Skip to content

Latest commit

 

History

History
173 lines (149 loc) · 7.65 KB

Abaqus_tip.md

File metadata and controls

173 lines (149 loc) · 7.65 KB
Raw

Changing license server from the client

  • In windows laptop
  • Edit c:\Program files \Abaqus\6.8.x\site\abaqus_v6.env
  • abaquslm_license_file="XXXX@YYY.abc.org" into new one

N. of license tokens

When scratch space is required

  • Run from /tmp to /work/$account
    • In CAE, Analysis-> Jobs-> Create, then edit “General” tab
    • in CLI, abaqus scratch=/work/$USER j=aaa int

Abaqus runs background, and this could be a problem in PBS batch queue

  • Use “interactive” command to prevent this: abaqus j=aaa int

When GUI looks weird (or message are becomes gray)

  • Delete ~/abaqus_v6.14.gpr then reload abaqus CAE

Plug-in installation

  • Users may ask to install 3rd party plugins
    • Just unzip them at abaqus/6.xxx/code/python2.7/lib/abaqus_plugins
    • Double-check if the plugin appears in the GUI menu
    • The menu may change per functionality wise
    • CAE -> VIEW or only at analysis mode etc.

User subroutine in Linux

  • Make sure that you have fortran compiler enabled: default is ifort
  • abaqus j=inp_name user=myForSub
  • Read *.log file, and check: Begin Compiling Abaqus/Standard User Subroutines
  • Sample file: $ abaqus fetch j=boltpipeflange_3d_usr_umat

Fortran compiler options

  • Check SimulationServices/V6R2017x/linux_a64/SMA/site/lnx86_64.env
  • License is registered in custom_v6.env:abaquslm_license_file="port@server"
    • In windows, check C:\Program Files\Dassault Systemes\SimulationServices\V6R2017x\win_b64\SMA\site\win86_64.env

Using intel compiler 2017 and vs2013 at abaqus 2017 on windows

  • Edit Abaqus Command icon: "C:\Program Files (x86)\IntelSWTools\compilers_and_libraries_2017\windows\bin\ifortvars.bat" intel64 vs2013 & C:\windows\system32\cmd.exe /k
  • Edit Abaqus CAE icon: "C:\Program Files (x86)\IntelSWTools\compilers_and_libraries_2017\windows\bin\ifortvars.bat" intel64 vs2013 & C:\SIMULIA\CAE\2017\win_b64\resources\install\cae\launcher.bat cae || pause

Using abaqus in windows

  • abaqus verify -all # verify installed abaqus
  • Linux command: abaqus j=aaa inp=bbb user=ccc cpus=12
  • Windows command: abaqus –j aaa –inp bbb –user ccc –cpus 12

Parallel computing in abaqus

  • abaqus j=input cpus=4 mp_mode=threads (or MPI)
    • MPI runs only on iterative solver. Edit #step as #step, solver=iterative

GPU run

  • abaqus j=test gpus=1
    • To monitor GPU usage, nvidia-smi -l

Running double precision

  • abaqus j=aaa double cpus=...

Installing scipy

  • Download Linux binary wheel of scipy 0.9 (0.10.1 works as well) from PyPi
  • Unzip
  • Copy scipy folder to SIMULIA/CAE/2017/linux_a64/tools/SMApy/python2.7/lib/python2.7/site-packages
  • Copy or make a static link the same scipy into SimulationServices/V6R2019x/linux_a64/tools/SMApy/python2.7/lib/python2.7/site-packages SIMULIA/Tosca/2019/linux_a64/tools/SMApy/python2.7/lib/python2.7/site-packages
    • match the version of scipy with built-in numpy of abaqus. Find the version made in similar time
    • EX: in abaqus 2019, numpy.version = 1.6.2 => May 2012 => scipy 0.10.1 as of Feb 2012

In Python scripting

Face vs FaceArray

  • A=mdb.models['Case_1'].parts['Monarch_Pipette'].faces.getSequenceFromMask(('[#1200 ]', ), )
  • B = A[0:1]
  • mdb.models[Model_Name].parts[Part_Name].Surface(name='Load_Contact',side1Faces= B)
    • There might NOT be a constructor of FaceArray
    • Instead of A[0], use A[0:1], keeping the sequence

Find nearby edges

  • tmp = instance.edges.findAt(((ix,iy,0),)) # id = tmp[0].index
  • next = tmp[0].getAdjacentEdges() # find nearby edges of edge "tmp"
for x in next:
	p1 = x.getVertices()[0]  # 1st point of nearby edge
	p2 = x.getVertices()[1]  # 2nd point of nearby edge
	x1 =  instance.vertices[p1].pointOn[0][0] #x coord of 1st point
	y1 = instance.vertices[p1].pointOn[0][1]

tmp = instance.vertices.findAt(((ix,iy,0),))
next = tmp[0].getEdges()
for seg in next:
    if not (seg in indices_seam_edges):
        edges += instance.edges[seg:seg+1]

Find a crack seam

seam_edges = seam_edges + assembly.sets['Crack Seam ' + str(i)].edges;
indices_seam_edges = []
for seg in seam_edges:
    indices_seam_edges.append(seg.index)

tmp = instance.edges.findAt(((ix,iy,0),)) # id = tmp[0].index
next = tmp[0].getAdjacentEdges()
for seg, n in zip(next,range(len(next))):
    if not (seg.index in indices_seam_edges):
        edges += next[n:n+1]

Installing Abaqus 2019

  • only ksh: bash/sh not working
  • The SIMUIA Established product installers aborts with an error message both in GUI and Command line (TUI) mode with the following error message:
  • ERROR: Cannot wait for process "/<Abaqus_Installer_directory_path>/AM_SIM_Abaqus_Extend.AllOS/1/inst/linux_a64/code/bin /DSYInsAppliGUI" "-CDpath" "/<Abaqus_Installer_directory_path>/AM_SIM_Abaqus_Extend.AllOS/1/"
  • Workaround: The workaround is to set the size of stack limit from unlimited to "10240".
  • sh/ksh/bash shell: $ ulimit -s 10240
  • csh/tcsh shell: $ limit stacksize 10m (where 10m=10240 KB)
  • ksh rpm must be installed
  • On a serial file system. Unpacking on Lustre could be an issue - media check fails.

Section defintion for 2D

  • Do not use generalized plane strain. This may yield an error of node set assembly not defined
  • Use Homogeneous and select 2D in the option
    • Thickness can be adjusted in the option

Environmental variables

  • abaqus information=environment # displays current configuration
  • Ex) export ABA_SINT_CAP=65536

Feeding hostnames

for n in $(sort -u $PBS_NODEFILE); do
mp_host_list="${mp_host_list}['$n',$(grep -c $n $PBS_NODEFILE)],"
done
mp_host_list=$(echo ${mp_host_list} | sed -e "s/,$/]/")
echo "mp_host_list=[${mp_host_list}" > abaqus_v6.env

Python scripting

  • abaqus cae nogui=script.py
  • Doing re-partitions may decrease the performance. Draw segment-wise in CAD or sketch steps

RDMA error

  • Error message: multiple pkey found in partition key table, please choose one via MPI_IB_PKEY
  • Or: MPI_Init: Can't initialize RDMA device
  • or: Cannot initialize RDMA protocol
  • Solution
    • export MPI_IB_PKEY=0xffff
    • For multiple nodes, hardcode export MPI_IB_PKEY=0xffff into .bashrc and .bash_profile

Plotting data along path

  • Extracting nodal number in sequene of 5,3,2 ... might not be easy
  • Draw a line segment in advance, and pick up points, which exist already
key_list = output.rootAssembly.nodeSets.keys()
if 'Some Edges' in key_list:
    path1 = session.Path(name='Path-1', type=POINT_LIST,
                 expression=((0,1.1,0), (0,-1.1,0)))
    session.viewports['Viewport: 1'].odbDisplay.setPrimaryVariable(
        variableLabel='S', outputPosition=INTEGRATION_POINT, 
        refinement=(INVARIANT, 'Pressure'))
    session.viewports['Viewport: 1'].odbDisplay.setFrame(step=0, frame=1) # first step is 0, second step is 1
    v1 = (('S', INTEGRATION_POINT, ((INVARIANT, 'Pressure' ), )),) 
    newXYData=session.XYDataFromPath(name='sectiondata',
                                     path=path1,\
                                     includeIntersections=TRUE,\
                                     shape=UNDEFORMED,\
                                     labelType=Y_COORDINATE,
                                     pathStyle=UNIFORM_SPACING,
                                     numIntervals=200,
                                     variable=v1)
    session.writeXYReport(fileName="section_pressure.txt", xyData=(newXYData,))