METNET Workshop

October 11-12, 2009, Poznań, Poland

Experimental and numerical analysis

of sandwich metal panels
Zbigniew Pozorski, Monika Chuda-Kowalska, Robert Studziński, Andrzej Garstecki

Poznan University of Technology, Institute of Structural Engineering, Poznan, Poland

OUTLINE
Aim of the study
Experimental determination of mechanical parameters
Local buckling (wrinkling)
Failure maps
Optimization of sandwich panels
1
AIM OF THE STUDY
1) Proper description of structural behaviour of sandwich panels
(non-homogeneous, anisotropic core, profiling of metal faces,
non-uniform boundary conditions)
2) Precise analysis and prediction of failure mechanisms

3) Optimal design
(minimum cost, maximum span, maximum load capacity)

2
 EXPERIMENTAL DETERMINATION
OF MECHANICAL PARAMETERS

Reliable methods of determination modulus GC of the core are

still expected by producers and designers:
- GC strongly influences the response of the panel
- methods suggested by code 14509 provide results depending
on the method and specimen’s scale

Tasks:
- verification of methods of identification the Kirchhoff modulus
GC proposed by code 14509
- improvement of identification methods
3
 CLASSICAL METHODS
1. Tests on short/long panels (w)

1.6 F [kN]

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0
0 2 4 6 8 10 12 14 ws [mm]

w = wB + wS , (1) where:
wB, wS – bending and shear deflection
23 ⋅ ∆F ⋅ L3 L, B – span and width of the panel
wB = , (2)
1296 ⋅ BS dC – depth of the core
BS – flexural rigidity
∆F ⋅ L
GC = . (3)
6 ⋅ B ⋅ d C ⋅ ∆wS 4
 NON--CLASSICAL METHODS
NON
1. Measurement of angles of rotation (γγ)

5
 Assessment of the shear modulus

where:
γ 01, γ 02 – total slope of deflection line
α0 – the angle of rotation of the cross-section
γ – the angle of rotation of the core

γ 0 = 0,5 ⋅ (γ 01 + γ 02 ), (4)

γ = γ 0 −α0 , (5)

V
GC = (6) where:
γ ⋅ B ⋅ dC V – the shear force

6
2. Shear test (γγ)
The force F acts on the middle plate and the vertical displacement w of
the rigid plate is measured

1 – rigid steel plate

(10x240x550mm)
2 – sandwich panel
(100x240x500mm)

w
γ= , (7)
dC

F (8)
GC = .
2 ⋅γ ⋅ B ⋅ L 7
3. Torsion test (ϕ
ϕ)
MS
ϕ'= , (9)
GC ⋅ I 0
ϕ
ϕ'= , (10)
L

π ⋅ D14
I0 = . (11)
32

where:
MS – torsion moment
ϕ – angle of specimen rotation
L – length of the sample
D1 – diameter of the cylinder cross-section
I0 – central second order moment of the
area of the cylinder cross-section
8
Discussion of the results
Experimental results of GC from methods proposed by code 14509
Type of Width of the GC δ
sample sample [MPa] [%]
Short panel
B=0.1m 3.33
L=0.6m
Long panel
L=4.9m
Total width 4.87 46.2

Experimental results of GC from different bending methods

Method of GC Influence of longitudinal

δ
Type of panel identification edge profiling on test results
[MPa] [%]
longitudinal Measured w 4.87
edge profiling
exists Measured γ 4.70 3.6 30.6
longitudinal Measured w 3.73
30.9
edge profiling
cut off Measured γ 3.59 3.9 9
LOCAL BUCKLING (WRINKLING)
WRINKLING)

wrinkling

Tasks:
- numerical modelling of structural
response representing
global and local effects
- analysis of progressive damage,
contact effects, wrinkling
- assessment of the influence of
discontinuity of the core
10
 LABORATORY TESTS
- supporting system
- load cell HBM C 6A 200kN 0.5 class
- displacement transducers HBM WA L 100mm
- tensometers HBM LY 10 mm
- HBM catman 4.5 software

11
 NUMERICAL SIMULATIONS

1. Class of problems

Core: PU (isotropic, homogeneous, continuous)

Mineral wool (anisotropic, macro-homogeneous, non-continuous)
Profiling of faces: flat, micro-profiled, deep-profiled
Boundary conditions: transversal loads, thermal actions

12
2. Bending of flat (or micro-profiled, deep-profiled) panels
Lo
S4R

D
COH3D8
C3D8R
q

S4R

L RIKS method

Material parameters:

Faces: EF=210 GPa, νF=0.3, fy=270 MPa (elastic, ideal plastic)

Core (MW): EC=12 MPa, νC=0.05, GC=5.714 MPa (isotropic elastic)
Et= E3=14 MPa, Ep= E1= E2 =3 MPa (ortotrophy)
Gp= G12=1.43 MPa, Gt= G13= G23 =6.36 MPa
νpt=ν13=ν23=0.03, νtp=ν31=ν32=0.14
Interface: Knn=12 MPa, Kss=Ktt=6 MPa (cohesive) 13
Damage modelling in the interface layer
• Elasticity (uncoupled) for cohesive elements
t n   K nn 0 0  ε n  tn – normal traction (stress)
t  =  0 K ss 0  ε s  ts , tt – shear tractions
 s  ε - corresponding nominal strain
 tt   0 0 K tt   ε t 

• Damage initiation 2 2 2
tno = 140 kPa
 tn   t s   tt 
- quadratic nominal stress  o  +  o  +  o  =1 tso = 100 kPa
 tn  t s  tt  tto = 100 kPa
• Damage evolution
Type: Displacement, Softening: Linear

• Contact in the core (between lamellas)

- hard contact
- exponential
- Coulomb friction model

14
Example 1. Core: isotropic, continuous

q = 2.205 kN/m2
faces: σ11= -234.2 / +102.7 MPa
deflection: u = 2.954 cm

15
Example 2. Core (mineral wool): orthotropic, non-continuous,
„hard” contact

„hard” contact
q = 0.521 kN/m2
faces: σ11= -34.24 / +22.81 MPa
deflection: u = 0.6037 cm

16
Example 2a. Core: orthotropic, non-continuous

COPEN – clearance between surfaces

CSLIPCPRESS – contact
– tangential motionpressure
of the
surfaces during contact

17
Example 3. Non-uniform loading, continuous core

q = 5.285 kN/m2
faces: σ11= -122.7 / +54.40 MPa
deflection: u = 1.659 cm

Discussion of the results

18
 FAILURE MAPS
Failure maps specify the failure modes for different combinations of the design
parameters.

Groups of interrelations:
thickness of the facings tFi vs thickness of the core d,
shear modulus GC vs d,
shear modulus GC vs stiffness of the external (k1) or internal (k2) support.

The isolines: the maximal allowable load p(x) or temperature difference ∆T=T2−T1
for various combinations of the analysed parameters.
19
 Failure maps – examples
Mechanical load p(x)=p [kN/m]
Isolines of p for variable GC and stiffness of the external (a) and internal (b) support

Comments:
the response of structure with varied internal support stiffness is different from that
with varied external support stiffness
apparently the limits of the failure modes converge with the maximal capacity of the
panel
the general optimal capacity level is nearly the same and the sensitivity of the
capacity of the sandwich panel is much less than in the case of the thermal actions 20
 Failure maps – examples
Thermal load ∆T [ C]
Isolines of ∆T for variable GC and stiffness of the external (a) and internal (b) support

Comments:
large sensitivity of the panel capacity along the limits of the failure modes
elastic supports leads to significant improvement of the capacity of the sandwich panel
(40% - 60%)
quite opposite results are observed in the case of mechanically loaded panels
21
 OPTIMIZATION
Motivation for optimal design
minimal
core variance in
failure types of
panels

maximal
facing range of
failure application

minimum
deflection cost

22
 Optimization – Problems formulation

I. Optimization of the material and geometrical parameters

design vector: two criterion fitness function: constraints:

s = [t1, t2, D, GC, L]
minimum cost FC(s)
ULS and SLS

maximum length
box conditions
of the span L

II. Optimization of the support conditions

design vector: two criterion fitness function: constraints:
s = [kj, δj, L], j=1,.., n
maximum load
ULS and SLS
where j is the support number multiplier λ
box conditions

maximum span L

The problems are nonconvex hence Distributed Parallel Evolutionary Algorithm

was used
23
 Optimization – Example
Influence of temperature and load intensity on cost function

24
 Optimization – Example
Elastic support with a gap, thermal action ∆T [ C]
Design parameters: Measure of the quality of the structure:
k1 - stiffness of the external support λ - load multiplier of the thermal load ∆T,
δ2– - gap in the internal support where λ =1 means maximal capacity of the
L - span length thermally loaded panel on rigid supports

25
 PART 2
A.Garstecki

Research activity in Steel Structures

Prof. M. Szumigała, Dr. K. Rzeszut, Dipl.Eng. M. Chybiński

(Division of Metal Structures)

1. Statical and stability analysis of lattice
thin-walled structures
2. Composite steel-concrete beams
Motivation

• Wide implementation in civil engineering

• Nonlinear behaviour due to deformation of
initial contour and geometrical
imperfections
Tasks

• Buckling and postbuckling analysis of

columns and beams made of single Σ and
double Σ cross-section
• Sensitivity to imperfections and clearances
• Optimal configuration of ribs in I beams
• Moment-curvature relation and limit curves
for steel-concrete composite beams
1. K. Rzeszut, Nonlinear stability analyses of thin-walled structures

Equilibrium equations:
F(P, U) = 0, (1)
where F is a nonlinear differential operator, P is a load vector and U denotes a
displacement vector. It can be written in incremental form

(K o + K σ + K U )∆U = ∆P, (2)

where:
KO - small-displacement stiffness matrix
Kσ - initial stress matrix
KU - load stiffness matrix.
Eq. 3 accounts for all nonlinearities of the problem.
Riks method is used to solve eq. 3 for both stable and unstable postbuckling
behaviour.
 GEOMETRIC IMPERFECTIONS
Postbuckling analysis by introducing a geometric imperfection
pattern to the “perfect” geometry.
Lowest buckling modes are scaled and added to the perfect
geometry to create the perturbed mesh.
Imperections have the form:
( K o + λK G ) U = 0 (3)
T
~ =[u~ ] = ∑αU  = UTα
u≅u
n
where: r i=1 i ir  (4)
Uir- the buckling mode, from measurements
of real structures
αi- the associated scale factor,
n - the number of eigenmodes used to create the perturbed mesh.
 Initial imperfections and clearances in stability analysis
Asymmetrical, unstable bifurcation point
a) η= 0.2, b) η= 1
P 2,0 P 3,2
3,0
P cr 1,8 P cr 2,8
1,6 2,6
1a
2,4
1,4 1a
1b 2,2
1b
1,2 2,0
2
1,8 2
1,0 1,6
2a 2a
0,8 1,4
2b 2b
1,2
0,6 2c
2c 1,0
0,8
0,4
0,6
0,2 0,4

0,0 ∆ = ∆/L 0,2 ∆ = ∆/ L

0,0
-0,60 -0,40 -0,20 0,00 0,20 0,40 0,60
-0,60 -0,40 -0,20 0,00 0,20 0,40 0,60

Where η=k1/k2 - non-dimensional stiffness coefficient

Connectors of cold formed double SIGMA elements

A B

“Link” - preserves only the constant distance between points A and B.

“Tie” - the rotations are constrained.
“Slot” - accounts for the clearances in the connection.
 Σ
MODELLING OF CONNECTIONS OF 2Σ

CONNECTOR
CONNECTOR TYPE

TRANSLATION
ROTATION

Different spacing l1 (l1=0.5m, l1=1.0m, l1=l/2, l1=l and l1=0) and

different numerical models of connectors was analysed.
 Non-linear stability analysis of a double Σ beam
λ/λcr
1

0,9

0,8
1l
0,7
1g
0,6 2l
0,5 2g
3l
0,4
3g
0,3

0,2

0,1

0
0 2,5 7,5 12,5 17,5 22,5
Displacement parameter

Load proportionality factor of column made of double ∑ cross section for

global (g), local (l) shapes of imperfections
Stability of laterally braced purlins

Fig 3. Shapes of buckling mode for purlins:

a) without bracings, b) with two bracings.
2. M. Chybiński, Optimal configuration of ribs in welded girders
 3. M. Szumigała, Composite steel-
steel-concrete
elements
Ciêgna
Badany slup
240

Silownik S1
2250 H/2
2250 1500 H/2
240

H/2 H/2
Silomierze S2
6000

1500
P
12

12
P

S3 S3
Silomierz
H H

240
216
240
216

Badany slup

12 6000
12

240 240
Limit curves M-
M-N for various numerical models

M-N (porównania)
1600000
1
1400000 2
4
1200000 5
9
1000000 12
Standard-Riks
800000 Warstwowy
M

600000

400000

200000

0
0 50000 100000 150000 200000 250000
N
Conclusions
• Small clearances influence the stability response in a
similar way as small imperfections.

•Thick head plates in girders improve torsional stiffness,

however usual habits are too conservative. Thick head
plates decrease rotational capacity of bolted joint.

• Diagonal ribs in welded girders are definitely better than

orthogonal ones (in linear and non-linear regimes).

• In steal concrete composite structures the normal force

drastically influences the moment-curvature relation.
Poznan University of Technology
Institute of Structural Engineering

Last information on the activity

Students’ MSc diploma works made for Rautaruukki

Corporation in 2008 and 2010.

Thank you
for your attention