The document discusses finite element analysis of structures and its applications, innovations, and challenges. It describes a presentation given by Tanguy Mertens of Siemens on structural design and mechanical integrity simulation. The presentation covered challenges in gas turbine thermal and structural simulation like multi-physics coupling and material modeling. It demonstrated Simcenter 3D's capabilities for cyclic symmetry analysis of multiple engine stages, combined 2D-3D modeling, and lifing analysis accounting for creep. Advanced modeling of manufacturing processes like blade manufacture from hot-to-cold deformation was also highlighted.
36. de structures par éléments finis :
applications, innovations et défis
Jean-Philippe PONTHOT, University of Liège, Belgium
Romain Boman, Luc Papeleux
Department of Aerospace and Mechanical Engineering
JP.Ponthot@uliege.be
M.L. Cerquaglia, B.J. Bobach, E.F. Sanchez-Fernandez, M. Lacroix, S. Février,
C. Laruelle, Y. Crutzen, G. Tanaka
Liège Creative, 26 avril 2022
37. Contents
2
1. Introduction/General context
2. Roll forming of complex parts
3. Wear/ Rotor Stator interactions in an aeroengine
4. Additive Manufacturing (macroscopic scale)
5. Alternative discretization techniques: PFEM
38. 3
J.-P. Ponthot
Our lab within the university
Numerical simulation
Solid mechanics
Fluid-Structure Interactions
Finite element method
Software development
Computational Mechanics
hydroforming of a tube
Dept of Aerospace and
Mechanical Engineering
(Faculty of Applied Sciences)
39. Our main simulation code: Metafor
4
Implicit Finite-Element solver
for the numerical simulation of large deformations of solids
ALE Formalism, remeshing.
Thermomechanical time-integration schemes.
Modeling of cracks,
fracture.
Contact algorithms.
Nonlinear constitutive laws.
Mesh generation from medical
images.
Fluid finite elements.
Monolithic schemes.
Coupling with
extenal solvers.
Metal Forming applications Crash / Impact
Biomechanics Fluid/structure
interaction
40. pin
shoulder
welded
zone
advancing
side
retreating
side
Software development and numerical simulation of
problems involving large strains, contacts, coupled thermo-
mechanics and complex material behavior modeling:
Metal forming processes (deep drawing and
springback, superplastic forming, cold rolling,
Impact simulation and crashworthiness
Tire mechanics & rubber
Biomechanics
Research interests
42. Contents
7
1. Introduction/General context
2. Roll forming of complex parts
3. Wear/ Rotor Stator interactions in an aeroengine
4. Additive Manufacturing (macroscopic scale)
5. Alternative discretization techniques: PFEM
44. U-channel
9
Industrial application Roll forming
Forming of a symmetrical U-channel
Experimental mill (ArcelorMittal R&D, Montataire, France)
6 stands (15°, 32°, 50°, 68°, 80°, 90°)
Final bending radii: 6 mm
Inter-stand distance : 0.5 m
Sheet : 2000 x 200 x 1.6 mm
Sheet velocity: v = 200 mm/s
Coulomb friction = 0.2
DP980 steel ( Y0 = 697.34 MPa)
Numerical parameters
Symmetry
Friction drives the sheet
Two layers of EAS elements
Dynamic implicit scheme (Chung-Hulbert)
Process parameters
45. U-channel
10
Industrial application Roll forming
Numerical vs. experimental springback
The final shape has been digitised using a high precision 3D measurement device
and fits well both numerical curves (courtesy of ArcelorMittal)
Lagrangian : (21 320 FEs)
ALE : (12 768 FEs)
CPU Times:
ALE is
2.6x faster!
DEFORMATION LONGITUDINALE J1 PEAU SUP EN FONCTION DE LA DISTANCE DE PROFILAGE
-0,20%
-0,10%
0,00%
0,10%
0,20%
0,30%
0,40%
9,00E+02 1,40E+03 1,90E+03 2,40E+03 2,90E+03 3,40E+03 3,90E+03 4,40E+03 4,90E+03
Distance de profilage en mm
Déformation
Ingénieur
METAFOR
MES. EXP.
46. Forming of a rocker panel
11
Industrial application Roll forming
Simulation of an industrial line
Process parameters
16 stands unsymmetrical shape
Material: DP980
Sheet: 5950 x 165 x 1.5 mm
Mesh
1 FE through the thickness
FE length: from 3mm to 30mm
155 652 dofs
stand #1
stand #16
forming
direction
closed
cross section
48. Contents
13
1. Introduction/General context
2. Roll forming of complex parts
3. Wear/ Rotor Stator interactions in an aeroengine
4. Additive Manufacturing (macroscopic scale)
5. Alternative discretization techniques: PFEM
49. Recent Fan Blade Out Problem
14
Southwest WN1380 New-York-Dallas, April 17, 2018
Boeing 737-700/ CFM56-7B24 flying at 32 000 feet
54. Context and motivation
One way of increasing engine efficiency is to
decrease the clearance between the
rotating blades and the casing (thus
avoiding leakage flows)
A reduction of 25% of the clearance means
an increase of 1% of the engine efficiency.
Increasing by 1% engine efficiency leads to
saving 200 000 liters of fuel per year for a
middle range aircraft*!
From the mechanical point of view, the
clearance becomes so small that sometimes
the blades come into contact with the casing
(the shaft deforms during e.g. brutal
manoeuvers or gusts
To mitigate the contact forces, aircraft
engine manufacturers use an abradable
coating
*Lattime S.B., Steinez B.M. Turbine Engine Clearance Control Systems:
Current Practices and Future Directions. Report NASA/TM-2002-211794
55. Wear in Blade-Casing interaction
20
Abradable seal :
Abradable
Casing
Blade
Abradable thickness ~ 2-3 mm
56. Low Pressure Compressor = Booster
Typical booster architecture
Industrial partner: Safran Aero Boosters
Yellow and red: rotating parts
Blue: fixed parts
Typical clearance: 2% blade chord e.g. 1mm for a 50 mm compressor blade
57. Typical abradable material
The ideal abradables material must resist erosion (due to particle impacts
but must be easily worn when hit by a blade.
Typical abradables material (e.g. METCO 601NS, DURABRADE)
Al-Si12% to resist erosion
Polyester to allow abrasion by the blade
Manufactured by thermal spray coating
E ~ 1500 MPa
58. In case of contact the blades start to vibrate
Worst case scenario:
What is sometimes observed is a synchronization of the blade
frequency with the engine configuration
In other words the blade vibrates an integer number of times per
revolution and interacts with the abradable (8 times per
revolution in the figure on the right)
Under some (unknown) conditions, the blade can start to vibrate with a large amplitude
and hits the abradables several times during a revolution
Two basic scenarios:
The abradables is worn and there is no longer any interaction
with the blades
The self-excited process quickly leads to blade failure
Blade tearing due to fatigue
59. A wear model for abradable materials
We have to manage contact between the blade and casing, as well as wear of
abradable material, while keeping computational time under control!
60. Wear update over a surface
Wear surface is represented
thanks to isoparametric
coordinates .
Wear profile is stored at the
nodes (green dots), and can
be interpolated.
3D Wear evolution algorithm in isoparametric space
62. ONERA bench test
27
Bench test developed at ONERA/Centrale Lille,
France, PhD of Sarah Baïz
Etude expérimentale du contact aube/abradable :
contribution à la caractérisation mécanique des
matériaux abradables et de leur interaction
dynamique sur banc rotatif avec une aube.
N.B. Curvature is opposite to a aeroengine casing,
but it allows a better view of the phenomenon
VP = drum velocity
DN & DT = Normal and tangential displacement
FN = Normal Force
T = strain at the base of the blade
68. Experimental approach: Safran Aero Boosters
Experimental set up:
Thermal camera imaging during the test
Wear pattern (8 lobes)
69. Experimental approach
EO = Engine Order
EO8 means 8 interactions per revolution
(8 lobes )
Drawbacks of experimental approach:
High cost!
Low flexibility to test different blade designs
Test rig availability
Numerical model?
70. Parametric studies can be undertaken
Typical wear pattern at different angular speeds (N is the number of lobes, T = torsional mode, F = Flexure/bending mode)
72. Wear pattern and gage signal evolution
Wear pattern evolution Gage signal evolution
N.B. The last two pictures show that the abradable has been broken in
73.
74. Blisk model
56 blades
One of the blades is a little bit longer
A small mass (29 gr) to trigger unbalance
Casing is not exactly centered!
12 926 hexahedral elements
67 776
Diameter =~500 mm
76. Contents
41
1. Introduction/General context
2. Roll forming of complex parts
3. Wear/ Rotor Stator interactions in an aeroengine
4. Additive Manufacturing (macroscopic scale)
5. Alternative discretization techniques: PFEM
77. Additive manufacturing test 3D (Metz)
& collaboration with A-M Habraken (Uliège)
42
107 Layers
(175mm)
[2] ment Bourlet. Développement de la fabrication additive par procédé arc-fil pour les aciers : caractérisation
microstructurale et mécanique des dépôts en nuances ER100 et 316L pour la validation des propriétés d'emploi de pièces
industrielles. Autre [cond-mat.other]. Ecole nationale supérieure d'arts et métiers - ENSAM, 2019. Français. NNT :
2019ENAM0058 . tel-02860062
88. Contents
53
1. Introduction/General context
2. Roll forming of complex parts
3. Wear/ Rotor Stator interactions in an aeroengine
4. Additive Manufacturing (macroscopic scale)
5. Alternative discretization techniques: PFEM
90. Motivation for PFEM
55
Avoid distortions such as those encountered in updated
Lagrangian formulation for solid mechanics
Combine the advantages of classical FEM and particle
methods (e.g. SPH)
at the nodes like in particle methods
Evolution is computed thanks to a FEM discretization
Use Lagrangian representation to easily track
evolution of interfaces
so there is no need for an interface tracking algorithm
91. PFEM: how does it work?
56
The first step in the PFEM is discretizing the continuum with
some particles/nodes
The particles carry all the physical and mathematical
The equations are written in their Lagrangian form.Thus
external boundaries are easily determined by following the
particle motion.
92. At each time step a new mesh is quickly built and
boundaries are determined thanks to the shape
technique.This mesh is used to solve the weak form
using classical FEM over one time step.
Distorted elements and external boundaries are
determined thanks to the shape algorithm
Classical FEM computation
93. PFEM in the literature
58
Seminal contribution from E. Onate and S. Idelsohn:
Idelsohn S.R., Oñate E., Del Pin F., The particle finite element
method: a powerful tool to solve incompressible flows with free-
surfaces and breaking waves, IJNME (2004)
Oñate E, Idelsohn SR, Del Pin F, Aubry R. The particle finite
element method. An overview. International Journal of
Computational Methods 2004; 1(2):267 307.
See e.g. Cremonesi et al., Arch. Of Comput. Methods in Eng.
2020, for a recent overview.
94. PFEM: Examples
59
Dam break Problem set
Physical parameters
Numerical parameters
PSPG+Picard
Pressure sensor location
97. PFEM
CUPyDO
(Python)
(C++)
(C++)
(C++)
Implementation: coupling codes through Python
Communications are
performed through
memory (No I/O files)
No full execution of
coupled codes
No system calls
SWIG: Simplified Wrapper and Interface Generator, http://www.swig.org
62
Multi-Physics/Multi-scale: coupling different codes
106. 71
Motivation: numerical simulation of weld pools
multi-physics simulations at meso-scale
Solid
Liquid
Solid
Liquid
Heat source (e.g. Laser LPBF, laser welding )
Heat source
Heat transfer
Melt pool fluid dynamics
Melting & solidification
Residual stresses
and distortion after cooling
transient & coupled
unknown & evolving
Interfaces
107. Equations to solve:
Liquid, mushy & solid regions solved using same mesh using:
Navier-Stokes equations (NSE) - Lagrangian form
Heat equation
72
Latent heat
absorption
flow resistance
of mushy zone
(and solid)
Surface tension/
Marangoni terms
108. Equations to solve
Momentum equation for newtonian fluids
Surface tension
Resulting surface force term
73
Normal force Tangential Marangoni force
Note:
Marangoni coefficient
usually negative
surface force drives
fluid away from heat
source
= curvature
= outward normal
= surface tension at
= Marangoni coefficient
Surface tension term
120. Conclusions
PFEM handles well
free surface deformation
Thermo-mechanical problems
phase change with latent heat
surface tension & Marangoni effect
adaptive mesh refinement
Fluid-Structure interactions
Still a lot of work till realistic SLM (Selective
on the way!
85