Arcweld FE simulation of large 3D structures using MSC'Marc for Volvo Car Corporation

1
Arcweld FE simulation of large 3D structures
using MSC.Marc for Volvo Car Corporation
MSC.Software The right partner in process
simulation
2
Problem definition

• thermo-mechanical coupling
• non-linear material properties at high
  temperatures
• material properties depending on strain and
  strain rate
• Definition of a moving heat source with input
  from ROBCAD
• multiple welding heads
• modelling of clamping conditions
• weld geometry (FE-modeling)
• FE-modelling strategy (solid, shell, beam)

3
Problem definition (contd.)

• User subroutines are standard features of
  MSC.Marc
• User subroutines are used for controlling flux
  distribution and work hardening
• Programming of arbitrary shaped moving heat
  source is done with user subroutine FLUX
• Material behaviour is programmed with user
  subroutine WKSLP

4
Main goal

• Predicting the residual deformation in the
  structure

5
Project phases

• Step 1
• Simple T-node geometry
• Development of welding routines
• Verification against welding tests

• Step 2
• Ladderframe structure with same topology as the
  underbody
• Development of routines for multiple weldingheads
• Verification agains welding tests (in progress)

• Step 3
• Complete underbody structure
• Model reduction to save computational time
• Verification against welding tests (in progress)

6
Step 1, T-node

• Decision to go for a shell model
• One robot welding 4 seams
• Good agreements reached on both temperatures and
  deformations

7
Step 1, T-node

• Modelling of welding seams using shell elements
• Thickness of shells in the seam have been set so
  that the correct cross section area is reached

8
Step 1, T-node

• Clamping modelling using stiff springs

9
Step 1, T-node

• Some simulation statistics
• 6500 elements
• 2.5mm elm size along seam
• 10-15mm global elm size
• 4 consecutive welding seams
• weld data input from ROBCAD files
• total simulation time is 200 seconds including
  150 seconds cooling
• 365 increments total in simulation
• computer time equals to appr. 50 hours on a
  single CPU ws

10
Step 1, T-node

• Simplified model to reduce calculation times
• Replace parts of the structure with beam elements
• Model size reduced to 1/3rd
• Calculation times reduced to appr 20 of original
• Correct principal behaviour can still be found
  with a much simpler model

11
Step 2, Ladderframe

• Modification to handle multiple welding robots
• Two robots welding 16 seams each
• A graphical pre-step was performed to verify
  welding data before simulation
• Parallel execution necessary

12
Step 2, Ladderframe

• Some simulation statistics
• 27000 elements
• 2.5mm elm size along seam
• 15 mm global elm. Size
• 2 welding heads welding in parallell
• each welding head is welding 16 seams
• weld data input directly from ROBCAD files, one
  file per welding head
• total simulation time is 4000 seconds including
  3674 seconds cooling to reach room temperature
• 3673 increments total in simulation
• computer time equals to 155 hours elapsed on a 8
  cpu HP V-2250

13
Step 2, Ladderframe

• Graphical pre-step to verify orientation of
  welding flame and welding power

14
Step 2, Ladderframe

• Graphical pre-step to verify orientation of
  welding flame and welding power

15
Step 2, Ladderframe

• Simulation results
• Graphical results available
• Avi-movies showing temperatures and deformations
• Curve-plots showing measurement location
  displacements as function of time
• Stress- and deformation fringe plots after
  cooling period

16
Step 2, Ladderframe

• Simulation results
• Avi-movie showing temperatures and deformations

17
Step 2, Ladderframe

• Simulation results
• Curve-plots showing measurement location
  displacements as function of time

18
Step 2, Ladderframe

• Simulation results
• Residual stresses and deformations available on
  fringe plots

19
Step 2, Ladderframe

• Simulation summary
• The FE-model is heated up faster then the
  verifying object
• The cooling is to rapid in the FE-model
• Correct principal behaviour in the FE-model

20
Step 3, Underbody

• Simplification of FE-model to reduce calculation
  time
• Two robots welding 9 seams each
• A pre-step was performed to verify welding data
  before start of simulation

21
Step 3, Underbody

• Some simulation statistics
• 11000 elements
• 2.5mm elm size along seam
• 15 mm global elm. size
• equivalent beam sections have replaced parts
  between welding locations
• 2 welding heads welding 9 seams each
• weld data input directly from ROBCAD files, one
  file per welding head
• total simulation time is 4000 seconds including
  3745 seconds cooling to reach room temperature
• 1839 increments total in simulation
• computer time equals to 27 hours on a 4P SGI
  O2000 (R10k/250Mhz)

22
Step 3, Underbody

• Simulation results
• Graphical results available
• Avi-movies showing temperatures and deformations
• Curve-plots showing measurement location
  displacements as function of time
• Stress- and deformation fringe plots after
  cooling period

23
Step 3, Underbody

• Simulation results
• Avi-movie showing node 1 right side

24
Step 3, Underbody

• Simulation results
• Avi-movie showing node 2 right side

25
Step 3, Underbody

• Simulation results
• Avi-movie showing node 3 right side

26
Step 3, Underbody

• Simulation results
• Avi-movie showing node 4 right side

27
Step 3, Underbody

• Simulation results
• Residual deformations after 4000 seconds

28
Step 3, Underbody

• Simulation results
• Deformation history

29
Step 3, Underbody

• Simulation summary
• Simulations are to be verified against welding
  tests
• Interesting behaviour on residual deformation due
  to clamping geometry seen from simulations
• Welding simulations possible to perform
  overnight

30
Conclusions

• All the features necessary for performing the
  complete welding simulation are available
• Correct principal structural behavior compared to
  measurements
• Welding of large 3D structures can be simulated
  with reasonable computing times
• Short loop times are possible with automation of
  simulations
• It is now possible to draw conclusions from
  different welding layouts before the prototype
  exists!

31
Future actions

• Simulation performance
• Use existing FE-models
• Material data
• Automation of simulation steps
• Clamping modelling
• Residual stress as input for fatigue analysis
• Activate weld elements automatically
• Consider deformations of structure before welding
• ..