TECHNICAL TRANSACTIONS 2/2019

ENVIRONMENTAL ENGINEERING
DOI: 10.4467/2353737XCT.19.027.10163
SUBMISSION OF THE FINAL VERSION: 15/01/2019

Dorota Anielska orcid.org/0000-0002-0368-6138

[email protected]
Department of Geoengineering and Engineering Structures, Faculty of Environmental
Engineering, Cracow University of Technology

Finite element analysis of reinforced concrete elements

subjected to torsion
Analiza skręcania elementów żelbetowych metodą
elementów skończonych

Abstract
This paper presents FEM techniques used for modelling concrete elements subjected to torsion. It compares
the results from a 3D numerical analysis and a numerical homogenization method analysis. Finally, the
results are compared to the reported experimental data.
Keywords: beam, torsion, homogenisation, FEM

Streszczenie
W artykule przedstawiono sposoby modelowania skręcanych elementów żelbetowych za pomocą Metody
Elementów Skończonych. Zaprezentowano porównanie wyników otrzymanych w trójwymiarowej analizie
numerycznej oraz w analizie numerycznej bazującej na teorii homogenizacji. Wyniki zestawiono z wynika-
mi eksperymentalnymi znanymi z literatury.
Słowa kluczowe: belka, skręcanie, homogenizacja, MES

129
 1.  Introduction

In a basic case, reinforced concrete structures are subjected simultaneously to axial forces (N),
shear forces (V), bending moments (M) and twisting moments (T). In engineering practice most
of these states can be analysed independently, or by taking into account the effect of a secondary
value on the principal value, such as considering the effect of shearing in bending analyses, or in an
interactive manner, as is the case in analysing eccentrically compressed sections or in analysing the
combined effect of torsion and compression or the effect of torsion and shear on the cross-section
capacity. In common practice a maximum of two internal forces are considered in such analyses.
In the design of sections subjected to combined bending and torsion, the strength analysis
is carried out separately for bending and torsion, as if the cross-section was subjected to pure
bending or pure torsion at one time. The procedures prescribed by the standards do not address
the combined effect of bending and torsion and assign the effect of bending to the longitudinal
reinforcement and the effect of torsion to the transverse reinforcement on exclusive basis.
Although very practical, this approach, which separates the bending and torsion effects
does not represent the actual behaviour of beam elements in reinforced concrete structures.
This concerns, in particular, the outermost beams, such as spandrel beams in column-and-
slab structures for which the torsional moment has a considerable effect.
Experimental studies of elements subjected to combined bending and torsion showed
that with an increasing share of the bending moment in relation to the torsional moment,
bending starts to govern the behaviour of the reinforced concrete element under loading.
Bending has a predominant effect not only on the strength and deflections but also on the
stress and strain state of the element. The situation is not so clear in the case of increasing the
share of torsion and, as a result, it is common practice to design additional reinforcement to
resist the entire torsion with no participation of the designed bending reinforcement.
However, before examining the interaction of two or more internal forces, it is important
to investigate the behaviour of a section subjected to pure torsion so as to eliminate
misrepresentation of results due to the effect of bending.
The following part of this paper presents a numerical approach to the problem of pure
torsion using non-linear models of concrete (Concrete Plastic Damage and Menétrey
plasticity model with softening behaviour). The results obtained in this way are compared
with the experimental data and with the provisions of the relevant standards.

2.  Numerical analysis

Two numerical modelling approaches are compared in this paper. The behaviour of the
reinforced concrete beam subjected to torsion in the experimental study whose results are
reported in [3] is represented in each case. The original dimensions were given in Imperial
units. They have been converted to SI and rounded. The rounding takes into account the
average dimensions of all elements subjected to the experimental test and does not affect the
results of numerical calculations.

130
 Fig. 1. Schematic representation of beam No. B11 – longitudinal and transverse sections

Fig. 2. Schematic representation of beam No. B21 – longitudinal and transverse sections

Fig. 3. Boundary conditions

In the first approach, a numerical model is created to represent all the geometric parameters
of the beam under analysis, including the boundary conditions and torsion is applied by force
control, i.e. the application of a pair of forces. In the second approach, the model is created for
a section of a beam subjected to torque without restraint, i.e. with the warping free boundary
conditions, and the effect of torsion is obtained by deformation-controlled action.
The ZSoil®.PC v2018 software was used in the analysis, with the Concrete Plastic Damage
constitutive model described in [5] and, for comparison, the Menétrey plasticity model with
softening behaviour given in [4].
The first approach seems the obvious choice and is available to any user of advanced FEM-
based programs. On the other hand, for obvious reasons it is also non-economical due to
the labour-intensive preparation of data and time-consuming calculations and processing of
results.
In the second approach, representation of the beam behaviour is limited to the middle
portion where the torsional moment is constant. The analysed portion of the beam has
a length equal to the distance between the stirrups and includes a centrally located stirrup.
Periodic boundary conditions are imposed on the cross-sectional surfaces to enforce the
same nodal displacement in the x, y and z directions. Moreover, six independent boundary
conditions are imposed in the analysed beam section to constrain translation and rotation.
The torsion effect is obtained by applying a macro-strain field in x-y and x-z directions.

131
 Fig. 4. 3D model of beam in ZSoil

Fig. 5. Beam model obtained with the homogenisation approach – a continuum model, system of bars,
periodic boundary conditions

The boundary conditions no longer play a role in this approach and the computation time
is reduced.
In numerical analyses of objects, as in the first approach, considerable perturbations can
occur at the shear zones due to stress concentrations. In order to avoid their dominating effect
a number of measures must be implemented, including, without limitation, use of different
material properties or different material models. In the analysed case, the measures used at
supports include application of an elastic constitutive model for concrete, an increased amount
of longitudinal reinforcement, and smaller distances between stirrups (as described in [3]).
However, perturbations were still found at the joint between continuous elements made of
Elastic and Concrete Plastic Damage materials, waning away no sooner than about the beam
midspan. Application of load was yet another challenge. Thus, in order to avoid the concentrated
load effect the load was applied through a membrane. The natural consequence of this approach
is the long time required for creating the model and for carrying out the calculations.

132
 The second approach is based on homogenisation theory. Similarly to the homogenisa-
tion theory described in [6], it considers a 3D element that is loaded, in the general case,

 
T
with all the six internal force componentsΣ N , M x , M y , M z
,Q x 0,
Q y 0 , (i.e. lon-
gitudinal force, two bending moment components, torsional moment and two transverse
force components) generating generalised strains (macro-strains) that describe the kine-

 
T
matics E  E0 , K x , K y ,  z ,  x ,  y . In the analysed case, in which the analysed element
is subjected solely to pure torsion, the vector of strain control macro-strains E takes the form

0,0,0,  z ,0,0
T
Ε . In the analysed linear element a repeatable 3D unit (periodic cell) is
.
distinguished of finite length which in the numerical solution stage does not go to zero at the
boundary, which differentiates the current solution from the cross sectional analysis of the
beam presented in [7].
The total strain field comprises two parts: macro-strains used for deformation-controlled
action and strains caused by perturbation of the displacement field up(x) on which the
periodic boundary conditions are imposed.

ε( x)= ε E ( x )  ε p  x  = L E (x)E  Bu p (x), (1)

u( x)= u E (x)  u p (x)= C E (x)E  u p ( x ). (2)

where:
 εxx
E

 E  0 0 0 0 0 0  0 
 ε yy  0 0 0 0 0 0  0 
 E    
 γ xy  0 0 0 0 0 0   0 

E
ε =  E =  = L (x)E , (3)
 εzz  1 y x 0 0 0   z  E
γ E  0 0 0 y 1 0  0 
 xz     
γ Eyz  0 0 0 x 0 1   0 
 

0 
0 1 2 
 0 z  yz z 0   0 
2  
  0 
1  
u E = 0  z 2 0 xz 0 z     = C E (x )E (4)
 2   z 
z yz xz 0 0 0  0 
   
  0
 
and B is the differential operator matrix obtained from the Cauchy-Riemann equations.

133
 This approach is a particular case of application of the method described in [6].
In the considered case of pure torsion  xz E   yz and  yz E  xz .
In this method, the numerical analysis considers a completely consistent 3D stress/ strain
state for which geometric and constitutive equations hold at any point. Control over the
cross-sectional values is maintained. This approach enables cross-sectional analysis without
the need to consider the effects of concentrated forces or boundary conditions.
It is sufficiently general to be applied to the analysis of any cross-sectional shape with
longitudinal and transverse reinforcement if needed. Moreover, any constitutive model can
be used to describe the mechanical behaviour of materials. Prestress and the effect of non-
uniform strains caused by shrinkage or thermal effects can be introduced depending on the
needs and capabilities of the constitutive model.
The three-dimensional approach with fully three-dimensional analysis of the strain and
stress fields enables the use of realistic and sophisticated material models, such as those which
describe damage at the micro-structure level, softening or plasticity. In such models one must
not leave out certain components of the strain or stress field as is the case in the so-called
engineer’s methods based on elastic behaviour of the material used due to their simplicity
and universality. Moreover, they must not be used for modelling the behaviour of structures
composed of one-dimensional (1D) linear elements. In more advanced constitutive models,
different couplings between the strain and stress field components can be observed, increasing
the accuracy of representation of the actual behaviour of the analysed structure.
The above approach can be used with any FEM program that enables defining initial
(imposed) strains and is provided with the periodic boundary conditions option.

3.  Standard procedures

The standard procedures contained in Eurocode 2 [8] require a torsion check for reinforced
concrete members when the structure’s stability is defined by the torsional resistance of its
members. However, as mentioned, there are no guidelines on including the interaction of
bending with torsion and in when torsion results from the strain compatibility conditions (as
in statically indeterminate structures) the provisions are limited to recommending the use of
reinforcement for the crack width control (minimum longitudinal reinforcement, transverse
reinforcement and additional bars over the beam height).
All the design procedures are based on the relationships in thin-walled box sections in
which equilibrium is satisfied by closed shear flow. Since reinforced concrete members have,
as a rule, a solid cross-section, in the design they are represented by thin-walled components.
As far as the torsional resistance of concrete members is concerned, EC-2 [8] distinguishes
the torsional resistance identified with the torsional cracking moment TRd.c limited by the
stresses generated in the wall that exceed the tensile strength of the concrete tt.i = fctd and the
torsional moment resistance TRd.max defined by the diagonal compression failure. This value
depends on the freely chosen value of angle θ. Similarly to shear, the important parameter is
cotθ which in Poland can take any value in the range between 1 and 2 (1 and 2.5 in Europe).

134
This gives 100% difference of capacity between the limit values. However, in most cases it is
governed by the TRd.c capacity. The required quantity of additional steel is determined from
the condition of equilibrium of vertical forces in the section wall where the sum of forces
caused by torsion and forces caused by the action of transverse forces is equal to the capacity
of stirrups.

TEd Aswt  f yd
 hk   v  VEd   hk  cot  (5)
2 Ak s

where αv – factor depending on the number of stirrup legs (αv = 0.5 and αv = 0.25 for two and
four-legged stirrups respectively). The amount of reinforcing steel determined in this way
depends directly on the value of cotθ.
The following two interaction requirements are used to check the capacity of a section
subjected to torsion:

TEd VEd
 1 (6)
TRd . max VRd .max
TEd V
 Ed 1 (7)
TRd . c VRd . c
which take into account the combined action of shear force and torsional moment.
In the case of more complex cross-sections, in particular if they are sensitive to deplanation,
the standards prescribe strut-and-tie or beam-truss modelling.
The behaviour of a section subjected to pure torsion and in response to interaction taking
into account all the relevant factors might be of interest in this context.
As shown, the standard procedures present a very simplified description of the problem
and do not allow more complicated cases to be taken into account.

4.  Comparison with the experimental data

In this section the results obtained from numerical modelling of the homogenized model
with macro-strain control are compared with the experimental data obtained by McMullen
and Warwaruk [3] for solid cross-sections subjected to torsion. The comparative analysis has
been carried out for beams No. B11 and No. B21.
The strength of concrete, according to [3] was measured on 15 × 30 cm cylindrical
specimens.
The graphs show the relationship between the unit torsion angle and torsional moment
for a full-length beam model with force-controlled action and beam modelled on the basis of
homogenization assumptions. In both cases, the Concrete Plastic Damage model was chosen,
the parameters of which are given below.

135
 Table 1. Material data of beams B11 and B21 based on [3]
Transverse
Bottom reinforcement Top reinforcement
Beam fcm [MPa] fctm [MPa] reinforcement
A s [cm2] fy [MPa] A s [cm2] fy [MPa] A s [cm2] fy [MPa]
B11 35.78 3.41 5.70 323.4 1.42 365.4 1.42 379.2
B21 39.64 2.87 5.70 323.4 5.70 323.4 1.42 370.2

Table 2. Parameters for the Concrete Plastic Damage model of material

Parameter B11 B21 Description
Compressive/ tensile strength
fc 27.78 MPa 31.64 MPa fcm – 8 MPa, according to EC-2
ft 1.80 MPa 1.9 MPa fctk0.05
E 22.5 GPa 22.5 GPa taking into account the type of aggregate
ν 0.2 0.2  
fco /fc 0.4 0.4 initial uniaxial compressive strength
fcbo /fco 1.16 1.16 initial biaxial compressive strength
Damage in compression
σc,D /fc 0.95 0.95 activation stress level

σ c /fc 1 1 ref. stress level for damage

Dc 0.55 0.55 damage at ref. stress level
Gc 3.33 kN/m 3.33 kN/m fracture energy
Damage in tension

σ t /ft 0.5 0.5 ref. stress level for damage

D t 0.5 0.5 damage at ref. stress level

Gt 0.066 kN/m 0.066 kN/m fracture energy

s0 0.2 0.2 stiffness recovery factor
Dilatancy
type variable variable  
αpo 0.35 0.35 tensile dilatancy parameter
αp 0.35 0.35 dilatancy parameter (compr.)
σc.dil /fc 0.95 0.95 activation stress level (compr.)
αd 2 2 appex smoothing parameter

136
 Fig. 6. Torsional moment – unit torsion angle relationship for beam No. B11

Fig. 7. Torsional moment – unit torsion angle relationship for beam No. B21

The graph shows the relationship between the unit torsion angle and torsional moment
for beam in the homogenization approach with reinforcement and for plain concrete cross-
section. Also in this case, the Concrete Plastic Damage model was used with the parameters
given above.

Fig. 8. Torsional moment – unit torsion angle relationship for beam No. B11 with reinforcement and for plain
concrete cross-section

137
 Moreover, the beam behaviour is presented in the homogenisation approach for elastic
plastic material with softening behaviour (M-W) for wr = 0.001 m and dilatancy angle of
Yc = 7

Fig. 9. Torsional moment – unit torsion angle relationship for beam No. B11 for Concrete Plastic Damage and
M-W material models

5.  Final conclusions

It has been demonstrated that numerical analyses can be limited to analysing a section of
a linear element with a user-defined periodic boundary condition imposed on its walls. This
approach significantly reduces the calculation time. Moreover, it enables the use of a simple
deformation-controlled procedure. The results obtained in this way exhibit a satisfactory
consistency with the experimental data.
Note is made of the problem of selecting a material model for the continuum. While more
accurate results in the pre-failure stress-strain state of cross-sections can be obtained with
complex descriptions of the concrete, these require determination of numerous parameters,
not always discernible in the straightforward, engineer’s approaches.
Agreement of the results of numerical simulation with the experimental data obtained in
simple loading cases gives grounds for further research on the behaviour of beams subjected
to combined loading.

138
References

[1] Knauff M., Obliczanie konstrukcji żelbetowych wg Eurokodu 2: zasady ogólne i zasady
dotyczące budynków, Wydawnictwo Naukowe PWN, 2013.
[2] Knauff M. et al., Podstawy Projektowania konstrukcji żelbetowych sprężonych według
Erokodu 2, Dolnośląskie Wydawnictwo Naukowe, 2006.
[3] McMullen A.E., Warwaruk J., The Torsional Strenght of Rectangular Reinforced Concrete
Beams Subjected to Combined Loading, PhD thesis, University of Alberta, 1967.
[4] Menetrey P., Numerical analysis of punching failure in reinforced concrete structures, PhD
thesis, Lausanne EPFL, 1994.
[5] Truty A., Elastic-Plastic Damage Model, ZSoil.PC 160102 report, www.zsoil.com (access:
10.04.2018).
[6] Urbański A., The unified finite element formulation of homogenization of structural members
with a periodic microstucture, Politechnika Krakowska, 2005.
[7] Urbański A., Analysis of a beam cross-section under coupled actions including transversal
shear, International Journal of Solids and Structures 71, 2015, 291–307.
[8] PN-EN 1992-1-1:2008 Eurokod 2. Projektowanie konstrukcji z betonu. Część 1-1:
Reguły ogólne i reguły dla budynków.

If you want to quote this article, its proper bibliographic entry is as follow: Anielska D., Finite element analysis of reinforced
concrete elements subjected to torsion, Technical Transactions, Vol. 2/2019, pp. 129–140. 139