness and mass matrix as
kij
=
d X
1
1
kij −kij uj1
m11 m12 üi1
Ri1
−
−
Ri =
= bi L
1
−kij kij
uj2
m21 m22 üi2
Ri2
2
j=1
(5.2)
with
ρAL
ρAL
; m12 = m21 =
.
3
6
For non-linear material behavior the residual must be computed using Equation 5.1
with the stress replaced by the value computed from the constitutive equation.
m11 = m22 =
The integration method for nodal quantities is taken as Newmark’s method described
in Section 5.2. The residual and tangent matrix for a Newton type method are now
available and may be inserted into R and S after noting that for the truss that the
damping matrix C is zero. The residual may be programmed directly from Equation
5.1 and an implementation using the two dimensional form r(ndf,nen) is shown in
Figure 5.5.
Similarly, using the results from Section 5.2, the tangent matrix for the truss may be
programmed as indicated in Figures 5.6 and 5.7.
CHAPTER 5. ADDING ELEMENTS
46
if(isw.eq.3 .or. isw.eq.6) then
c
Compute element length
L2= 0.0d0
do i = 1,ndm
L2 = L2 + (xl(i,2) - xl(i,1))**2
end do
L = sqrt(L2)
c
Compute strain-displacement matrix
Lr = 1.d0/L2
eps = 0.0d0
do i = 1,ndm
bb(i,1) = -(xl(i,2) - xl(i,1))*Lr
bb(i,2) = -bb(i,1)
eps
= eps + bb(i,2)*(ul(i,2,1) - ul(i,1,1))
end do
c
Compute mass terms
cmd = rhoA*L/3.0d0
cmo = cmd*0.5d0
c
Form body/inertia force vector (dm = prop. ld.)
&
&
sigA = EA*eps*L
body = 0.5d0*L*dm
do i = 1,ndm
r(i,1) = body*d(6+i)
- cmd*ul(i,1,5)
r(i,2) = body*d(6+i)
- cmo*ul(i,1,5)
end do
- bb(i,1)*sigA
- cmo*ul(i,2,5)
- bb(i,2)*sigA
- cmd*ul(i,2,5)
Figure 5.5: Element residual for two node truss
CHAPTER 5. ADDING ELEMENTS
if(isw.eq.3) then
c
Compute element length
L2= 0.0d0
do i = 1,ndm
L2 = L2 + (xl(i,2) - xl(i,1))**2
end do
L = sqrt(L2)
c
Form stiffness multiplier
dd = ctan(1)*EA*L
c
Compute strain-displacement matrix
Lr = 1.d0/L2
do i = 1,ndm
bb(i,1) = -(xl(i,2) - xl(i,1))*Lr
bb(i,2) = -bb(i,1)
db(i,1) = dd*bb(i,1)
db(i,2) = -db(i,1)
end do
Figure 5.6: Truss Tangent Matrix. Part 1
47
CHAPTER 5. ADDING ELEMENTS
c
Compute stiffness terms (N.B. ndm < or = ndf)
i1 = 0
do ii = 1,2
j1 = 0
do jj = 1,2
do i = 1,ndm
do j = 1,ndm
s(i+i1,j+j1) = db(i,ii)*bb(j,jj)
end do
end do
j1 = j1 + ndf
end do
i1 = i1 + ndf
end do
c
Compute mass terms and correct for inertial effects
cmd = ctan(3)*rhoA*L/3.0d0
cmo = cmd*0.5d0
do i = 1,ndm
j
= i + ndf
s(i,i) = s(i,i) + cmd
s(i,j) = s(i,j) + cmo
s(j,i) = s(j,i) + cmo
s(j,j) = s(j,j) + cmd
end do
endif
Figure 5.7: Truss Tangent Matrix. Part 2
48
CHAPTER 5. ADDING ELEMENTS
5.4
49
Additional Options in Elements
FEAP permits some additional options to be included within element tasks.
5.4.1
Task 1 Options
Often it is necessary to use several element types to perform an analysis. For example it
may be necessary to use both truss and frame (bending resistant) elements to perform
an analysis. As developed in Section 5.3, the truss element has one degree of freedom
for each spatial dimension, whereas, the frame element must have additional unknowns
to represent the bending behavior. For nodes connected only to truss elements it is not
necessary to have the additional degrees-of-freedom active and a user would be required
to specify restraint conditions for these nodes and degrees-of-freedom. By inserting the
following lines of code into the truss element routine for the isw = 1 task FEAP will
automatically eliminate any unneeded degrees-of-freedom.
do i = ndm+1,ndf
ix(i) = 0
end do ! i
Note that for isw = 1 the ix parameter is not used to pass the nodal connection array
but is used to return the list of unused degrees-of-freedom.
Utility routines are also provided to assist users in providing the necessary list of nodes
needed to properly draw the mesh each element type during plot outputs. The names
of the routines are listed in Table 5.10 and each routine is called as
call plname (iel)
where iel is an integer parameter defined in common eldata. If no call to a subprogram is included each element is assumed to be a 4 to 9 node quadrilateral and default
drawing order will be assigned. Users may construct their own drawing order also by
following the steps employed in one of the routines defined in Table 5.10.
5.4.2
Task 6 Options
The TPLOt solution command includes an option to save specific element quantities
(e.g., stress,strain, etc.). This option is implemented for user elements by including the
common
CHAPTER 5. ADDING ELEMENTS
Routine Name
PLTLN2
PLTRI3
PLQUD4
PLTRI6
PLTET4
PLBRK8
50
Description
2-node line element
3-node triangular element
4-node quadrilateral element
6-node triangular element
4-node tetrahedron element
8-node brick element
Table 5.10: Element Plot Definition Subprograms
real*8
tt
common /elplot/ tt(100)
and then inserting the statement
tt(i) = value
at an appropriate location in the isw = 3 task.
For example if it were desired to save the force and strain in the truss element the
statements
tt(1) = EA*eps
tt(2) = eps
! Element axial force
! Element axial strain
could be placed anywhere after the stress and strain are defined. These values would
be output by using a solution command sequence such as
batch
tplot
end
stress,nn,1
stress,nn,2
show
5.5
! saves force for element nn
! saves strain for element nn
! writes tplot items to output file
Projection of element variables to nodes
The STREss,NODE solution command and the PLOT,STRE command require a projection
of element variables to nodes. A continuous stress field is assumed to obtain the nodal
values. Accordingly,
σ = Nα σα
CHAPTER 5. ADDING ELEMENTS
51
where σ is any value which is to be projected to nodes (e.g., a stress or strain), Nα are
shape functions for the element type considered, and σα nodal values of the projected
quantity. The projection routine uses a diagonal weight matrix to project the values.
For simple elements the matrix is computed by a procedure identical to mass lumping.
For example,
Z
Mαα =
Nα dΩ
Ω
defines a ‘row sum’ form of projection[2, 3] . Using the above results in the set of equations and a least square fit with the finite element values σ̂ gives the equation set
Z
Nα σ̂ dΩ .
Mαα σα =
Ω
This defines nodal values for projected quantities. Since the coefficient matrix is diagonal the solution to the set of equations for each component is trivial. The actual
solution is performed automatically by FEAP.
To permit each element to project its own quantities it is necessary to add the projection operations for each element under ISW = 8.3 These are performed locally for
each element similar to all other operations. Figure 5.8 shows a simple routine for
two-dimensional elements with 4-stress components begin projected. When multiple
element types are used in an analysis users must be careful to project like quantities
to common values of the ST(nen,*) array so as to get correct results. Also, when
results are displayed it is necessary to plot results by material type to obtain correct
indications of stress discontinuities at material interfaces.
5.6
Elements with History Variables
FEAP provides options for each element to manage variables which must be saved
during the solution. These are history variables and are separated into three groups: (a)
Variables associated with the last converged solution time tn ; (b) Variables associated
with the current solution time tn+1 ; and variables which are not associated to any
particular time. All history variables are associated with the allocation name H which
has a pointer value 49. Users are not permitted direct access to the data stored as
H (of course, it is possible to access from hr(np(49)) but this should not normally
be attempted!). Before calling the element routine for each element, FEAP transfers
the required history variable to a local storage for each type. Users may then access
the history data for each element and if necessary update values and return them
FEAP. Only for specific actions will the local history data be transferred back to the
appropriate H locations. The element history data associated with tn starts at the
memory address of the pointer for NH1 using the double precision dummy array HR in
3
An implementation of the Zienkiewicz-Zhu projection method is implemented using ISW = 24.
CHAPTER 5. ADDING ELEMENTS
52
subroutine slcn2d(sig,shp,xsj,sg,lint,nel,nes, p,s)
c-----[--.----+----.----+----.---------------------------------------]
c
Purpose: Project element variables to nodes
c
c
c
c
c
c
c
c
Inputs:
sig(nes,*)
shp(nel,*)
xsj(*)
sg(3,*)
lint
nel
nes
-
Stresses at quadrature points
Shape functions at quadrature points
Volume element at quadrature points
Gauss points (1,2) and weights (3)
Number of quadrature points
Number nodes on element
Dimension of stress array
c
Outputs:
c
p(nen)
- Weights for ’lumped’ projection
c
s(nen,*) - Integral of variables
c-----[--.----+----.----+----.---------------------------------------]
implicit none
include
include
integer
real*8
’cdata.h’ ! Contains ’nen’
’strnum.h’ ! Contains ’iste’
i,l,lint,nel,nes
xg,p(*),s(nen,*),xsj(*),sig(nes,*),shp(nel,*),sg(3,9)
do l = 1,lint
do i = 1,nel
xg
= shp(i,l)*xsj(l)
p(i)
= p(i) + xg
s(i,1) = s(i,1) + sig(1,l)*xg
s(i,2) = s(i,2) + sig(2,l)*xg
s(i,3) = s(i,3) + sig(3,l)*xg
s(i,4) = s(i,4) + sig(4,l)*xg
end do ! i
end do ! i
iste = 4 ! Returns number projections
end
Figure 5.8: Element variable projection routine
blank common; similarly data for tn+1 starts at the memory address of the pointer for
NH2, and that not associated with a time at NH3. The three pointers are passed to each
element routine in the labeled common
integer
nh1,nh2,nh3
common /hdata/ nh1,nh2,nh3
CHAPTER 5. ADDING ELEMENTS
5.6.1
53
Assigning amount of storage for each element
The specification for the amount of history information to be associated with each
material set is controlled in the isw = 1 task of an element routine. For each material
type specified within the element routine a value for the length of the NH1 and the NH3
data must be provided (the amount of NH2 data will be the same as for NH1). This is
accomplished by setting the variables nh1 and nh2 in common hdata (see above) to
the required values. That is, the statements required are:
if(isw .eq. 1) then
. . .
nh1 = 6
nh3 = 10
. . .
reserves 6 words of NH1 and NH2 data and 10 words of NH3 data for each element with
the current material number. Care should be taken to minimize the number of history
variables since, for very large problems, the memory requirements can become large,
thus reducing the size of problem that FEAP can solve.
5.6.2
Accessing history data for each element
As noted above the data for each element is contained in arrays whose first word is
located at hr(nh1), hr(nh2) (where nh1 and nh2 are pointers) for tn , tn+1 , respectively;
and at hr(nh3) for that not associated with time (note that there are values for each
only if non-zero values are assigned to nh1 and/or nh3 during the isw = 1 task. Any
other allocated data follows immediately after each first word It is a users responsibility
to manage what is retained in each variable type; however, the order of placing the tn
and tn+1 data into the NH1 and NH2 arrays should be identical. There are no provisions
to store integer history variables separately from double precision quantities. It is
necessary to cast the integer data as double precision and move to the history location.
For example, using the statement
hr(nh3+5) = dble(ivarbl)
saves the value for the integer variable ivarbl in the sixth word of the NH3 element
history array. At a subsequent iteration for this element the value of the integer would
be recovered as
ivarbl = int(hr(nh3+5))
CHAPTER 5. ADDING ELEMENTS
54
While this wastes storage for integer variables, experience indicates there is little need
to save many integer quantities and, thus, it was not deemed necessary to provide for
integer history variables separately.
Although users may define new values for any of the hr(nh1), hr(nh2), or hr(nh3)
types, the new quantities will be returned to the H history for the element only for
isw tasks where residuals are being formed for a solution step (i.e., solution command
FORM, TANG,,1, or UTAN,,1 and for history reinitialization during a time update (i.e.,
solution command, TIME). These access the task options isw equal to 3 or 6 and 14,
respectively.
If a user adds a new option for which it is desired to save the history variables, it is
necessary to set the variables hflgu and h3flgu to true as required, if no update is
wanted the variables should be set to false. These parameters are located in
logical
hflgu,h3flgu
common /hdatam/ hflgu,h3flgu
5.7
Energy Computation
FEAP elements provide an option to accumulate the total momenta and energy during
the solution process. The values are accumulated in the array EPL(20) when the switch
parameter isw is 13 and written to a file named Pxxxx.ene (where xxxx is extracted
from the problem input filename) whenever the solution command TIME is used. The
array EPL(2) is in the common block named ptdat6 which has the structure:
real*8
epl
integer
iepl,
neplts
common /ptdat6/ epl(20),iepl(2,20),neplts
For problems in solid mechanics the linear momenta are stored as follows:
The linear momenta are computed as:
Z
p =
ρ v dΩ
Ω
the angular momenta as:
Z
(I ω + x × p) dΩ
π =
Ω
the kinetic energy
Z
ρ v · v dΩ
K =
Ω
CHAPTER 5. ADDING ELEMENTS
55
Component
EPL(1) - EPL(3)
EPL(4) - EPL(6)
EPL(7)
EPL(8)
EPL(9)
EPL(10)
Description
Linear momenta
Angular momenta
Kinetic energy
Stored energy
Work by external loads
Total energy
Table 5.11: Momenta and Energy Assignments
the stored energy as
Z
U =
W (C) dΩ
Ω
and the work by external loads as
Z
(x − X) · Fext dΓ .
V =
Γ
The value of the displacement and velocity at the current time tn+1 are passed in
ul(i,j,1) and ul(i,j,4), respectively. Note that this is true no matter which time
integration algorithm is specified.
5.8
A Non-linear Theory for a Truss
A simple non-linear theory for a two or three dimensional truss which may undergo
large displacements for which the strains remain small may be developed by defining
the axial strain approximation in each member as
2
d−1 ∂us
1 X ∂unj
ss =
+
∂s
2 j=1
∂s
where unj is a displacement component normal to the axis of the member. The virtual
strain from a linearization of the strain is given as
d−1 X
∂δus
∂δunj
∂unj
δss =
+
.
∂s
∂s
∂s
j=1
An algorithm to define the two orthogonal unit vectors which are normal to the member
may be constructed by taking
v = ek
CHAPTER 5. ADDING ELEMENTS
56
where k is a direction for which a minimum value of the direction cosine li exists (for
a 2-dimensional problem defined in the x1 , x2 plane v may be taken as e3 ). Now,
n1 =
v×l
|v×l|
and
n2 = l × n1 .
Using these vectors the two normal components of the displacement are given by
unj ( s , t ) = nj · u( s , t ) =
d
X
nji ui ( s , t )
i=1
and the derivative by
d
X
∂ui
∂unj
=
nji
.
∂s
∂s
i=1
Collecting terms and combining with previously defined quantities the virtual strain
may be written as
∂δu δss =
· g
∂s
where
d−1
X
∂unj
g = l +
nj .
∂s
j=1
After differentiation of the displacement field the discrete form of the virtual strain is
given by
1
−g
δu1 δu2 ·
δss =
.
g
L
Substituting the above virtual strain expression into the weak form gives the modified
residual expression
1
L 2 1 üi1
1
− gi
Ri = b i L
− σss A
− ρA
.
(5.3)
1
gi
2
6 1 2 üi2
The tangent tensor is obtained by linearizing the residual as shown previously. The
only part which is different is the term with σss . Noting that
dss =
and
d δss =
∂du
g ·
∂s
∂δu
∂du
· (n1 ⊗ n1 + n2 ⊗ n2 ) ·
.
∂s
∂s
CHAPTER 5. ADDING ELEMENTS
57
If the ni are constructed as column vectors then the tensor product becomes a matrix
defined as
G = n1 ⊗ n1 + n2 ⊗ n2 = n1 nT1 + n2 nT2 .
With these definitions, the tangent matrix for the non-linear problem is given as
σss A
EA − gi Gij − Gij
− gj gj +
Kij =
.
gi
Gij
L
L2 − Gij
Notice that for the linear problem
gi =
∆xi
L
thus, the only difference between the linear and non-linear problem is the definition of
ss in terms of displacements, the modification for geometric effects for the gi and the
second term on the tangent matrix which is sometimes called the geometric stiffness
part.
Chapter 6
UTILITY ROUTINES
The FEAP system includes several subprograms that can assist developers in writing
new modules. In the next sections we describe some of the routines which perform
numerical integration, compute shape functions and their derivatives, etc.
6.1
Numerical quadrature routines
Details on quadrature formula types and the layout and location of points and weights
may be found in standard references.[4, 5, 2, 3] Here only the description of subroutine
calls is included together with the available options on number of points.
6.1.1
One dimensional quadrature
Line integrals may be evaluated using Gaussian quadrature in which the approximation
to an integral is given as
Z +1
L
X
f (ξ) dξ ≈
f (ξl ) Wl
(6.1)
−1
l=1
where ξl are quadrature points and Wl are the weights to be applied at each point.
The weights satisfy the condition.
L
X
Wl = 2 .
(6.2)
l=1
The Gauss-Legendre formula has points |ξl | which are all less than unity. The subprogram call
58
CHAPTER 6. UTILITY ROUTINES
59
CALL INT1D ( L , SG , WG )
in which L is assigned an integer value between 1 and 5 returns the points in the array
SG(*) and the weights in WG(*), both of which are of type REAL*8. The Gauss-Legendre
formula integrates exactly polynomials up to order 2*L - 1.
The Gauss-Lobato formula has two of its points at -1 and 1 with the remainder in the
interior of the interval. A routine to perform quadrature is obtain by using the call
CALL INT1DL ( L , SW )
in which L is assigned an integer value between 1 and 6. The values of the points and
weights are returned in the two dimensional array SW: Points in SW(1,*) and weights
in SW(2,*).
6.1.2
Two dimensional quadrature
Two dimensional quadrature on quadrilateral domains may be performed by repeated
one-dimensional integration. The two dimensional integrations are approximated by
Z Z +1
L
X
f (ξl , ηl ) Wl
(6.3)
f (ξ, η) dξ dη ≈
−1
l=1
where L is the total of all quadrature points. A routine to compute n × n order
Gauss-Legendre quadrature is obtained by the call
CALL INT2D ( L , LINT, SW )
where L is assigned to the number of points in each direction, LINT is returned as
the total number of points and SW(3,*) is an array containing the points and weights
according to: SW(1,l) contains values of the points ξl ; SW(2,l) contains values of the
points ηl ; and SW(3,l) contains values of the weights Wl .
Two dimensional quadrature on triangles may be performed using the subprograms
call
CALL TINT2D ( L , LINT, SW )
where L is a type indicator, LINT returns the number of points, and SW(4,*) is an array
which returns three area coordinates and the quadrature weight: SW(1,l) returns the
area coordinate L1l (as defined in [2, 3]); SW(2,l) returns the area coordinate L2l ;
SW(3,l) returns the area coordinate L3l ; SW(4,l) returns the weight Wl ; Table 6.1
describes the admissible types, number and location of quadrature points.
CHAPTER 6. UTILITY ROUTINES
Type
1
3
-3
4
6
7
Number
Points
1
3
3
4
6
7
60
Location
Centroid (O(h2 ))
Mid-sides (O(h3 ))
Interior (O(h3 ))
Interior (O(h4 ))
Nodal (O(h3 ))
Interior (O(h6 ))
Table 6.1: Quadrature for triangles
6.1.3
Three dimensional quadrature
Three dimensional quadrature on brick domains may be performed by repeated onedimensional integration. The three dimensional integrations are approximated by
ZZZ
+1
f (ξ, η, ζ) dξ dη dζ ≈
−1
L
X
f (ξl , ηl , ζ) Wl
(6.4)
l=1
where L is the total of all quadrature points. A routine to compute n × n × n order
Gauss-Legendre quadrature is obtained by the call
CALL INT3D ( L , LINT, SW )
where L is assigned to the number of points in each direction, LINT is returned as
the total number of points and SW(4,*) is an array containing the points and weights
according to: SW(1,l) contains values of the points ξl ; SW(2,l) contains values of the
points ηl ; and SW(3,l) contains values of the points ζl ; and SW(4,l) contains values
of the weights Wl .
Three dimensional quadrature on tetrahedra may be performed using the subprograms
call
CALL TINT3D ( L , LINT, SW )
where L is a type indicator, LINT returns the number of points, and SW(5,*) is an array
which returns three area coordinates and the quadrature weight: SW(1,l) returns the
volume coordinate L1,l (as defined in [2, 3]); SW(2,l) returns the volume coordinate L2,l ;
SW(3,l) returns the volume coordinate L3,l ; SW(4,l) returns the volume coordinate
L4,l ; SW(5,l) returns the weight Wl ; Table 6.2 describes the admissible types, number
and location of quadrature points.
CHAPTER 6. UTILITY ROUTINES
Type
1
2
3
4
Number
Points
1
4
11
16
61
Location
Centroid (O(h2 ))
Interior (O(h3 ))
Interior (O(h4 ))
Interior (O(h5 ))
Table 6.2: Quadrature for tetrahedra
6.2
Shape function subprograms
Finite element approximations commonly use shape function subprograms to perform
computations of the functions and their derivatives at preselected points (often the
quadrature points). FEAP includes options to obtain the shape functions for some low
order elements (linear and quadratic order) in one and two dimensions and linear shape
functions for three dimensions. In addition a cubic Hermitian interpolation routine is
available. The calling arguments for routines is summarized below.
6.2.1
Shape functions in one-dimension
Lagrangian interpolation in one-dimensional isoparametric forms may be obtained using the call
CALL SHP1D ( S , XL , SHP, NDM, NEL, XJAC )
where
Parameter
S
XL(NDM,*)
NDM
NEL
SHP(2,NEL)
XJAC
Description
Natural coordinate ξ
Nodal coordinates for element
Spatial dimension of mesh
Number element nodes (2 or 3)
Shape function and derivative
Jacobian transformation
The shape functions are evaluated as: SHP(1,i shape function derivative along the
axis of the element and SHP(2,i) the shape function Ni . In calculations integrals are
represented as
Z
Z 1
f (Ni , Ni,s ) ds =
f [Ni (ξ), Ni,s (ξ)] XJAC(ξ) dξ
(6.5)
L
−1
CHAPTER 6. UTILITY ROUTINES
62
and quadrature may be used for evaluation.
Calculation of natural coordinate derivatives only may be obtained with the call
CALL SHAP1DN( S , SHP, NEL )
where
Parameter
S
SHP(2,NEL)
NEL
Description
Natural coordinate ξ
Shape function and derivative
Number element nodes (2 or 3)
where SHP(1,i contains Ni,ξ and SHP(2,i) the shape function Ni .
Cubic Hermitian interpolation (e.g., for use in straight linear beam elements) given by
w = N1w w̄1 + N2w w̄2 + N1θ θ̄1 + N2θ θ̄2
is obtained using the call
CALL SHP1DH ( S , LEN , SHPW, SHPT )
where
Parameter
S
LEN
SHPW(4,2)
SHPT(4,2)
Description
Natural coordinate ξ
Length of the element (2-node)
Shape functions for wi
Shape functions for θi
The arrays are evaluated as follows:
1. SHPW(1,i), SHPT(1,i) are first derivatives (e.g. Ni,x );
2. SHPW(2,i), SHPT(2,i) are second derivatives (e.g. Ni,xx );
3. SHPW(3,i), SHPT(3,i) are third derivatives (e.g. Ni,xxx ); and
4. SHPW(4,i), SHPT(4,i) are shape functions (e.g. Ni ).
(6.6)
CHAPTER 6. UTILITY ROUTINES
6.2.2
63
Shape functions in two-dimensions
Two-dimensional C0 isoparametric interpolation on quadrilaterals of linear or quadratic
order may be obtained using the subprogram call
CALL SHP2D ( SS, XL, SHP, XJAC, NDM, NEL, IX, FLG )
where
Parameter
SS(2)
XL(NDM,NEL)
NDM
NEL
IX(NEL)
FLG
SHP(3,NEL
XJAC
Description
Natural coordinates ξ, η
Element coordinates in local order
Spatial dimension mesh (2 or 3)
Largest local node number on element
Element global node numbers
Return ξ − η derivatives if true or
x − y derivatives if false
Shape functions and derivatives
Jacobian transformation from x − y to ξ − η.
The array SHP stores the values in the order: SHP(1,i) derivative with respect to ξ or
x; SHP(2,i) derivative with respect to η or y; SHP(3,i) shape function.
Two-dimensional C0 isoparametric interpolation on triangles of linear or quadratic
order may be obtained using the subprogram call
CALL TRISHP ( SS, XL, NDM, IORD, XJAC, SHP )
where
Parameter
SS(3)
XL(NDM,*)
NDM
IORD
XJAC
SHP(3,NEL
Description
Area coordinates L1 , L2 , L3
Element coordinates in local order
Spatial dimension mesh (2 or 3)
Order of interpolation (1,2 or 3)
Jacobian transformation from x − y to ξ − η
Shape functions and derivatives
The array SHP stores the values in the order: SHP(1,i) derivative with respect to
ξ or x; SHP(2,i) derivative with respect to η or y; SHP(3,i) shape function. The
parameter IORD defines the order of interpolation. If it is 1 simple 3-node triangles
with linear interpolation is returned; if 2 quadratic interpolation; if 3 the interpolation
is generated plus a cubic bubble in the seventh function. Giving the IORD parameter
as a negative returns hierarchical form for mid side nodes.
CHAPTER 6. UTILITY ROUTINES
6.2.3
64
Shape functions in three-dimensions
Three-dimensional C0 isoparametric interpolation on bricks of linear order (i.e., 8-node
elements) may be obtained using the subprogram call
CALL SHP3D ( SS, XJAC, SHP, XL, NDM, NEL )
where
Parameter
SS(3)
XL(NDM,8)
NDM
NEL
SHP(4,8)
XJAC
Description
Natural coordinates ξ, η, ζ
Element coordinates in local order
Spatial dimension mesh (2 or 3)
Number nodes on element: 8 = linear brick; 20 = serendipity
quadratic; 27 = lagrangian quadratic; 64 = lagrangian cubic
Shape functions and derivatives
Jacobian transformation from xyz to ξηζ.
The array SHP stores the values in the order: SHP(1,i) derivative with respect to x;
SHP(2,i) derivative with respect to y; SHP(3,i) derivative with respect to z; SHP(4,i)
shape function.
Three-dimensional C0 isoparametric interpolation on tetrahedra of linear order (i.e.,
4-node elements) may be obtained using the subprogram call
CALL TETSHP ( SS, XL, NDM, ORDER, XJAC, SHP )
where
Parameter
SS(4)
XL(NDM,4)
NDM
ORDER
XJAC
SHP(4,4
Description
Volume coordinates L1 , L2 , L3 , L4
Element coordinates in local order
Spatial dimension mesh (2 or 3)
Interpolation order: 1 = linear, 2 = quadratic
Jacobian transformation from xyz to ξηζ
Shape functions and derivatives
The array SHP stores the values in the same order as for the brick element.
6.3
Eigenvalues for 3 × 3 matrix
Three dimensional problems often require the solution of a 3 × 3 eigenproblem to
generate principal values and directions. FEAP includes a special routine to calculate
CHAPTER 6. UTILITY ROUTINES
65
the values and vectors for symmetric arrays. The routine is used by a call to the
subprogram as
CALL EIG3 ( V, D, ROT )
On call to the routine V(3,3) is a REAL*8 array containing the symmetric array to
be diagonalized. On return the eigenvalues are contained in D(3) and the vectors
for each value in the columns of the V array. A Jacobi method is used with ROT an
integer parameter returning the number of rotations to diagonalize. The routine is
quite efficient compared to any attempt to compute vectors after closed form solution
of the cubic for roots.
In addition to the general eigensolution above FEAP includes options to compute
principal values of a symmetric second order tensor for two and three dimensional
problems. In two dimensional use, the call to
CALL PSTR2D ( SIG, PV )
is used where SIG(4) stores stresses in the order σ11 , σ22 , σ33 , σ12 and returns principal
values and directions in PV(3) in the order σ1 , σ2 , and θ, where the angle is in degrees
between x and the 1-axis. This routine does not use SIG(3).
In three dimensions the principal values are obtained using the call
CALL PSTR3D ( SIG, PV )
where SIG(6) stores stresses in the order σ11 , σ22 , σ33 , σ12 , σ23 , σ31 , and returns
principal values in PV(3) in the order σ1 , σ2 , σ3 . Roots are ordered from most positive
to most negative.
6.4
Plot routines
Several options exist in the FEAP system to create graphical plots for data and results.
6.4.1
Mesh plots
FEAP has plot capabilities to represent some standard element shapes. By default it
is assumed that all plots are for two dimensional elements of quadrilateral shape and
with between 4 and 9 nodes (numbered as shown in the manuals - vertex nodes first,
mid side nodes, center node). If other shapes are to be represented it is necessary to
CHAPTER 6. UTILITY ROUTINES
66
include drawing instructions in the element routine segment where material properties
are input (i.e., ISW = 1). A parameter describing the type of element and named IEL
is contained in the labeled common ELDATA and, thus, this common must be included
to activate the plot option call. The known types of plots are:
1. Point element with one node obtained by call
CALL PLTPT1 ( IEL )
2. Line element with two nodes obtained by call
CALL PLTLN2 ( IEL )
3. Triangular element with 3-nodes obtained by call
CALL PLTRI3 ( IEL )
4. Triangular element with 6-nodes obtained by call
CALL PLTRI6 ( IEL )
5. Quadrilateral element with 4-nodes obtained by call
CALL PLQUD4 ( IEL )
6. Tetrahedral element with 4-nodes obtained by call
CALL PLTET4 ( IEL )
7. Brick element with 8-nodes obtained by call
CALL PLBRK8 ( IEL )
Using these and internal extraction of element surfaces the program is able to make
some hidden surface plots in three dimensions. Failure to include a plot sequence will
usually result in unpredictable plots of the mesh and contour values.
CHAPTER 6. UTILITY ROUTINES
6.4.2
67
Element data plots
Users may construct plots within their elements (i.e., an ELMTnn) and access using the
plot command:
PLOT,PELE,v1,v2,v3
In interactive mode in the plot environment it is only necessary to enter
PELE,v1,v2,v3
The values entered in v1,v2,v3 are optional and are passed to the element through a
common block as
REAL*8
ELPLT
COMMON /ELPDAT/ ELPLT(3)
The PELE option calls each element with the switch parameter ISW = 20. Users merely
code whatever option they wish to include within their element module.
The standard color table is available through use of the subroutine call
CALL PPPCOL(ICOL, 0)
in which ICOL designates the color to be assigned according to Table 6.3. An exception
occurs for PostScript outputs where black and white are switched (since the background
then is assumed to be white).
A straight line segment may be drawn to the screen in the current color between the
coordinates (x1 , y1 , z1 ) and (x2 , y2 , z2 ) using the commands
CALL PLOTL(X1,Y1,Z1, 3)
CALL PLOTL(X1,Y1,Z1, 2)
Here the basic command is
CALL PLOTL(X1,Y1,Z1, IP)
where the three cartesian coordinates relate to mesh coordinates (not screen values)
and IP is a parameter defined according to Table 6.4.
The perimeter of a panel is drawn with standard line drawing commands starting with
CHAPTER 6. UTILITY ROUTINES
ICOL
0
1
2
3
4
5
6
7
8
9
COLOR
Black
White
Red
Green
Blue
Yellow
Cyan
Magenta
Orange
Coral
68
ICOL
10
11
12
13
14
15
16
17
18
COLOR
Green-yellow
Wheat
Royal blue
Purple
Aquamarine
Violet-red
Dark slate blue
Grey
Light grey
Table 6.3: Color pallet for FEAP plots
IP
1
2
3
Action
Start panel fill
Move to point
Draw to point
Table 6.4: Values for control of plots
CALL PLOTL(X1,Y1,Z1, 1)
and continuing with a sequence of draw commands
CALL PLOTL(Xi,Yi,Zi, 2)
(however, no lines appear on the screen) and the fill of each panel is completed by the
statement
CALL CLPAN
It should be noted that all plots within FEAP are performed in three dimensions.
For two dimensional problems no zi coordinates are available in the XL(NDM,NEN)
array and, hence, it is necessary to assign zero values for the zi coordinates before
calling a plot subprogram. If a perspective view has been requested a full use of a
xi , yi , zi specification is made. In this case a user may wish to pass the value of some
solution variable as the zi value (scaled so that it will make sense relative to the xi , yi
coordinate values). Similarly, if deformed plots are being performed it is necessary
to add (scaled) displacements to the coordinates. The current value of the scaling
parameter (i.e., variable CS) is available in labeled common PVIEW. In this case one can
CHAPTER 6. UTILITY ROUTINES
69
add the statements (assuming here that the displacements correspond to the coordinate
directions)
DO NE = 1,NEL
DO I = 1,NDM
XP(I,NE) = XL(I,NE) + CS*UL(I,NE)
END DO ! I
END DO ! NE
(NEL is the number of connected nodes to each element and is passed through labeled
common ELDATA) before performing any deformed plots and then plot the appropriate
values of XP. Indeed, this may always be performed as the value of CS will be zero for
an undeformed plot.
6.4.3
Other user plots
It is also possible for users to prepare plot outputs unrelated to elements. The plot
command
PLOT UPLOt v1 v2 v3
initiates a call to the subroutine UPLOT which has the basic structure
SUBROUTINE UPLOT(CT)
IMPLICIT
NONE
REAL*8
CT(3)
...
END
The argument CT contains the values for the three parameters v1, v2, v3. The default
color is white. Direct plots in screen coordinates [lower left at (0,0); upper right at (1,1)]
may be given using the statement
CALL DPLOT(XS,YS, IP)
where XS, YS are between zero (0) and one (1) and IP is interpreted according to Table
6.4. Panels are closed using
CALL CLPAN
and colors treated according to values specified in calls to PPPCOL.
Bibliography
[1] G.G. Weber, A.M. Lush, A. Zavaliangos, and L. Anand. An objective timeintegration procedure for isotropic rate-independent and rate-dependent elasticplastic constitutive equations. International Journal of Plasticity, 6:701–744, 1990.
[2] O.C. Zienkiewicz and R.L. Taylor. The Finite Element Method, volume 1. McGrawHill, London, 4th edition, 1989.
[3] O.C. Zienkiewicz and R.L. Taylor. The Finite Element Method: The Basis, volume 1. Butterworth-Heinemann, Oxford, 5th edition, 2000.
[4] M. Abramowitz and I.A. Stegun, editors. Handbook of Mathematical Functions.
Dover Publications, New York, 1965.
[5] T.J.R. Hughes. The Finite Element Method: Linear Static and Dynamic Analysis.
Prentice-Hall, Englewood Cliffs, N.J., 1987.
70
Index
Help files, 1
Adding elements, 26
ELMTnn subprogram, 26
Arguments on subprogram ELMTnn, 28
Array allocation, 8
PALLOC subprogram, 9
Removing arrays, 10
UALLOC subprogram, 9
Include statements, 3
Locating arrays
PGETD, 13
Material models, 16
Auto time step control, 19
History variables, 19
Internal variables, 19
UCON use, 16
UMATIn subprogram, 17
UMATLn subprogram, 17
Material property variables, 30
Mesh array names, 11
Mesh input
UMESHn subprogram, 14
Mesh manipulation
UMANIn subprogram, 21
Mesh plots, 65
Brick element, 66
Line element, 66
Point element, 66
Quadrilateral element, 66
Tetrahedral element, 66
Triangular element, 66
Color in plots, 24
Common block definitions, 34
Common statements, 3
Data input
Converting text to real, 7
PINPUT, 5
TINPUT, 6
Data output
Integer arrays, 7
Real arrays, 7
Debug output, 3
Eigenvalues for 3 × 3 matrix, 64
Element array names, 11
Element common block variables, 31
Element connection array, 12
Element control array, 12
Element energy computation, 54
Element history variables, 51
Accessing for each element, 53
Element options
Projection to nodes, 50
Task 1 (ISW=1), 49
Task 6 (ISW=6), 49
TPLOt definitions, 49
Element plots
User, 67
Element task definitions, 29
Newton solution
Element residual array storage, 40
Element tangent array storage, 39
Non-linear problems, 32
Tangent array definition, 34
Numerical integration, 58
One dimension, 58
Three dimension, 60
Two dimension, 59
71
INDEX
Plot command
UPLOT subprogram, 69
UPLOTn subprogram, 23
Plot utility function
Fills, 67
Lines, 67
Plot utility functions
Color, 24
Fills, 24
Lines, 24
Precision
Integer, 1
Real, 1
Principal stress
Three dimensional, 65
Two dimensional, 65
Problem size, 1
Quadrature, 58
One dimension, 58
Three dimension, 60
Two dimension, 59
Setting colors, 24
Setting options
Help level, 2
Parsing statements, 2
Plot prompts, 2
PostScript, 2
Shape functions, 61
Brick isoparametric, 64
One dimensional cubic Hermitian, 62
One dimensional isoparametric, 61
One dimensional natural derivative,
62
Quadrilateral isoparametric, 63
Tetrahedral isoparametric, 64
Three dimensional isoparametric, 64
Trianglular isoparametric, 63
Two dimensional isoparametric, 63
Solution array names, 11
Solution command
UMACRn subprogram, 22
Table of colors, 25
72
Transient solution
CTAN - definition, 41
Newton method, 32
Non-linear, 30
Truss element
Newmark method, 45
Non-linear theory, 55
Tangent modulus, 56
Theory, 41
Variational equation, 44
Weak form, 44

advertisement

Was this manual useful for you? Yes No
Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Download PDF

advertisement

Languages