IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS You may also like

- DARK MATTER ADMIXED TYPE Ia
Analysis of a reinforced concrete dome SUPERNOVAE
S.-C. Leung, M.-C. Chu and L.-M. Lin

To cite this article: P Czumaj et al 2021 IOP Conf. Ser.: Mater. Sci. Eng. 1015 012006 - An implementation of the phase-field
model based on coupled
thermomechanical finite element solvers
for large-strain twinning, explicit dynamic
fracture and the classical Stefan problem
Milovan Zecevic, M J Cawkwell, K J
View the article online for updates and enhancements. Ramos et al.

- 2d finite element analysis of inconel 718

under turning processes
G Chandra Mouli, K Prakash Marimuthu
and T Jagadeesha

This content was downloaded from IP address 37.238.90.22 on 19/11/2022 at 18:54

XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

Analysis of a reinforced concrete dome

P Czumaj, S Dudziak and Z Kacprzyk

The Faculty of Civil Engineering, Warsaw University of Technology, al. Armii
Ludowej 16, 00-637 Warsaw, Poland

E-mail address: [email protected]

Abstract. FEM models of axi-symmetrical reinforced concrete dome with two rings have been
analysed. Different complexity level of computational models (2D and 3D), geometry
simplifications and FEM codes (Abaqus, FEAS, ARSAP) have been compared. Assessment of
building structure deflections has been performed with several approaches, which gave
opportunity to confront them and estimate mistakes of most commonly used models.

1. Introduction
Nowadays, static analysis of building structures is most often performed with the Finite Element Method
(FEM) and 3D models consisting of shells and bars [1]. During phase of geometry preparation for such
models, structural designers have to make many decisions concerning e.g. relative position of neutral
axis and mid-surfaces (resulting in eccentricities). It is obvious that every designer adopts various
simplifications [2–4]. Hence, modern design codes take into account the uncertainty of models in safety
factors [5]. However, some simplifications can lead to major mistakes [6], therefore, in case of complex
systems, it is worthy to perform validation and verification of models before final static analysis.
Axial symmetry of geometry is rarely used in static analyses due to unsymmetrical variable actions
(especially wind or thermal actions). On the other hand, axi-symmetry allows significant reduction of
model variables, which means, that it is possible to prepare 2D models based on the theory of elasticity,
which ensure almost exact results for symmetrical loads [7]. Such models can be used in the verification
and validation procedure for more complex 3D models consisting of shells and bars.
The main objective for this paper is to compare models of different level of complexity, concerning
axial-symmetrical dome with two circular flanges, shown in figure 1.

Figure 1. Concrete dome with two circular flanges.

Three groups of models have been analyzed:
 P1: 2D axial-symmetrical problem of theory of elasticity (this model will be treated as a
reference for further 3D models),
 P2: 3D task with axial-symmetrical cylindrical finite elements,
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

 P3: 3D models consisting of bar and shell type of finite elements.

Each group contained at least two models created with different FEM codes:
 Abaqus FEA (Abaqus) [8]
 Finite Element Analysis System (FEAS) [9]
 Autodesk Robot Structural Analysis Professional 2020 (ARSAP) [10]

Figure 2. Dimensions of the dome.

Each software was operated by a different person, so personal influences in modelling process were
also included.

2. Geometry and material parameters

Geometry of the dome has been shown in figure 2. However, it is not always possible to model accurate
geometry. Important issue of modelling structures is approximation of geometry. In first group of
models, it could have been almost accurate, while in further models, volumetric structural components
have been represented with bar and shell finite elements. On the left side of figure 3, exemplary
approaches to geometry simplifications have been presented respectively: in the first model, mid-surface
of shell has been assumed on the top surface of the dome and its span has been extended to centroids of
both flanges. In the second model, surface in the middle of dome thickness has been taken as shell mid-
surface, while its span has been assumed in the same way as in the first model. The third model takes
mid surface in same way as the second one, though the span is cut by flange faces.

2.1. Material parameters

Mechanical properties for concrete have been adopted in the analysis. Parameters have been taken, as
follows:
 Young modulus E = 27027 MPa,
 Poisson ration ν = 1/6,
 unit weight ρ = 27,5 kN/m3.

2.2. Loads
Two load cases has been analysed: dead load of the structure and linear force of 500 kN/m over top ring
perimeter. Since top flange radius equals 0,751 m, total force applied is 2359,34 kN.

2
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

2.3. Boundary conditions

In this example only bottom flange has been supported in central point of its bottom surface. One degree
of freedom has been taken: vertical translation in one node, or more precisely: over circuit (see figure 4)

Figure 3. Dome models: a) top face

adjust, b) computational model, c)
middle surface adjustment, d)
computational model, e) middle
surface adjustment and cut with bar
sections, f) computational model.

3. Analysis

3.1. Model p1. Two-dimensional problem of theory of elasticity: axial symmetry

2D Calculations have been performed with all aforementioned three FEM codes. With each, at least two
different size of finite elements have been checked, to ensure displacement convergence. In FEAS and
ARSAP models, triangle finite elements with three nodes have been used, whereas triangle elements
with six nodes (AX6) have been adopted in Abaqus.
Figures 4÷6 depicts final finite elements meshes for each software.

Figure 4. FEAS, 1482 nodes, 2486 elements. Figure 5. ARSAP, 3345 nodes, 4974 elements.

3
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

Figure 6. Abaqus, 3642 nodes, 1571 elements.

In ARSAP, finite elements generator for curvilinear geometry structures is bound with geometry
definition. High accuracy of geometry imposes large number of elements (see figure 5), although high
mesh density is not always desired. In this example, the amount of elements is rather too high.

3.2. Model p2. 3D task with axi-symmetrical cylindrical finite elements

In this point, the dome has been modelled with rarely used cylindrical finite elements, which are not
implemented in ARSAP, nonetheless are available in FEAS and Abaqus. These elements can be
considered as one-dimensional as was shown in figure 7 (mid-surface is parametrised by one natural
coordinate s). Such elements are quite similar to beam elements since each node has three degrees of
freedom – two translations and one rotation angle.

Figure 7. Cylindrical finite element.

3.2.1. Abaqus
With Abaqus, 17 three-node shell elements (SAX2), which gives total of 35 nodes, have been used.
Two-parameter springs, with rotation (kφ) and translation (ku) stiffness as in equations (1) and (2), have
substituted flanges. Both stiffness were acquired from [11], though eccentricities between shell and rings
have been neglected.
EA
ku  (1)
R2
EJ
k  (2)
R2
Horizontal translation and rotation described with (1) have been added to boundary conditions from
previous models. with parameters from equations (1) and (2).

4
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

3.2.2. FEAS
Model created with FEAS consists of 120 nodes, 119 two-node shell elements and 2 single-node
cylindrical elements for top and bottom flange. It is worth pointing out that initial mesh studies revealed
that 59 elements, 60 nodes and 2 do not guarantee correct results.

3.3. Model p3. 3D models consisting of bar and shell type of finite elements
This group contains models from each of three presented FEM codes. Table 1 gathers all models and its
main properties.

Table 1. Analysed computational models.

no shell element bar element flange concentricity FEM code

P3-1 8-nodal curvilinear 3-nodal curvilinear no Abaqus
P3-2 8-nodal curvilinear 3-nodal curvilinear yes Abaqus
P3-3 3-nodal plain 2-nodal yes FEAS
P3-4 3-nodal plain 2-nodal yes ARSAP

Figure 8. Abaqus, computational model P3-1, Figure 9. Abaqus, computational model P3-2,
3312 nodes, 1008 shell elements, 64 bar 2900 nodes, 812 shell elements, 116 bar
elements. elements.

Figure 10. FEAS, computational model P3-3 Figure 11. ARSAP, computational model P3-4,
(plan), 2800 nodes, 5400 shell elements, 200 bar 9190 nodes, 3836 shell elements, 360 bar
elements elements

5
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

Two Abaqus models with same types of elements (S8R and B32) have been created, as shown in
figures 8 and 9. They differed with approach to flange position: model P3-1 represents exact geometry,
therefore flange concentricity has been preserved. Model P3-2 neglected eccentricity of flanges,
according to figure 3f.
In FEAS (figure 10) and ARSAP (figure 11) models, similar finite elements have been used. In the
FEAS model, finite elements mesh has been obtained with use of geometry primitives generator. It
allows obtaining regular elements, according to requested parameters.

4. Results summary
In tables 2 and 3, vertical displacements of top ring over all models have been presented for two analysed
load cases. Ratios between results obtained from model P1 (considered as reference – the most accurate
model) and other models were shown in brackets. For comparable models which were analysed in all
three FEM codes mean, standard deviation and coefficient of variation (COV) were calculated (see
tables 4 and 5).

Table 2. Vertical displacement of the top ring – dead weight [m ∙ 10-5].

P1 P2 P3
5,39 (1,07)
Abaqus 5,05 5,75 (1,14)
5,75 (1,14)
FEAS 5,34 5,37 (1,01) 5,43 (1,02)
ARSAP 5,32 - 5,64 (1,06)

Table 3. Vertical displacement of the top ring – line load [cm].

P1 P2 P3
1,06 (1,06)
Abaqus 0,968 1,27 (1,31)
1,27 (1,31)
FEAS 0,948 1,01 (1,07) 1,15 (1,21)
ARSAP 0,953 - 1,30 (1,36)

Table 4. Vertical displacement of the top ring – dead weight

– comparison of results [m ∙ 10-5].
P1 P3
Abaqus 5,05 5,75
FEAS 5,34 5,43
ARSAP 5,32 5,64
mean 5,24 5,61
standard deviation 0,16 0,16
COV 3,09% 2,90%

6
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

Table 5. Vertical displacement of the top ring – line load –

comparison of results [cm].
P1 P3
Abaqus 0,968 1,27
FEAS 0,948 1,15
ARSAP 0,953 1,30
mean 0,956 1,24
standard deviation 0,010 0,08
COV 1,09% 6,40%
Sample deformations of models loaded with dead weight are presented in figures 12÷15.
Figures 16÷20 shows deformations of models loaded with linear load over top ring.

Figure 12. P1 ARSAP, dead weight. Figure 13. P1 FEAS, dead weight.

Figure 14. P2 Abaqus, dead weight. Figure 15. P3 Abaqus, dead weight.

Figure 16. P1 ARSAP, line load. Figure 17. P1 FEAS, line load.

7
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

Figure 18. P2 Abaqus, line load. Figure 19. P3 ARSAP, line load.

Figure 20. P3 FEAS, line load.

5. Result discussion and conclusions

Differences between reference P1 model and models consisting of structural elements – P2 and P3 are
significant (up to 30%). On the other hand, deflections predicted by these models are overestimated,
therefore from engineering point of view, they are “on the safe side”. However, using models which
overestimate deflections so significantly, can lead to uneconomical design. In P3 model family, the best
agreement between accurate model and 3D model, was obtained for Abaqus model taking into account
the eccentrics between the shell and the rings (7% for dead weight and 6% for line load). For programs
analyzed separately the smallest differences for different class of models occurs for FEAS (1-2% for
dead weight and to 21% for line load).
Dispersion between results obtained with models prepared in different programs by different users
are smaller than for models of different classes. The smaller COV value (which is measure of result
relative dispersion) was obtained in case of P1 model for line load, while the greatest was in case of P3
model and line load as well. The small dispersion of results from different software, indicate that users
have not made major mistakes.
The paper presents the verification procedure of the concrete dome with two rings. The dome was
modelled with different class models (2D and 3D) in three different FEM codes. The 2D axi-symmetrical
models were used to verify the complex 3D shell-bar model, which later could be applicate to the static
analysis covering unsymmetrical load cases also.

References
[1] Perelmuter A V and Slivker V I 2003 Numerical Structural Analysis: Methods, Models and
Pitfalls (Berlin Heidelberg: Springer)
[2] Zobel H, Zbiciak A, Oleszek R, Michalczyk R and Mossakowski P 2014 Numeryczna
identyfikacja cech dynamicznych stalowo-betonowego mostu kolejowego Roads Bridg. -

8
XXIX R-P-S Seminar 2020 IOP Publishing
IOP Conf. Series: Materials Science and Engineering 1015 (2021) 012006 doi:10.1088/1757-899X/1015/1/012006

Drog. i Most. 13 pp 275–301

[3] Czumaj P, Dudziak S and Kacprzyk Z 2018 Computational models of reinforced concrete
ribbed floor MATEC Web Conf. 196 pp 1–8
[4] Kacprzyk Z and Czumaj P 2019 Modelling of Multi-Storey Frame Interacting with Rigid Core
of the Building IOP Conf. Ser. Mater. Sci. Eng. 661 pp 0–8
[5] EN 1990 Eurocode - Basis of structural desing
[6] Sidorov V N and Nowak K 2018 Numerical investigation of the long-term work of arches under
material creep conditions IOP Conf. Ser. Mater. Sci. Eng. 456
[7] Lewiński P M and Dudziak S 2018 Nonlinear interaction analysis of RC cylindrical tank with
subsoil by adopting two kinds of constitutive models for ground and structure AIP
Conference Proceedings 1922 (AIP Publishing LLC) p 130007
[8] Abaqus Theory Manual (6.12)
[9] Kacprzyk Z and Postek E 1993 System FEAS Isoparametric Finite Element Library Comput.
Methods Civ. Eng. 3 pp 89–108
[10] Robot Structural Analysis Professional 2020 - manual
[11] Mazurkiewicz Z 2004 Cienkie powłoki sprężyste: teoria liniowa (Warszawa: Oficyna
Wydawnicza Politechniki Warszawskiej)