See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283117231

OPTIMIZATION OF UHP-FRC INVERTED UMBRELLA SHELL STRUCTURES

Conference Paper · May 2014

CITATIONS READS
0 1,191

3 authors:

Hesham Othman Magdy Tayel

Kinectrics Menoufia University
25 PUBLICATIONS   379 CITATIONS    32 PUBLICATIONS   583 CITATIONS   

SEE PROFILE SEE PROFILE

Hesham Marzouk
Ryerson University
169 PUBLICATIONS   2,753 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

FEM Analysis for UHP-FRC subject blast loading View project

Structural Health Monitoring on concrete Bridges View project

All content following this page was uploaded by Hesham Othman on 22 December 2015.

The user has requested enhancement of the downloaded file.

 CSCE 2014 4th International Structural Specialty Conference - 4e Conférence internationale
spécialisée sur les structures 2014 de la SCGC

Halifax, NS
May 28 to 31, 2014 / 28 au 31 mai 2014

OPTIMIZATION OF UHP-FRC INVERTED UMBRELLA SHELL

STRUCTURES

H. Othman1, M. Tayel2, H. Marzouk1

1Civil Engineering Department, Ryerson University, Toronto, Ontario, Canada
2Civil Engineering Department, Minufiya University, Shebin El-Kom, Minufiya, Egypt

Abstract: In current paper, shape and size optimization techniques are combined to set some
geometrical relationships as designing aids for inverted hyperbolic paraboloid (hypar) umbrella shell.
Material optimization is achieved by using Ultra High Performance Fibre Reinforced Concrete (UHP-FRC)
as construction material. UHP-FRC provides ductility, durability, strength, and excellent surface finish
quality. The use of UHP-FRC concrete improved the initial size optimization considerably. Therefore, the
size can be assumed to be constant and a shape optimization is performed to obtain the optimal overall
dimensions of the required shell. Then, the shape is considered to be constant and a size optimization is
performed to obtain the optimal range for shell thickness and other supporting beams inertia. The chosen
objective functions are to minimize maximum deflection, and to minimize the volumetric displacement,
under stresses and construction constraints. A manual optimization model was developed to investigate
the effect of certain parameters on the overall structural response. The determination of displacements
was made discretely at fixed points located at the nodes of meshes using ABAQUS/Standard version
6.10. The result of this optimization process provided a proper values range for the inverted umbrella
shells dimensions. It is concluded that significant improvements in the structural behavior may be
achieved with only slight geometric changes.

1. INTRODUCTION

Shell structures present an attractive lightness and elegance from an aesthetic point of view, Shell
structures strength depends on their geometry rather than material strength. It is important to set some
relationships between shell dimensions to be considered in the beginning of a project design. These
relationships should be done with the aims of improving shell mechanical behavior, particularly avoiding
or reducing the moments and deflections. Structural Optimization techniques are the most effective mean
to achieve this aim (Tomás & Martí 2010).

The basic idea of the optimization process consists of an analysis module and an optimization module.
The analysis module is used to calculate the structural response and to perform a sensitivity analysis.
Based on the sensitivity analysis, the optimization module calculates the change in the structural design
that improves the response. Typically, the optimal design is not achieved after only one optimization step.
However, the procedure consists of the analysing, the sensitivity analysis parameters and the
optimization step, is repeated several times. After a number of iterations, the design cannot be further
improved and an optimize structure can be reached.

Structural optimization methods can be divided into four main categories: Material optimization, shape
optimization, size optimization, and topology optimization (Sigmund 2000). Although, these optimization
techniques may have a lot of overlapping, the material optimization is related to the optimal use of

CST-70-1
material in the structure. The aim of material optimization is to achieve an optimum material usage with
minimum stresses. In fact, shells are already material optimized by covering a long-span with minimum of
material use. Moreover, the use of UHP-FRC for shell structures instead of normal strength concrete
offers outstanding material characteristics. UHP-FRC exhibits outstanding mechanical, and durability
properties. Such properties include: ultra-compressive strength exceeding 150 MPa, ductility, flexibility,
toughness, impact resistance, dimensional stability, durability, impermeability, corrosion resistance,
abrasion resistance, aggressive, environment resistance (Walraven 2009), and excellent surface finish
quality. Therefore, the used of UHP-FRC for shell structures limits most of problems reported in literature,
such as brittle and sudden collapse of shells reported in (Hoogenboom 2006), and cracks in top and
bottom surfaces due to creep and fatigue reported in (Burger & Billington 2006). Shape optimization
refers to a technique in which the shape of a structure is optimized. Solutions obtained from shape
optimization methods maintain the same thickness of structures. Size optimization refers to the
optimization of the size of structural elements in a shell, i.e. shell thickness. The results of size
optimization have same initial geometry without any change. Topology optimization finds the optimal
layout of the structure within a specified design space. The topology optimization combines all three
categories (Tomás & Martí 2010).

Structural optimization can be performed using Computational optimization algorithms, or manually

formulated functions. Computational optimization algorithms are not restricted to the classification
presented above, but often use a combination between different categories of optimization, in this type
the final optimal geometry may be different from initial geometry completely according to the choice of
objective function. Moreover, it is difficult to obtain geometrical relationships between shell dimensions by
using computational algorithms. Manual free optimization refers to a process in which the user modifies
certain parameters of a finite element model of a structure in order to evaluate the effects of that
modification. This method can be used easily to set some relationships between shell dimensions to be
used as designing aids. In current paper, manual shape and size optimization techniques are combined,
with the aim of minimizing displacement for the inverted umbrella shell under stresses and construction
constrains. This geometry has been chosen due to its simplicity, extended use, and enables comparison
with the results found by other authors.

2. OPTIMIZATION METHODOLOGY

The aim of the optimal structural design is to obtain a set of values for the design variables, which
minimizes an objective function and complies with the constraints that depend on the variables. Structural
optimization can be formulated as a mathematical optimization problem (Tomás & Martí 2010):

[1] Minimizes: f (x) Subject to

hj (x) = 0 j = 1,2,3, … … … … . . , mi
g k (x) ≥ 0 k = 1,2,3, … … … … . . , md
xil ≤ xi ≤ xiu i = 1,2,3, … … … … … , n

Where, x is the n-dimensional vector of the design variables; f(x) is the objective function; hj(x) is the jth
equality design constrain; gk(x) is the kth inequality design constraint; mi is the number of equality
constraints; md is the number of inequality constraints; n is the number of variables; and xL (xu) is the
lower limit (upper limit) of the variable i.

The objective function is usually chosen according to the purpose of optimization process. For example,
to minimize instability phenomena, the buckling load can be used as objective function. To minimize
displacements in the whole shell, a function called volumetric displacement may be used. Volumetric
displacement (VD) is defined according to the following expression (Tomás & Martí 2010; Ortega &
Robles 2003):

[2] VD = ∑ni=1 di ∗ Si ∗ Thick i

CST-70-2
Where di is the displacement vector at each point i; Si is the area of influence at this point; Thicki is the
average thickness of the structure at the mentioned area. The advantage of (VD) function is that provides
a wider view of displacement throughout the structure (shell and beams) and gives a deeper analysis of
the displacement (Tomás & Martí 2010; Ortega & Robles 2003). The design variables may be shell
thicknesses, beams moment of inertia, shell rise, etc. The constraints are the conditions that the design
must comply with in order to be regarded as valid.

Many authors utilized the manual optimization method for shell structures optimization. For example,
Ortega & Robles (2003) used volumetric flexure and volumetric displacement as objective functions, to
optimize hypar shell with straight edges, the design variables were rise, edge beam size, and shell
thickness. Burger & Billington (2006) used maximum deflection as objective function; the design variables
were shell rise and side length ratio. Holzer et al. (2008) used the maximum deflection as objective
function in the optimization of front mouth height of the famous Felix Candela’s Chapel Lomas de
Curnavaca.

3. OPTIMIZATION OF INVERTED UMBRELLA SHELL

In current study, in addition to material optimization achieved by using UHP-FRC, the optimization
process of inverted umbrella shell as shown Figure 1, fall into two categories: shape optimization and size
optimization. Initially, a shape optimization is performed to obtain the optimal overall dimensions of the
studied shell. The shell thickness and beams size are considered with constant values, i.e. the size is
fixed during shape optimization, and then the shell shape is fixed and size optimization is performed.

W
L
Shell element

Edge beam

Interior beam

Figure 1: Typical geometry and structural elements of inverted of Inverted Umbrella hypar shell

In current study two objective functions have been considered, namely maximum deflection and
volumetric displacement. Additionally, the optimization process is subjected to the following constraints:
The rise f > 0 (f = 0 flat plate), the inclined roof should have a convenient slope i.e. less than 30o with the
horizontal.

Optimization process usually takes place in preliminary analysis stage (conceptual design stage), in this
stage, it is usual to model the material as homogenous and isotropic. Steel reinforcement contribution is
not considered, except in the effect of total density. The primary loads usual considered in primary
analysis stage are gravity loads that include the dead weight and perhaps a uniform live load. In current

CST-70-3
study, the shell is modeled with homogeneous concrete material. The main properties of used concrete
are: 25 kN/m3 for specific weight, 140 MPa for the characteristics compressive strength, 50,000 MPa for
Young's Modulus, and 0.20 for Poisson’s ratio. The structure dead weight plus superimposed load of 1.00
kPa distributed all over the surface was adopted as load state.

3.1 Computation Procedure Used in Current Study

Since optimization process is a chain of events, each leading towards the optimal design. A total of 70
different shells have been studied. The analysis and optimization tools necessary for solving the
optimization problem tools are briefly discussed below as illustrated in Figure 2.

1. The starting point is the use of a CAD drafting program to generate the initial geometry. In current
study Rhinoceros (Robert & Associates 2013) was used to generate the hypar geometry.

2. the objective functions were evaluated using the finite element program ABAQUS/Standard 6.13
(Simulia 2013). The first objective function was maximum deflection as obtained directly from
ABAQUS, which occurs in all cases at the outer corner of the studied shells. While the second
objective function was the volumetric displacement as calculated using a MATLAB routine (m.file)
which gets the sum of each node displacement multiplied by its surround volume as shown in
Equation [2], the validation of this subroutine is mentioned in the beam size optimization section.

3. After evaluating the objective functions from pervious step and applying constrains. The next step
towards the optimal design is using the sensitivity analysis. The result of sensitivity analysis is
optimized range for certain design variable. The previous three steps have been repeated for
each design variable till the optimization process completed.

Figure 2: Flow chart of optimization process

3.2 Shape Optimization of Inverted Umbrella Shell

The design variables to be optimized in this section are the rise (f), side length ratio (L/W), edge beam
offset, and Interior beam offset. The other variables, like material properties, type and magnitude of the

CST-70-4
applied loads, shell covering area, shell thickness, edge beam size, and interior beam size are
considered constant. The optimum range chosen in each optimization process, in the following sub-
sections, is shown between to dot lines in graphs.

3.2.1 Optimization of shell rise

It was decided to start optimization from the rise with different values from 1 to 10 m, as the design
variable, adopting constants values for the other design variables, the models used in the analysis are
square shells with the following constant values for side length equal 10 m, i.e. side length ratio equal 1,
shell thickness of 0.10 m, beams dimension of 0.20 by 0.50 m, and beam offset equal zero, i.e. the
section located above and below the shell is the same. The rise optimization results for maximum
deflection volumetric displacement are shown in Figure 3.

Figure 3: The ratio of rise-to-area versus: a) maximum deflection (left); (b) volumetric displacement (right)

As shown the increase in the total rise (rise-to-area) associated with a decrease in maximum deflection
and volumetric displacement. The volumetric displacement versus the rise-to-area curve follows
approximately the same shape of max deflection versus rise-to-area curve, but with minimum value at a
total rise of 6.00 m, i.e. (rise-to-area ratio = 0.06). A useful analysis to the rise-to-area ratio results in
Figure 3, take into consideration the stress and construction constrains show that a range 0.02 to 0.03
can be taken as optimum, it can be seen that at the left of optimum range the curves slope is steeper than
at the other side. This is due to the fact that as the rise is gradually reduced, the mechanical behavior is
getting close flat plate (rise = 0.0). In this range, the possibility of minimum concrete volume is used and
the minimum overall weight, the possibility of buckling is lower, and roof slope less than 30o with the
horizontal, i.e. less construction cost.

3.2.2 Optimization of shell side length ratio

In current section, shell side length ratio is considered as the design variable, a series of
inverted umbrella shells will be analyzed with the same design variables values of past section
for loads, materials, thicknesses, and the covered area equal 100 m 2. Finite element models
were created and analyses for two cases of rise, shell with rise-to-area ratio equal 0.02 and
other with 0.03 m-1 (in optimum range obtained from rise optimization above). The two different
rise-to-area ratio are taken to investigate the effect of rise change in the results of optimization
process, different side lengths ratio (L/W) were considered. The results of the two objective
functions are shown in Figure 4.

CST-70-5
 Figure 4: Side length ratio versus: a) maximum deflection (left); b) volumetric displacement (right)

As shown in figure 4, for first objective function, deflections decrease consistently as the ratio of side
lengths goes between 1 and 2, the optimal form for an umbrella shell is not square. For second objective
function, the optimal form for an umbrella shell is a square. Taking into consideration the aesthetic point
of view as constrain, the optimum value for side length ratio equal to 1 i.e. the shell is square, if a range is
wanted to be set, a range from 1 to 2 is optimum as deflection and volumetric displacement curves are
approximately flat in this range and after that the curves are steep.

3.2.3 Optimization of shell beams offset

In current section, shell beams offset are the design variable as shown in Figure 5 and other design
variables are taken with constant values as in past sections. Finite element models were created and
analyses for edge beam offset as design variable for three cases (dropped, inverted, and beam with the
shell center), and interior beam offset is taken equal to zero until the optimization of edge beam
completed. Two different values of rise-to-area ratio are taken, 0.02 and 0.03 m -1.

The results of optimization process are shown in Table 1 for the two objective functions, for the two
studied cases of rise-to-area considered. The inverted edge beam, case with under line in Table 1, has a
higher efficiency in its structural behavior which is reflected in a smallest value of max. Deflection and
volumetric displacement. By taking into consideration the aesthetic point of view as constrain, this
optimum case is very good for the shell view from bottom.

Dropped edge beam Inverted edge beam Beam coincide with shell

Figure 5: Edge beam offset cases

CST-70-6
 Table 1: Results of the optimization process for edge beam offset
Rise/Area (m-1) 0.02 0.03
Deflection VD Deflection VD
Edge Beam offset
(mm) (m.m3) (mm) (m.m3)
Dropped -7.27 -0.027 -4.43 -0.016
Coincide -4.5 -0.022 -3.02 -0.014
Inverted -2.85 -0.020 -2.04 -0.013

The design variable for the interior beams offset, while the edge beam offset is taken as constant. The
optimum case obtained for inverted edge beam with the same constant values as for case for edge beam
offset optimization. The results are shown in Table 2.

Table 2: Results of the optimization process for interior beam offset

Rise/Area (m-1) 0.02 0.03
Deflection VD Deflection VD
Interior Beam offset
(mm) (m.m3) (mm) (m.m3)
Dropped -2.72 -0.018 -1.96 -0.012
Coincide -2.85 -0.020 -2.04 -0.013
Inverted -2.96 -0.022 -2.13 -0.014

By taking into consideration the aesthetic point of view as constrain, this optimum case has unacceptable
view, for the same reason the case of shell with the beam center may be refused, it should be noted that
the three values for two objective functions are approximately with same value because the membrane
force transmitted from shell to interior beams is compression force. While for edge beams the difference
between the three cases is quite large, because the membrane force transmitted from shell is tension
force. Although the case of inverted beam is the worst case for deflection, it will be used.

3.3 Size optimization of inverted umbrella shell

The design variables to be optimized in following sub-sections are: shell thickness and beams size. The
other variables are considered constant and in optimum range obtained from shape optimization process.

3.3.1 Optimization of shell thickness

In current section, the objective functions are evaluated for the two cases of rise-to-area values usually
taken. The design variable is the shell thickness, a variation between a minimum of 50 mm, which was
considered from the construction viewpoint, and to a maximum of 200 m, that the same thickness for
beams, all other variables are taken with constant values as usual. From results shown in Figure 6,
deflection decreases with thickness increases due to the increase in the shell stiffness. In the other hand
volumetric displacement variation increases as thickness is increased due to the increase in dead weight,
with a minimum value when the thickness equal 120 mm. From results in Figure 6, the shell thickness
should have a thickness between 80 and 120 mm. when taken into consideration the maximum
compressive stress in shell does not exceed allowable compressive strength of concrete, and the
displacements must be less than the limit set by standards, the values are safe.

CST-70-7
 Figure 6: Shell thickness versus: a) maximum deflection (left); b) volumetric displacement (right)

3.3.2 Optimization of shell beams size

In this section, beams size is the design variable, the shell thickness taken equal to 100 mm, in optimum
range. Finite element models were created and analyses for four cases, two different values of rise-to-
area ratio and two different values of side lengths ratio, all values chosen are in optimum ranges found in
previous sections, shell with rise-to-area ratio equal 0.02 and other with 0.03 m-1 are used, and shell with
side length ratio equal 1 and other with 1.5, these cases are taken to check the effect of these changes in
the results of optimization process. For each case mentioned above all beams width, edge and interior,
have the width of 500 mm. this means that the design variable is the beam height, the beams heights
ranged from shell thickness 100 mm to 500 mm as the design variable in this section, beam offset equal
zero. The results of the two objective functions are shown in figures 7 and 8.

As shown in figure. 7 for all cases studied, the disadvantage that might appear when beams dimensions
are reduced is the increase in deflection. In the other hand, in figure 8, the relation between volumetric
displacement and beams height is linear, when the beam height increase, volumetric displacement
increases which resulting from increasing of beams weight, This linear relation has been also obtained by
Ortega,4 in which anther hypar shape was studied using volumetric flexure as objective function.

From results shown in Figures 7 and 8, the shell beams should have the smallest possible section
provided that the maximum compressive stress does not exceed allowable compressive strength of
concrete. On the other hand, tensile stresses are not an important inconvenient due to the fact that these
are absorbed by steel reinforcement. However, the displacements must be less than the limit set by
standards. For inverted umbrella hypar shell, there are two types of beams, edge beams, and interior
beams. Edge beams subjected to tension force, so no need to edge beam or edge beam with same shell
thickness is sufficient, and the need is only for additional steel to absorb tensile stresses (ACI-334 1992),
while interior beams subjected to compression force, so the need for beam with smallest possible section
provided that the maximum compression stress does not exceed allowable compression strength of
concrete is sufficient. For the four cases of rise-to-area ratio and side length ratio, the curves follow the
same behavior for maximum deflection and volumetric displacement, and this conclude that the
optimization process did not change by the changes of rise and side length ratio. The relation between
volumetric displacement and beam size was straight line, this relation used to validate the written

CST-70-8
subroutine. As this relation was obtained also by (Ortega & Robles 2003) in which anther hypar shape
were studied using same objective function.

Figure 7: Beams height versus maximum deflection: a) shell with side length ratio =1 (left); b) shell with
side length ratio =1.5 (right)

Figure 8: Beams height versus volumetric displacement: a) shell with side length ratio =1 (left); b) shell
with side length ratio =1.5 (right)

4. CONCLUSION

The shape and size optimization techniques were combined in the current paper, by minimizing maximum
deflection and volumetric displacement. Proposed technique of optimization can be considered as a
valuable tool for the shells designer to set a proper values range for different dimensions of different shell
types. In addition to the use of UHPF-RC enhances and overcomes the traditional concrete limitations

CST-70-9
 such as cracks in tension surface, durability problems, seismic resistance and surface finish quality. From
the results obtained in current study, the following aspects can be highlighted for inverted umbrella
hyperbolic parabolid shell:
 The optimum range for the shell rise to its area ratio is in the range 0.02 to 0.03.
 The optimal form for an umbrella shell is square or rectangular with side lengths ratio less than 2.
 The optimum range for the shell rise to its area ratio is in the range 0.0008 to 0.00012.
 The inverted edge beams is the optimum for considered objective functions and from aesthetic
view point.
 However, for the interior beam offset optimum case is dropped interior beam.
 For inverted umbrella shell there is no need to optimize beams offset, since it is subjected to
tension force.
 The shell beams should be the smallest possible section such that the maximum compressive
stress does not exceed allowable compression strength of concrete. The deflection must be less
than set by standards.

5. REFERENCES

ACI-334, 1992. Concrete Shell Structures Practice and Commentary. ACI 334.1R-92, (Reapproved
2002), pp.1–10.

Burger, N. & Billington, D., 2006. Felix Candela, elegance and endurance: an examination of the
Xochimilco shell. Journal of the international association for shell and spatial structures: IASS, 47.

Holzer, C.E., Garlock, M.E.M. & Prevost, J.H., 2008. Structural optimization of Felix Candela’s chapel
Lomas De Cuernavaca Candela's optimization of form. In Fifth International Conference on Thin-
Walled Structures. Brisbane, Australia.

Hoogenboom, P., 2006. Stability of Shells. Lecture Notes CT5143 Delft University of Technology.

MathWorks, 2011. MATLAB program R2011b.

Ortega, N. & Robles, S., 2003. The design of hyperbolic paraboloids on the basis of their mechanical
behaviour. Thin-Walled Structures, 41(8), pp.769–784.

Robert, M. & Associates, 2013. Rhinoceros®. Available at: http://www.rhino3d.com/features.htm#NURBS.

Sigmund, O., 2000. Topology optimization: a tool for the tailoring of structures and materials.
Transactions of the Royal Society, (358), pp.211–226.

Simulia, 2013. ABAQUS user’s manuals. Dassault Systèmes Simulia Corp.

Tomás, A. & Martí, P., 2010. Shape and size optimisation of concrete shells. Engineering Structures,
32(6), pp.1650–1658.

Walraven, J.C., 2009. High performance fiber reinforced concrete: progress in knowledge and design
codes. Materials and Structures, 42(9), pp.1247–1260.

CST-70-10

View publication stats