RESEARCH ARTICLE | MARCH 01 2019

Selected aspects of numerical modeling of the short span

thin-walled beams 
Robert Studziński  ; Katarzyna Ciesielczyk

AIP Conf. Proc. 2078, 020001 (2019)

https://doi.org/10.1063/1.5092004


View Export
Online Citation

18 April 2024 12:16:29

 Selected Aspects of Numerical Modeling of the Short Span
Thin-Walled Beams
Robert Studziński a) and Katarzyna Ciesielczyk b)

Poznan University of Technology, Institute of Structural Engineering, Piotrowo 5 Street, 60-965 Poznań, Poland.
a)
Corresponding author: robert.studzinski@put.poznan.pl
b)
katarzyna.ciesielczyk@put.poznan.pl

Abstract. This paper concentrates on selected aspects of numerical modeling of a short span thin-walled beams. The
performed numerical simulations cover different types of finite elements (shell or solid finite elements) and different
mesh sizes. Moreover the shape and magnitude of an initial geometrical imperfection is also investigated. The numerical
analysis were made with the Newton-Raphson procedure with geometrical and material nonlinearity. The numerical
results were compared with the laboratory four point bending test of a thin-walled beam. Furthermore, some guidelines
for numerical simulations of a short span thin-walled beams in bending were proposed by authors.

18 April 2024 12:16:29

INTRODUCTION
The cold formed thin-walled elements are commonly used in many structural engineering applications. For
example thin-walled beams are use as both the main and the secondary structural elements of single- or multi-storey
steel structures. The wide range of application of these elements results from their high load bearing capacity
coupled with small weight. The minimization of the weight is mainly realized by the reduction of a thickness of
walls of a cross section. It leads to the increase of both the slenderness of the walls and of the element itself.
Therefore, the capacity of the thin-walled beams is usually limited by the buckling phenomena. In the case of the
thin-walled beams there are three basic buckling modes namely local, distortional and global (flexural or flexural-
torsional), see Schafer 2002. The local buckling mode is governed by the slenderness of the cross section walls i.e.
c/t ratio, where c and t refer to width and thickness of the cross section walls respectively. The distortional buckling
mode is mainly governed by the slenderness of the cross section i.e. h/b ratio , where h and b refer to height and
width of the cross section respectively. The global buckling mode is governed by the slenderness of the element i.e.
ratio L/imin, where L and imin refer to span length of the element and radius of gyration of the cross section
respectively.
The theoretical background concerning the phenomenon of loss of stability of thin-walled beams was set-up by
Vlasov 1961 and Timoshenko and Gere 1961 and later developed by Bazant and Nimeiri 1973. Please note that
shape and magnitude of the loss of the stability depends inter alia on the initial geometrical imperfections, the
residual stresses (Moen et al. 2008, Schafer et al. 2008) and the boundary conditions. The way of allowing for
geometrical imperfections in the designing of the thin-walled beams was discussed for example by Dubina et al.
2001. In this paper the method of arbitrary scaling of the eigenvalues of the elastic buckling in order to allow for
geometrical imperfections was proposed. This approach was further developed by Lechner and Pircher 2005 by
introducing the linear superposition of the eigenvalues.
The aspects of numerical modeling of thin-walled elements have been widely discussed in subject literature. For
example a broad overview of computational models which refers to elastic buckling and nonlinear collapse was presented
by Schafer et. al 2010. Moreover in this paper the importance of imperfections, residual stresses, material models,
boundary conditions and element choice in numerical modeling of thin-walled elements has been also addressed. In turn,
Jachimowicz et al. 2008 presented the Finite Elements (FE) simulation results of thin-walled beams with rivet connections

Computational Technologies in Engineering (TKI’2018)

AIP Conf. Proc. 2078, 020001-1–020001-7; https://doi.org/10.1063/1.5092004
Published by AIP Publishing. 978-0-7354-1806-6/$30.00

020001-1
in the tension zone of the cross section. Similar problem was addressed by Pařenica et al. 2017 where, the accuracy of the
numerical modeling of the connection between the thin-walled element and supports was investigated.

PROBLEM FORMULATION
The main aim of the paper is the convergence analysis of numerical models of short span thin-walled beams,
which usually fail due to the local buckling coupled with plastic deformation. The term “short beams” is used for
beams which slenderness meet the following condition L/imin < 100. The numerical simulations were preceded by
laboratory four point bending tests of Z cross section of thin-walled beams, see Fig. 1.

(a) (b)

18 April 2024 12:16:29

FIGURE 1. Four point bending test: a) dimensions of a Z cross section, b) static scheme of a four point bending test

The convergence analysis of the numerical model of the short span thin-walled beam comprises not only the mesh
size and the type of a finite element (shell vs. solid) but also the type of a shape function (linear vs. quadratic), the
number of nodes [4, 8, 20] and the number of integration points. All finite element analyses were carried out in Abaqus
CEA program. The set of parameters included in the convergence analyses is presented in Table 1.
TABLE 1. The set of the parameters included in convergence analysis
Task Type of the FE Name of the FE Mesh size [mm]
Z-01 S4 2 / 5/ 10 / 15 / 20
Z-02 shell S4R 2 / 5/ 10 / 15 / 20
Z-03 S8R 2 / 5/ 10 / 15 / 20
Z-04 C3D8 2 / 5/ 10 / 15 / 20
Z-05 C3D8R 2 / 5/ 10 / 15 / 20
brick (solid)
Z-06 C3D20 2 / 5/ 10 / 15 / 20
Z-07 C3D20R 2 / 5/ 10 / 15 / 20
Z-08 tetrahedron (solid) C3D4 2 / 5/ 10 / 15 / 20

In order to obtain the nonlinear static response, the Newton-Raphson procedure was implemented. The initial
geometrical imperfections were obtained from linear buckling analysis (LBA) by an arbitrary scaling (multiplier one
and five) of selected buckling modes (first local and first interactive). The method of allowing for geometrical
imperfection in the designing of the thin-walled beams was discussed by Dubina et al. 2001. The local buckling
of a thin-walled element is depicted in Fig. 2a and 2b, while in Fig. 2c the shape of scaled buckling mode applied in the
analysis is presented.

020001-2
 (a) (b) (c)
FIGURE 2. The four point bending test of a thin-walled beam of a Z cross section: a) isometric view of a local bucking
of a thin-walled beam, b) cross section view from laboratory, c) applied and scaled buckling mode

LABORATORY TESTS RESULTS

The static scheme of the four point bending laboratory test is depicted in Fig. 1b. The following mechanical
parameters of the material of the thin-walled beam were obtained from the static tensile test: E = 198.7 GPa,
yield = 336.9 MPa. The measured mean thickness of the Z section walls equals t = 1.492 mm. The thin-walled
beams under investigation were of a 0.93 m span length (the total length was 1.1 m). The fork support condition
were assumed at both end supports. The velocity of the traverse beam was 10 mm/min. The load was applied at the
upper flange of the cross section. The statistical sample consisted of the eight trials. During the test the force and the
displacement of the traverse beam was measured. The obtained kinematical response of the thin-walled short beam
under four point bending was nonlinear, see the load paths depicted in Fig. 3.

18 April 2024 12:16:29

C
B

FIGURE 3. Load paths of a four point bending of a thin-walled Z beam

Please note, that the obtained load paths can be divided into three stages: stage A, stage B and stage C. Stage A
represents the adjustment and the elimination of the structural clearances between the elements of the test bed. From
the engineering point of view stage A can be neglected, therefore in further consideration will be omitted. Stage B
represents the elastic-plastic behavior of thin-walled element, while stage C represents the release phase i.e. stage
after the failure of the short span thin-walled beam. The obtained consistency of the results is acceptable. The
detailed results from experiment are presented in Table 2 and are illustrated in Fig. 4. In Table 2 the elastic stiffness,
and the secant stiffness of the beam is represents by keL and ksec respectively. The elastic stiffness of the thin-walled
beam is determined in the manner of the least square regression line y = ax + b where a and b are given by Eq. 1 and
Eq. 2 respectively.

n i 1 x i y i   i 1 x i  i 1 y i
n n n

a (1)
 x 
2
n i 1 x i2 
n n
i 1 i

020001-3
 1 n n

b   y i  a xi  (2)
n  i 1 i 1 
The secant stiffness was determined in accordance to the Eq. (3), where Fa = 2.0 kN, Fmax represents the ultimate
force, umax represents the displacement which refers to ultimate force and ua represents the displacement which
refers to Fa.
F Fmax  Fa
k sec   (3)
u u max  u a
TABLE 2. Data from the four point bending of a thin-walled short span beam
Name unit 1 2 3 4 5 6 7 8 x s.d.
keL [kN/m] 1840 1880 1740 1850 1860 1890 2030 1960 1880 86
ksec [kN/m] 1350 1410 1270 1360 1350 1410 1380 1530 1381 74
FEd,max [kN] 10.8 10.7 11.5 11.1 12.1 10.0 11.1 9.1 10.8 0.9
umax [mm] 12.5 10.6 11.7 12.4 13.0 11.2 13.8 10.5 11.9 1.2

(a) (b) (c) (d)

FIGURE 4. Results from four point bending test of a short span thin-walled beam: a) elastic stiffness keL, b) secant stiffness ksec,
c) ultimate load FEd,max, d) displacement at the ultimate load umax

18 April 2024 12:16:29

One can observe that short span thin-walled beams fail due to local buckling of a web coupled with plastic
deformations of the upper flange. The evolution of the failure is presented in Fig. 5. The failure process was initiated
by local buckling of a web, see Fig. 5a. Please note that, while the load increases the local buckling amplitude grows
and plastic deformations appear, see Fig. 5b and c. Moreover the tensioned flange (bottom) also is deformed, mainly
due to both the large buckling deformations of the web and the plastic deformations of the upper flange.

(a) (b) (c)

FIGURE 5. The evolution of the failure of the short span thin-walled beam: a) local buckling of the web, b) local buckling
of the web and small plastic deformations of the upper flange, c) local buckling of the web and large plastic deformations

NUMERICAL SIMULATIONS
For the purpose of FE simulations, it was assumed that the material of the thin-walled element is homogeneous
and isotropic. The material model was assumed as elastic-plastic. The Young modulus and the yield strength were
taken from static tensile test. The list of the considered FE models in convergence analysis is presented in Table 1.
The basic criterion for differentiating numerical simulations is the type of finite element i.e. shell (S4, S4R and S8R)
and solid (C3D8, C3D8R, C3D20, C3D20R and C3D4).

020001-4
 The linear shell finite elements the S4 and the S4R are 4-node, linear and of a general purpose. Both elements are
applicable for finite membrane strains, nevertheless the S4R elements are characterized by uniformly reduced
integration which allows for avoiding shear and membrane locking. Moreover, the S4R due to hourglass modes
allows for propagation over the mesh. Thus, the S4R converges to shear flexible theory for thick shells and classical
theory for thin shells. In convergence analysis also quadratic with reduced integration shell finite element (S8R) was
considered. The S8R is a 8-node doubly curved thick shell finite element with reduced integration. This element
should to be used for thicknesses which are larger than 1/15 of a characteristic length on the surface of the shell. In
the convergence analysis also brick and tetrahedron solid 3D finite elements were used. The applied brick finite
elements were linear or quadratic with or without reduced integration. The C3D8 element is a general purpose linear
brick element, fully integrated. This element tend to be stiff in bending. The C3D8R element is a general purpose
linear brick element, with reduced integration. Note that the shape functions of the C3D8R are the same as for the
C3D8. This element, due to reduced integration, is not sensitive for shear locking but shows tendency to be not stiff
enough. The C3D20 element is a general purpose quadratic brick element. In the case of bending this element tends
to be too stiff. The C3D20R element is a general purpose quadratic brick element, with reduced integration. This
finite element only causes problems in contact calculations, but it is a general condition of all quadratic finite
elements. The convergence analysis also included the tetrahedron finite element C3D4. The C3D4 is a general
purpose tetrahedral element (1 integration point). In structural calculations this element usually is too stiff.
Note that the group of tasks Z-01 to Z-03 and Z-04 to Z-08 refers to shell and solid finite elements respectively.
Within each group the five different mesh sizes were investigated namely: 1 – 20 mm, 2 – 15 mm, 3 – 10 mm, 4 –
5 mm and 5 – 2 mm. Additionally for the FE models with 15 mm mesh size (case B) two geometrical imperfections
were separately introduced. The shapes of the initial geometrical imperfections were obtained from the linear
buckling analyses. The selected shapes of the initial geometrical imperfections refer to first local and first distortional
buckling mode. The selected modes were introduced and once and five times rescaled in the right one FE model.

DISCUSSION OF THE RESULTS

18 April 2024 12:16:29

In Fig. 6 the results obtained from the use of shell finite elements in numerical models are compared with the
results from the laboratory experiments. In the case of short span thin-walled beams the numerical simulations with
the shell finite elements lead to the following conclusions:
 compacting the finite element mesh makes the numerical model too stiff (see the results with the mesh 2 mm
and 5 mm in size),
 the introduction of both the geometrical imperfections with the shape obtained from the local buckling mode
and the geometrical imperfections with the shape obtained from the interactive buckling mode (local with
global) insignificantly reduces the stiffness of the short span thin-walled beam,
 the numerical models with the S8R finite elements are generally stiffer than the numerical models with the
S4 and S4R finite elements,
 the best convergence of the shell finite element model with the laboratory results is obtained for the S4R
finite elements.
In Fig. 7 the results obtained from the numerical models with the use of the brick finite elements are compared
with the results from the laboratory experiments and the following conclusions can be formulated:
 similarly to shell finite element models, compacting the solid finite element mesh makes the numerical
model too stiff (see the results with the mesh 2 mm and 5 mm in size),
 the introduction of both the geometrical imperfections with the shape obtained from the local buckling mode
and the geometrical imperfections with the shape obtained from the interactive buckling mode (local with
global) slightly reduces the stiffness of the short span thin-walled beam,
 the best convergence with the laboratory equilibrium paths is obtained for the numerical models with the
C3D8 finite elements.
In Figure 8 the results obtained from the numerical models with the use of the tetrahedron finite elements are
compared with the results from the laboratory experiments and the following conclusions can be formulated:
 all considered mesh sizes of the tetrahedron finite element make the numerical model too stiff,
 the introduction of the geometrical imperfections with the shape obtained from the global buckling mode
does not reduce the stiffness of the short span thin-walled beam,
 this element gives the worst convergence with the laboratory results.

020001-5
FIGURE 6. Equilibrium paths of the FE models vs. laboratory experiment: a) S4 shell FE, b) S4R shell FE, c) S8R shell FE

18 April 2024 12:16:29

FIGURE 7. Equilibrium paths of the FE models vs. laboratory experiment: a) C3D8 brick FE, b) C3D8R brick FE,
c) C3D20 brick FE, d) C3D20R brick FE

020001-6
 FIGURE 8. Equilibrium paths of the FE models vs. laboratory experiment: C3D3 tetrahedron solid FE

ACKNOWLEDGMENTS
The research was financially supported by Poznan University of Technology Grant no. 01/11/DSPB/0006.

REFERENCES
1. B. W. Schafer, J. Struct. Eng.-ASCE 128, 3 (2002).
2. V. Z. Vlasov, Thin-walled elastic beam, English translation (Published for NSF and Department of
Commerce), 1961.
3. J. M. Gere and S. P. Timoshenko, Mechanics of materials (PWS-KENT Pub. Co. Boston).
4. Z. P. Bazant and M. Nimeiri, J. Eng. Mech.-ASCE 6, 1259–1281 (1973).

18 April 2024 12:16:29

5. C. D. Moen, T. Igusa, and B. W. Schafer, “Prediction of residual stresses and strains in cold-formed steel
members”, Thin-Wall. Struct. 46, 11, 1274–1289 (2008).
6. D. Dubina, V. Ungureanu and I. Szabo, “Codification of imperfections for advanced finite analysis of cold-
formed steel members”, in Proceedings of the 3rd ICTWS 2001, pp. 179–186.
7. B. Lechner and M. Pircher, Thin-Wall. Struct., 43, 361–374 (2005).
8. B. W. Schafer, Z. Lia, and C. D. Moen, Thin-Wall. Struct. 48, 10-11, 752–762 (2010).
9. J. Jachimowicz, J. Mańkowski, and J. Osiński, J. of KONES, 15, 1, (2008).
10. P. Pařenica, M. Rosmanita, and J. Flodra, Procedia Engineer., 186–192 (2017).

020001-7