Chimney Buckling Final Report Corrected
Chimney Buckling Final Report Corrected
Chimney Buckling Final Report Corrected
FINAL REPORT
By
Michael Angelides(1), Charis J. Gantes(2), Konstantinos E. Kalochairetis(2)
(1)
AMTE Consulting Engineers
(2)
Institute of Steel Structures, National Technical University of Athens
1 INTRODUCTION .................................................................................................... 1
1.1 PARTICIPANTS ................................................................................................................ 1
1.2 STATEMENT OF THE PROBLEM ........................................................................................ 1
1.2.1 Impact of compression stress field .......................................................................... 1
1.2.2 Impact of stiffeners ................................................................................................ 2
1.2.3 Impact of breeching ............................................................................................... 3
1.3 AIM OF THE RESEARCH PROJECT .................................................................................... 3
8 REFERENCES ...................................................................................................... 85
1 INTRODUCTION
1.1 PARTICIPANTS
Dr. Charis Gantes, Structural Engineer, Professor at the National Technical University of Athens
Dr. Konstantinos Kalochairetis, Structural Engineer, Research Associate at the National Technical
University of Athens
Steel chimneys are commonly used in engineering practice, offering structurally efficient solutions.
Until recently, steel chimney diameters did not use to exceed 4 meters and this is reflected in the
pertinent technical literature. However, in recent years the development and proliferation of combined
cycle power plants has led to boiler exhaust stacks with diameters in the order of 7 meters.
Steel chimneys are usually designed with respect to shell buckling according to one of the following
codes:
The provisions of all three above codes contain uncertainties with respect to frequently encountered
issues concerning the design of modern large diameter steel chimneys. These are outlined hereunder
and have provided the motivation for the initiation of the present research work.
Buckling is initiated by compressive stresses. In a steel chimney, the vertical compressive stresses are
predominantly due to bending arising from horizontal actions (wind, earthquake, pressure) and to a
lesser degree due to vertical dead and live loading. The three above codes deal with the issue of the
compression stress field responsible for buckling in very different ways. CICIND provides allowance for
the weighted contribution of both axial and bending compressive stresses through the factor alpha,
which reduces the critical elastic buckling stress. EN 1991-1-6 provides closed-form expressions for
the calculation of the elastic critical buckling stress in Appendix D, which are all explicitly applicable to
axial compression loading. The ASME STS-1 code explicitly states that the same allowable compressive
stress formulae are valid regardless of whether compressive stresses result from pure axial
compression or from combined axial compression and bending.
The first pertinent analytical work referred to a cylindrical shell with simple boundary conditions under
axial compression (Timoshenko (1910)), and resulted in the evaluation of the classical elastic critical
buckling stress, which was found equal to:
In the above expression E is the elastic modulus of the shell material, t is the shell thickness and r the
radius of the cylinder. It can be seen that the elastic critical buckling stress depends on the material
and the geometrical characteristics of the cross-section, while the effect of the cylinder length is not
included. Important discrepancies from this value were observed when experimental tests were
conducted, which were mainly attributed to initial imperfections (including residual stresses,
geometrical imperfections, load eccentricities, material inhomogeneity), prebuckling deformations and
the effect of boundary conditions.
The first researcher who investigated the buckling stress of cylinders under combined axial force and
bending was Flϋgge (1932), who assumed specific radius to thickness ratio and buckling wavelength.
Based on these results, he concluded that the buckling stress is 30% larger than the one found for
axial compression. This conclusion was also presented by Timoshenko (1932) and it was then used by
many other researchers. It was not until 1961, that Seide and Weingarten proved that the wavelength
used by Flϋgge was incorrect and led to misleading conclusions. According to their work, the buckling
stresses due to axial compression and due to flexural bending are very close to each other, provided
that no ovalization of the cross-section takes place. This conclusion was also verified analytically by
Reddy and Calladine (1978) and experimentally by Bailey and Kulak (1984).
This issue needs to be investigated and reflected in modern design codes of chimneys in a consistent
manner. In the context of the present work, the effect of the stress field on the buckling stress will be
investigated by using numerical procedures and by considering many different types of loading.
Both CICIND and EN codes either explicitly or implicitly assume that the structures are unstiffened
cylindrical shells. CICIND states that "Stiffeners may be used to increase the shell's resistance to
buckling" (8.3.4). EN 1991-1-6 provides closed form expressions for the calculation of the elastic
critical buckling stress in Appendix D, which are all explicitly applicable to unstiffened cylinders. ASME
STS-1 does not specify whether the allowable axial compressive stress formulae are applicable to
stiffened or unstiffened chimneys.
It is nevertheless known that the presence of stiffeners (both circumferential and radial) directly
influence the buckling strength of cylindrical shells (Teng & Rotter 2004, Angelides 2013 & 2014). The
impact of both longitudinal and circumferential stiffeners on the buckling strength of cylindrical shells
is addressed in design codes from the offshore industry such as API Bulletin 2U: Bulletin on Stability
Design of Cylindrical Shells and DNV-RP-C202: Buckling Strength of Shells.
This issue also needs to be reflected in modern chimney design codes, particularly since stiffeners are
almost always present in large diameter chimneys for ovalling control, and neglecting their impact on
buckling strength may be unrealistically conservative.
Large diameter chimneys often have breechings of significant dimensions, particularly in the case of
chimneys serving Heat Recovery Steam Generators (boilers) in combined cycle power plants. These
openings are frequently as wide as 150 degrees and as high as 20 to 25 meters. The fact that the
openings are usually placed very close to the chimney base means that the breeching is in a very
highly stressed region, and these stresses are often compressive. The influence of even smaller
openings on the stability behaviour of steel shells is significant, as has been demonstrated for example
for the case of tubular wind turbine towers (Dimopoulos & Gantes 2012, 2013). All three design codes
at best provide only general reference to this problem, without offering any explicit guidance. CICIND
provides a statement in a commentary that "if the vertical height of the opening is more than twice its
horizontal width, a stability check is needed". EN 1991-1-6 does not contain any reference to the
impact from shell openings on buckling strength. The EN code for steel chimneys (EN 1993-3-2) only
makes general statements on properly addressing opening issues, although it is clear from the
relevant figures (Fig. 1.1 & Fig. 6.1) that the openings considered are relatively small. ASME STS-1
refers to openings in 4.6.1, but only from a strength point of view.
However, openings of significant dimensions near the base are a common fact for modern large
diameter steel chimneys and the absence of any guidance on this issue may lead to a
misinterpretation of its significance. This needs to be reflected in modern design codes.
It follows from the above that significant issues concerning the buckling design of modern large
diameter steel chimneys are not directly addressed in current international steel chimney design
codes. Furthermore, a significant discrepancy has been observed in the buckling design results for
large diameter chimneys between the ASME code on one hand and the CICIND and EN codes on the
other (Angelides, 2013 & 2014). The aim of the proposed research project is to address the issues
outlined above and to provide a quantification of their impact on the buckling strength of large
diameter steel chimneys. The results will be useful for proposing a follow-up research project,
consisting of an experimental program in order to calibrate and validate the numerical models and a
subsequent wide range numerical parametric analysis to enable extrapolation, so that specific
suggestions may be made for improved design guidelines.
In this section, the prototype chimney under investigation is presented. This chimney is based on
experience from actual chimneys, but it incorporates simplifying assumptions, which facilitate the
calculations without having an important impact on the structural response. The chimney geometry
has been created using experience from past projects. It emulates a typical steel chimney serving a
HRSG (boiler) at a combined cycle power plant. The chimney is described briefly hereunder.
In addition to the prototype chimney, the procedure that an engineer of practice would follow
nowadays for the analysis and design of such a chimney, is also presented in this chapter. To that
effect, the simplified analysis and the accompanying analytical calculations of buckling stresses
according to three different design procedures are described, and are then used for the evaluation of
the utilization ratios of the chimney cross-sections.
The basic geometrical characteristics of the prototype chimney are presented in this section.
A view of the prototype chimney including stiffeners, breeching and platforms at different levels is
presented in Figure 2-1. In this case the breeching is achieved by providing a large number of holes,
which is not the only possible solution for its configuration, as will be shown in the next chapters.
The steel grade used is S235 JR. European material grades and European code material properties
have been used throughout this work, for consistency purposes.
An operating temperature of 200 deg C has been assumed. This leads to the following reduced
mechanical properties for S235 (as per EN 13084-7):
Yield stress fy = 160 MPa, Elastic modulus E = 2.025×108 kN/m2 = 202.5 GPa, Poisson ratio ν=0.3.
In the elastic numerical calculations only the weight of the steel, the elastic modulus and the Poisson
ratio are taken into account. In the elastoplastic analyses, apart from these parameters, the yield
stress is also included by considering a bilinear stress-strain relationship without strain hardening.
The chimney is considered to be fixed at the base, which restricts all displacements and rotations of
the corresponding nodes. At its top, the chimney is assumed to have a relatively stiff circumferential
ring that restricts any possible deformations of the top cross-section. Therefore, the top cross-section
maintains its circular shape, regardless of the level of loading. In the analytical calculations that will be
presented in the next sections, the stiff ring at the top is not taken into account.
For the purposes of the present work, only vertical loads and wind loads are considered. The vertical
loads are selected so that they are representative of actual dead loads expected to be present in such
a chimney, while the wind loads are calculated according to the European code for the specific
geometry, using assumed wind velocity and terrain category. Since the investigation targets the
buckling strength of the chimney, additional loads, such as snow, live loads, internal pressure, etc,
and possible variations of the load values used, will not affect the calculations.
Only the stiffness of the shell and stiffeners are considered in the calculations. All other attachments
are accounted for by their weight only. The following attachments and corresponding weights have
been used:
o Platforms: 3 full perimetric platforms at levels +30.00, +50.00, +55.00 with 7.0 tons weight
each.
o Inlet duct: The inlet duct is simulated through a total weight of 35.00 tons applied at the
vertical edges of the breeching. At the stage where no breeching is considered this weight is
ignored.
Silencer: 150 kN
For the evaluation of wind loads, the methodology described in EN 1991-1-4 is used, which is briefly
described hereunder:
Vb,0 = 30 m/s (project assumption)
kr = 0.19×(z0/z0,11)0.07 = 0.190
z0,II = 0.05
cf = cf,0×ψλ
cf,0 = 1.2 + 0.18log(10k/b)/(1 + 0.4log(Re/106)
k = 0.2 mm
ψλ = 0.675
A = 7.20 m (including eventual external insulation)
The wind force distribution corresponding to diameter and width equal to 7.20m is presented in Table
2-1.
The wind pressure is distributed around the circumference according to EN 1991, as shown in Figure
2-2, and its values are calculated according to the formula below:
we = qp(ze)× cpe,
where:
c pe= c po,h ψλ
o
30 .
a aA
.
0o 180o
b
+1
The corresponding pressure coefficients and resulting pressures per m of height of the stack are
shown in Table 2-2.
The wind pressure values that are used in the calculations, by considering for simplification that the
wind pressure is constant over 5m intervals, are presented in Table 2-3. In the numerical modelling,
the wind pressure distribution is assumed to have a linear variation along the perimeter of the shell
structure, as illustrated in Figure 2-3.
Figure 2-3: Wind pressure distribution along the perimeter and its simplified linear variation used in the numerical
model
In the present section, a brief presentation of the common design practice for such chimneys is
presented, as applied to the chimney under investigation. The prevailing approach proposed by
modern codes for carrying out ultimate limit state checks consists of obtaining action effects by means
of linear elastic analysis of the structure subjected to design loads, and comparing them to resistances
that account for both types of eventual nonlinearity, material and geometric. Thus, prediction of
collapse, a highly nonlinear phenomenon, is accomplished by means of linear elastic analysis. Collapse
may be due to development at certain cross-sections of stresses exceeding the material capacity, a
condition known as material nonlinearity, or due to buckling, associated with sudden increase of
deformations for a small increase in applied load, a condition known also as geometric nonlinearity, or
due to a combination of both types of nonlinearity.
In order to evaluate strength, modern codes assume full exploitation of material capacity and
appropriate reduction factors which take into account geometric nonlinearities. The reduction factor is
obtained from so-called buckling curves, which have been calibrated for specific types of structures
and cross-sections by means of a wide range of experimental results.
The above approach offers significant advantages as it is straightforward and the prediction of
collapse is elaborated by means of linear elastic analysis. This means that it can be carried out
through the use of commonly available commercial software, it is computationally inexpensive and
does not require special expertise from the engineer in setting up the structural model, selecting the
analysis parameters and interpreting the results. Furthermore, the results are quite reliable for
ordinary structural systems. Whether they are also reliable for more complicated structures such as
steel chimney with large diameter is a subject under investigation in the present research work.
The stresses are calculated by making use of the basic assumptions of beam theory. According to it,
plane sections remain plane and the cross-sectional shape remains circular during loading. The whole
column is assumed to be a cantilever with constant circular cross-section with diameter equal to 7m.
o D=7,000mm, t=13 mm
8.61
8.44
-107
8.27
-425
8.07 -949
7.86 -1676
7.62 -2598
7.65
7.35
-3711
-5010
7.03
-6486
6 .66
6.66
-8128 6.18
-9924
5.54
-11856
4.52
-13901
418
3.30
X
Y
Figure 2-4: Beam model with the applied distributed load and the corresponding bending moment diagram
Therefore, based on beam theory assumptions, the maximum normal stress appears at the bottom
support and consists of two parts:
For ease of comparison, the same load factors defined in EN 1990 for the load combinations at the
Ultimate Limit States are used in the calculations for all different codes:
For completeness, it is mentioned that the ULS combination in CICIND is: 1.10×Dead Loads +
1.50×Wind Loads. ASME STS-1 uses an allowable stress approach with a global safety factor of 1.5
against the ultimate stress calculated.
fy
λ= = 0.83 < √2 → σk = (1 - 0.412×λ1.2)×fy = 0.67×fy = 107.3 MPa
α × σ cr
αN × σN* + αB × σB*
α= = 0.51
σ*N + σB*
The utilization of the cross-section can be evaluated as the ratio of the maximum compressive stress
at ULS to the design buckling stress: 52.3/97.5 = 0.536
The procedure that should be followed for the evaluation of buckling strength according to EN1993-1-
6 is briefly described hereunder:
σX,Rd = σX,Rk /γM = 110.7/1.1 = 100.7 MPa (a material safety factor of 1.1 is used throughout this
approach, for uniformity and ease of comparison).
αx
λ x0 = 0.20, λ p = = 1.01
1-β
β = 0.60, η = 1.0
0.62 = 0.405
αx =
1 + 1.91 × (Δwk / t)1.44
1 r
Δwk = × t = 0.0053
Q t
E × t × Y = 94.05 MPa (no safety factor is applied to this formula, in order to compare directly
Scl =
4 ×D
to ULS stresses from the other codes).
Le /R = 120/3.5 = 34.3 < 60 → Y = 1.0
For a material safety factor of 1.1 that is compatible with the other codes, the utilization ratio is
obtained as: 52.3/(94.05/1.1) = 0.612
If the safety factor of the specific code is used, the allowable stress approach of the code leads to the
following results:
The three different design approaches used for the evaluation of the ultimate strength of the
prototype chimney result in different utilization ratios. The difference of the CICIND approach from
the EN provisions is only 3.2%. On the other hand, the difference of the ASME provisions from the EN
approach is larger and equal to 15.1%. A summary of the results presented in the previous sections is
provided in Table 2-4. The present research work will focus on the design provisions of CICIND.
Table 2-4: Summary of analytical results according to CICIND, EN1993-1-6 and ASME STS-1
2.4 CONCLUSIONS
In the present chapter, the geometry, material characteristics and applied loads of a prototype
chimney that will be used in this study were presented. The simplifying assumptions that are used in
the calculations were also described with emphasis on the ones related to the geometrical features
and to the application of the wind pressures normally to the thin-walled shell surfaces.
Additionally, the analysis and design process that is commonly used by engineers of practice was
described.
Analysis: In engineering practice the analysis of such involves beam modelling and simplified
application of the wind loads to a beam-type structure. Thus, a three-dimensional problem is
transformed into a simplified two-dimensional one. The type of analysis is usually linear elastic and
provides internal forces necessary for the design process. The simplifications induced in the analysis of
a chimney structure constitute a first source of inaccuracy.
Design: Three design approaches were used for the evaluation of the utilization ratio of the cross-
section of the chimney model and it was concluded that there are differences between the design
methods. The assumptions inserted in the design methods constitute a second source of inaccuracy.
In the next chapters, the investigation focuses on the CICIND provisions, which are compared with
the results obtained numerically.
3 NUMERICAL MODEL
For the numerical model the well-known finite element software ADINA is used, which has already
been employed by other researchers for the evaluation of the structural response of many types of
structures, including shell structures. Therefore, it can be considered as an appropriate tool for the
achievement of the goals of the present research work. As shown in Figure 3-1, in the numerical
model the horizontal plane is XY (transverse to the chimney axis) and the vertical axis is Z (parallel to
the chimney axis).
For the modeling of the shell structure 4-noded shell elements are used, as it is considered that the
thickness of the shell is much smaller than its other two dimensions. It is well-known that shell
buckling is a mesh-dependent problem. The fact that in tall shell structures subjected to wind loading
the region of failure is not known in prior (Teng and Rotter, 2004) makes it necessary to keep the
mesh size constant throughout the whole structure. Additionally, as the employed model is based on
an actual chimney and is subjected to realistic types of loading, using a half-model or making
assumptions for scaling down the structure is considered inappropriate.
During the first stage of the analyses a mesh sensitivity investigation takes place, varying the mesh
size from 25cm to 7cm (25cm, 20cm, 18cm, 15cm, 12cm, 10cm and 7cm) and keeping it constant
throughout the entire structure. The smallest considered mesh size (7cm) leads to very
computationally demanding analyses and any further decrease is not possible. Nevertheless, as will be
shown by the following numerical investigation, a mesh size of 7cm is considered as sufficient for the
purposes of the present work.
The boundary conditions at the bottom and at the top of the shell structure are modeled by using
rigid links, connecting all nodes of the respective cross-section to corresponding master nodes, at
which the supports and/or loads are applied, as shown in Figure 3-2. Hence, the distances between
nodes belonging to the bottom and top chimney sections remain constant, these sections do not
deform, and are forced to retain their circular shape. As mentioned above, the thickness of the shell is
equal to 13mm (including corrosion allowance), while the material has an elastic modulus equal to
20250kN/cm2 and a Poisson ratio equal to 0.3. In elastoplastic analyses, and unless it is stated
otherwise, the yield stress is taken equal to 16kN/cm2 and no strain hardening is taken into account.
Figure 3-2: Rigid links used at the top cross-section of the shell structure
In this section, the load cases that are used in the context of the present research work are
presented. They can be summarized as:
In this loading case uniform global bending is encountered due to end bending moments of opposite
direction. Based on beam theory the bending moment is constant along the structure.
LC4 deals with pure compression due to vertical loads-weights applied at different levels. The values
of the vertical loads used are the weights of platforms and equipment placed at different levels. Based
on beam theory the maximum axial force develops between the lowest level’s axial load and the
bottom support.
LC5 also deals with pure compression due to the vertical loads of LC4 applied at different levels and
the shell self-weight. According to beam theory the maximum axial force develops at the bottom.
In this loading case vertical loads at different levels, the self-weight of the shell and other parts of the
structure as well as wind loading applied normally to the shell are considered. Wind is assumed to act
in the +X-direction. In the case that openings exist, the additional weight of the inlet duct is also
included. This type of loading is considered to be the most realistic one and would be normally used
for the analysis and design of such structures. Based on beam theory, the maximum normal stress
due to axial force and bending moment is encountered at the bottom.
3.2.7 Combined loading including a concentrated wind load at the top (LC7)
This type of loading includes vertical loads at different levels, the self-weight of the shell and a lateral
concentrated load in the +X-direction applied at the top of the structure, having a magnitude equal to
the total shear force caused by the wind load of LC6. A loading of this type would be commonly used
for the analysis and design of wind turbine towers, in which the lateral load would be primarily due to
wind actions on the blades, hence it would be concentrated at the top of the tower. LC7 is considered
for reasons of assessment of the importance of wind pressure applied normally to the shell surface.
Based on beam theory, the maximum axial force and bending moment would also develop at the
bottom.
3.2.8 Combined loading including a concentrated wind load at the top (LC8)
This loading consists of vertical loads at different levels, the self-weight of the shell structure and a
lateral concentrated load applied at the top of the structure having a magnitude such that it causes a
bending moment at the bottom support equal to the one caused by the actions of LC6. The lateral
load is applied in the +X-direction. This loading is similar to the one presented in section 3.2.7.
The main types of analyses that are used in the present research work are the following:
This commonly used and simple type of analysis refers to the perfect structure with elastic material. It
is used in the context of the present research work only for the observation of the distribution of
stresses. The equilibrium conditions are based on the undeformed configuration.
In this type of nonlinear analyses, equilibrium is based on the deformed configuration. No material
nonlinearity is considered and for this reason the collapse load cannot be directly calculated. The
results are presented by making use of equilibrium paths that show the relation between the applied
load and the resulting deflections.
In this type of nonlinear analyses, equilibrium conditions are based on the undeformed configuration
and only material nonlinearity is considered. The results are presented by using equilibrium paths as
in the case of GNA. The material nonlinearity used for the shell structure is based on the flow
plasticity rule.
This type of analysis includes both geometrical and material nonlinearity, either considering also initial
imperfections (GMNIA) or not (GMNA). If properly employed, this type of analysis can result in the
most accurate prediction of the response of the structure and of the collapse load. Material
nonlinearity is also based on the flow plasticity rule.
The computational effort required for the three types of nonlinear analysis is much larger than the one
for LA and LBA.
4.1 INTRODUCTION
In this chapter, the unstiffened chimney shell structure is investigated, fixed at the support and having
a rigid ring at the top. Although such an unstiffened chimney is not commonly encountered in
practice, many useful conclusions can be drawn by performing appropriate parametric studies and
analyses using it. The numerical procedure used in this investigation is based on the guidance
provided by Gantes and Fragkopoulos (2010). Both linear and nonlinear analyses are performed and
the results are presented by using appropriate equilibrium paths, characteristic deformed shapes and
snap-shots of plastification distribution.
The shell element size plays an important role in the buckling loads and mode shapes. In this section,
LBA is used in order to investigate this effect and to obtain the shell element size that can be
considered as sufficient for the purposes of the present research work. The results are summarized in
Figure 4-1. On the vertical axis, the ratio of the numerical buckling load to the corresponding
analytical one is plotted. On the horizontal axis, the shell element size is plotted, ranging between a
maximum value of 25cm and a minimum one of 7cm.
The analytical buckling load factor λb is evaluated by making use of the basic assumptions of beam
theory and the classical buckling stress as:
(N/A+M/W)λb=σcr (4-1)
where N is the axial force at the base support and M is the bending moment at the same location for
the design loads without any safety factors. The other parameters have already been defined in the
previous sections. The classical buckling stress for the case investigated has been found in section
2.3.2 to be equal to 455MPa. From the description of load cases it is deduced that in LC1, LC4 and
LC5 only an axial force N develops, in LC2 and LC3 only a bending moment M, while in case LC6 both
a bending moment and an axial force are encountered. Loading cases LC7 and LC8 are not included in
this mesh sensitivity investigation.
From Figure 4-1 it is concluded that as the mesh size is reduced, the numerical results approach the
analytical ones for all cases, except for LC6. It is also seen that the mesh size that is required for each
loading case to approach the analytical results is different. Additionally, it is observed that in LC6 the
numerical buckling loads are always larger than the analytical ones, and this is more pronounced for
small shell element sizes. A separate description of the results for each load case is presented in the
following sections.
Figure 4-1: Variation of the ratio between the numerical and the analytical buckling loads for various element
sizes and load cases
From the buckling mode shape (Figure 4-2) it is observed that buckling takes place over the entire
height of the cylinder, as could be expected due to the constant axial force diagram induced by the
concentrated load at the top. Close to the supports the lateral deflections of the buckled shell are
reduced due to the effect of boundary conditions.
Figure 4-2: First buckling mode of the shell structure for LC1 using 7cm element size
In LC2 the chimney is subjected to uniform global bending due to the application of end bending
moments of opposite direction. From the buckling mode shape (Figure 4-3) it is observed that
buckling develops along the compressed side of the cylinder and becomes more prominent far from
the supports, which is in accordance with beam theory. The observed wavelength is significantly
shorter than the one corresponding to LC1. A bending moment approximately equal to 23180kNm is
required for buckling. In Figure 4-1, the convergence of the critical bending moment to the theoretical
one (which is based on the critical buckling stress σcr and beam theory), as the element size is
reduced, is verified. It is noted that the mesh size required for convergence of the critical bending
moment is less than 10cm. Making use of a mesh size equal to 25cm, which proved to be sufficient for
LC1 (axial compression), leads to significant overestimation of the buckling moment.
Figure 4-3: First buckling mode of the shell structure for LC2 using 7cm element size
Figure 4-4: First buckling mode of the shell structure for LC3 using 7cm element size
In LC4 pure compression develops in the chimney structure due to the application of vertical loads at
different levels, representing the weights of attached structural components. A load factor
approximately equal to 277 is required for the 1st buckling mode, corresponding to an axial force at
the bottom of the chimney equal to 130000kN. Buckling appears along the lowest part (Figure 4-5)
and is similar to the diamond-shaped buckling pattern observed in LC1. The main difference between
the two load cases is that the part of the cylinder that is more susceptible to buckling has a smaller
length in LC4, and therefore the effect of the boundary conditions is larger when compared with LC1.
Nevertheless, as in LC1, a mesh size equal to 25cm seems to be satisfactory for the purposes of
engineering practice.
Figure 4-5: First buckling mode of the shell structure for LC4 using 7cm element size. The vertical loads
applied at different levels are also presented with red colour.
In LC5 also pure compression develops in the chimney structure due to the application of vertical
loads at different levels and self-weight. In this case, due to the self-weight that is uniformly
distributed along the structure, the maximum axial force develops at the bottom, and this is clearly
demonstrated by the buckling mode in Figure 4-6. A load factor approximately equal to 75 is required
for the 1st buckling mode to occur. The load factor of LC5 is smaller than the one corresponding to
LC4, as expected, due to the addition of the self-weight. This load factor corresponds to an axial force
approximately equal to 136000kN. The wavelengths are relatively small, as in LC2 and LC3, and a
mesh size less than 10cm seems to be necessary for this type of loading.
Figure 4-6: First buckling mode of the shell structure for LC5 using 7cm element size and detail
In this load case vertical loads applied at different levels, self-weight and wind pressure are combined.
This is the first load case, in which pressures normal to the surface are applied. The buckling mode
shape is shown in Figure 4-7 and it is observed that buckling occurs at the top of the structure and at
a location that does not correspond to the more compressed side according to beam theory (see
Figure 4-8). A buckling load factor approximately equal to 12.3 is required for the 1 st buckling mode to
occur. It is pointed out that the use of beam theory leads to larger buckling load factors for all mesh
sizes considered and there is a significant difference from the numerical results. Therefore,
considering that the analytical method, which is based on beam theory, is a rough simplification of the
much more complex real situation, it is deduced that its use is inappropriate for LC6. Results that will
be presented in the following comparison between unstiffened and stiffened shell structures will shed
more light on this observation.
Figure 4-7: First buckling mode of the shell structure (local buckling at the top) for LC6 using 7cm
element size
Figure 4-8: Wind direction and region in which failure is predicted based on beam theory
The reason behind this differentiation can be investigated by focusing on linear elastic analyses, in
order to observe the way, in which the structure deforms under wind pressure, and compare it with
the case that the wind load is applied as a concentrated load at the top of the structure (LC7). The
results are summarized and compared in Table 4-1. It is interesting to observe that the cross-sectional
shape of the unstiffened shell structure under the application of LC6 is significantly modified and is not
circular. The largest modification of the shape takes place far from the supports where the boundary
conditions (at the bottom) and the rigid ring (at the top) restrict changes of the cross-sectional shape.
The fact that the shape of the cross-section changes is clearly illustrated in the stress distribution, too.
On the other hand, if the wind load is applied as a concentrated load at the top (LC7), the cross-
sectional shape is not significantly affected and the stress distributions are very close to the ones that
classical beam theory would predict.
Therefore, it can be concluded that in the case that wind pressures are directly applied to the thin-
walled shell surface of the unstiffened chimney, the cross-sectional shape is significantly changed and
the stress distributions are very different from the ones predicted by beam theory assumptions. As a
result, the numerical buckling load is very different from the one found analytically. Additionally, the
location of buckling is different from the one predicted analytically.
Table 4-1: Comparison between unstiffened chimney’s response under LC6 and LC7
3D View
Plan view
Circumferential stress
distribution
Longitudinal stress
distribution
N is the axial force at the base support and M is the bending moment at the same location for the
design loads without any safety factors. The other parameters have already been defined in the
previous sections. The failure stress obtained with CICIND provisions does not contain any safety
factors, so that it is comparable with the numerically obtained results. The numerical collapse load
factor is relatively close to the one found by using the provisions of CICIND (≈3.76) although they are
not totally comparable, as the latter includes initial imperfections of specific patterns.
In Figure 4-10, a detail of the equilibrium paths is presented in the region in which the GMNA’s
equilibrium path reaches its maximum load factor. It is observed that plastification is encountered for
a smaller load factor than the one corresponding to failure according to CICIND (decrease of the
structure’s stiffness in the case of MNA and divergence of MNA and GNA). This means that the CICIND
load factor is not associated with the first yield in this specific case according to the numerically
obtained results. Nevertheless, it remains on the safe side, being smaller than the collapse load factor
found with GMNA.
Figure 4-9: Equilibrium paths for the unstiffened structure for load case LC6 from different types of analysis
The magnified deformed shape of the unstiffened structure at failure from GMNA is shown in Figure 4-
11. It is interesting to note that the deformed cross-sectional shape is significantly different from the
undeformed circular one, due to the wind pressure application. The application of pressures directly
on the surface of the shell structure leads to deformation of the thin-walled shell that is associated
with a three-dimensional response, which cannot be captured by beam theory.
Overall bending deformation as a cantilever, without cross-sectional shape changes, thus with
the cross-sections remaining circular.
Local deformation of the surfaces, associated with significant modification of the cross-
sectional shape, thus with the cross-sections losing their circular shape.
Figure 4-10: Detail of equilibrium paths for the unstiffened structure for load case LC6 from different types of
analysis
It is also interesting to note that the deformed shape is significantly different from the buckling mode
shape found by using LBA. This fact is one of the reasons, for which the appropriateness of selecting
buckling mode shapes as initial imperfection patterns is open to debate. More discussion on the
influence of imperfections will be presented in a following section. The distribution of plastification
over the shell structure at failure is presented in the views of Figure 4-12. It can be observed that
plastification is encountered in several regions at the top and bottom of the chimney, in contrast to
what beam theory would predict. According to beam theory the failure of a cantilever subjected to
these loads would occur at the base.
Figure 4-11: 3D view and plan view of deformed shapes of the unstiffened structure at failure from GMNA for
load case LC6
Figure 4-12: Plastic and elastic regions of the unstiffened structure at failure from GMNA for load case LC6
The selection of initial imperfections in cases of wind loads is even more complicated due to the fact
that buckling modes obtained by LBA may have different shapes from the deformed configurations
obtained with nonlinear analyses, as shown in previously investigated cases. Moreover, it should be
pointed out that the deformation of the shell under the application of the wind pressure acts itself as a
type of imperfection. Nevertheless, whether this load-induced type of imperfection is sufficient for
practical purposes or not, is unclear and not yet resolved.
Greiner and Derler (1995) studied the effect of imperfections on wind loaded cylindrical shells. They
used different types of initial imperfections including local shapes, global shapes (specific initial
ovalization of the cross-section), combined shapes (including local and global shapes) and
eigenmodes. One of their main conclusions is that a great variety of imperfection forms should be
studied before safe conclusions can be drawn and the actual sensitivity of shells can be reliably
assessed.
In this section, the results for the unstiffened structure are presented. The results are summarised in
Figure 4-13 for imperfections according to the 1st buckling mode shape due to LC1. It can be seen
that the effect of imperfections on the structural response is negligible and that the equilibrium paths
obtained from GMNA and GMNIA are almost identical. The deformed shapes and distributions of
plastification in the unstiffened structure at collapse load level for imperfection’s magnitude equal to
0.5t are shown in Table 4-2. It is concluded that they are not very different between each other and
no significant local buckling appears at any location for this specific comparison.
The results for imperfections according to the 1st buckling mode shape due to LC6 are presented in
Figure 4-14. It is observed that the insertion of this shape of initial imperfection leads to a decrease in
the collapse load factor. For the largest investigated imperfection the decrease is approximately equal
to 10%. It should be pointed out that in the specific cases investigated the material yield stress is
relatively low and reduces the detrimental effect of geometrical nonlinearity and imperfections.
The result obtained by using CICIND provisions (and the formulas provided for imperfection
magnitude less than 0.01 ℓ) is satisfactory and close to the numerically obtained collapse load factors.
The deformed shapes and plastification distributions of the unstiffened structure at collapse load level
for imperfection’s magnitude equal to 0.5t are shown in Table 4-2.
Figure 4-13: Comparison between equilibrium paths obtained from GMNA and GMNIA for the unstiffened
structure with imperfections based on the 1st buckling mode shape due to LC1
Figure 4-14: Comparison between equilibrium paths obtained from GMNA and GMNIA for the unstiffened
structure with imperfections based on the 1st buckling mode shape due to LC6
Table 4-2: Comparison between unstiffened chimney’s response under LC6 for imperfection magnitude equal to
0.5t
3D deformed
view
Deformed view
in plan
3D view of
elastic and
plastified
regions
Plan view of
elastic and
plastified
regions
In this section, the efficiency of CICIND is investigated by making comparisons with numerically
obtained results for LC6, LC7 and LC8. The numerical procedures included both GMNA and GMNIA and
can provide a reliable estimation of the structural response and collapse load factor. For this reason,
they are considered as the comparison basis. The structure in all cases is considered to be
unstiffened.
The difference between the three load cases is in the way in which the horizontal loading is applied.
LC6 deals with wind loading applied directly to the thin shell surfaces as pressure, while LC7 is related
to a concentrated lateral force, at the top of the structure, equal to the shear force caused by the
wind loading of LC6. LC8 is related to a concentrated lateral force, at the top of the structure, which
causes a bending moment at the support, equal to the one caused by the wind loading of LC6.
It is reminded that the collapse load factor according to CICIND is calculated as:
In this section, the CICIND results obtained for LC6, LC7 and LC8, are compared with the ones
obtained with GMNA. The CICIND provisions account for initial imperfections, while in GMNA no
imperfections are considered. Nevertheless, the application of the wind pressure directly on the shell
surface in LC6 induces a type of imperfection due to the deformation of the shell. This type of
imperfection is not present in LC7 and LC8.
The results are presented in Figure 4-15. According to CICIND, LC6 and LC8 lead to the same collapse
load factor as the axial force and bending moment at the base are the same for these load cases.
Nevertheless, GMNA results show that this is not the case and that the capacity of LC8 is larger than
the one found for LC6. This implies that the concept behind the analytical calculations is not in
accordance with GMNA results. In LC8, failure takes place at the bottom (which is in accordance with
the analytical method) and therefore probably the conservatism behind the analytical predictions lies
in the calculation of the failure stress by CICIND. In LC6, failure takes place at many locations and no
conclusion can be drawn as far as the efficiency of the analytical method is concerned.
Figure 4-15: Comparison between equilibrium paths obtained from GMNA for the unstiffened structure for LC6,
LC7 and LC8 and analytical collapse load factors
The collapse load factor according to CICIND for LC7 is much smaller than the collapse load factor for
LC6 and LC8. It is also smaller than the one calculated numerically with the use of GMNA. As in LC8,
in LC7 failure takes place in the region predicted by the analytical method, however the analytical
method is conservative, which is again attributed to the conservatism of the strength proposed by
CICIND.
In this section, the CICIND results obtained for LC6, LC7 and LC8, are compared with the ones
obtained with GMNIA. The CICIND provisions account for initial imperfections and in GMNIA
appropriate imperfections are incorporated. The results are presented in Figure 4-16. It can be seen
that the numerical results are much closer to the analytical ones, when compared with GMNA. In LC7
and LC8, the conservatism of the analytical procedure may be attributed to the conservatism of the
failure stress of CICIND. In LC6, the failure predicted by the numerical model differs from the one
found with the analytical procedure and the excellent accuracy of the latter, as far as the collapse load
factor is concerned, is probably a coincidence.
Figure 4-16: Comparison between equilibrium paths obtained from GMNIA for the unstiffened structure for LC6,
LC7 and LC8 and analytical collapse load factors
4.6 CONCLUSIONS
Several conclusions can be drawn from the above numerical investigation and most of them are
thoroughly presented in the text. The main conclusions that refer to the specific structure and loading
types of the present research work are:
The stress distribution does not affect significantly the normal buckling stress. Whether the
stresses are due to axial forces or due to bending moments, the buckling stress does not
change significantly, a conclusion drawn by other researchers, too (Rotter et al., 2014).
Therefore, the buckling stress of axial compression could be applied to other load cases also.
The mesh size has an important effect on the evaluation of the structural response of shell
structures. It was observed that a mesh size that may prove to be sufficient for a type of
loading may not be adequate for another type of loading. The smaller the buckling
wavelength is, the smaller the mesh size required for adequate accuracy to be achieved.
In most loading cases, the buckling load (or load factor) was satisfactorily predicted by
making use of the simplified assumptions of beam theory and the buckling stress due to axial
compression. These were the cases, in which the cross-sectional shape did not change
significantly, so that the main assumptions of beam theory were valid. The locations at which
buckling occurred were those of maximum compressive stresses according to beam theory.
In the case of direct application of wind pressure to the perimeter of the thin-walled shell
structure the cross-sectional shape was significantly modified. For this reason, beam theory
was not sufficient, leading to results that were unsafe. The locations at which buckling
occurred were not the ones predicted by beam theory, appearing across the wind direction.
Nevertheless, the collapse load factor predicted by CICIND was relatively close to the
numerical one obtained with GMNA.
Selecting an appropriate imperfection pattern and magnitude is still open to debate in the
scientific community investigating the structural response of shells. Considering in nonlinear
analyses imperfections in areas of high stresses (imperfections based on LBA and in
accordance with the load case used for the nonlinear analyses) deteriorates the structural
response.
The collapse load factor predicted by CICIND was very close to the numerical one obtained
with GMNIA in LC6. However, this conclusion cannot be generalized due to the fact that the
fundamental assumptions of CICIND’s analytical approach cannot be valid, as the analytical
types of failure are different from the numerical ones.
LC6 and LC8 collapse load factors should have been equal according to the analytical method
but the numerical results showed an important difference between them. Therefore, treating
LC6 as LC8 cannot be acceptable for design purposes.
5.1 INTRODUCTION
In many cases, stiffening rings are used in industrial chimneys in order to increase the capacity of the
structure. The distance between successive stiffening rings may vary. It is also common practice to
use stiffening rings at levels at which platforms and other types of distributed loads are encountered.
It is commonly accepted that the use of stiffening rings provides restraint against changes of the
cross-sectional geometry (ovalization). On the other hand, no significant increase of the bending
stiffness is achieved with the use of circumferential rings. Nevertheless, there are no specific code
provisions for the analysis and design of shell structures that are strengthened by making use of
circumferential stiffeners. The numerical and experimental works on stiffened structures are also
limited.
The stiffened structure considered in the present section consists of the same shell as the one of the
unstiffened structure of the previous section, and of the stiffening rings, that are assumed to have
either an angle cross-section or rigid properties. Each stiffening ring is connected with the shell
through one of its tips. The results for various distances between stiffening rings will be investigated
in comparison to a basic case in which the distance between successive stiffeners is equal to 5m. The
ring stiffeners are also modeled with the use of shell finite elements. A view of the stiffened shell
(green colour) and a detail of a ring stiffener (red colour) running along the perimeter of the shell are
shown in Figures 5-1 and 5-2, respectively.
It should be pointed out that the results that will be presented in this chapter are obtained with a
mesh size of 7cm that was found to be sufficient for the purposes of the present research work. This
mesh size was obtained with a mesh sensitivity investigation that is not shown in detail for brevity.
Figure 5-1: Stiffened shell structure. The green color is used for the shell surface while the red one for the
stiffening rings
Figure 5-2: Detail of a stiffening ring (red color) running along the perimeter of the considered shell structure
(green color)
In this section, the effect of the stiffening rings on the buckling load is investigated by making use of
LBA. The load cases that were presented in Chapter 4 are also used in this section. The stiffened
structure has stiffeners with L120X120X10 angle cross-sections (for the sake of simplicity described as
L120X10 in the figures) that are placed at a distance equal to 5m between each other.
The effect of the stiffening rings can be easily detected if the ratio of the buckling load factor of the
stiffened structure to the one of the unstiffened structure is evaluated for the load cases considered.
The results are presented in Figure 5-3. It can be seen that the stiffening rings do not modify
significantly the buckling load in LC1, LC2, LC3, LC4 and LC5. These five cases are the ones in which
the analytical method, which is based on the simplified assumptions of beam theory, proved to be
sufficient in the previous section. On the contrary, the stiffening rings increase significantly the
buckling strength of the structure in LC6. Beam theory proved to be insufficient for the satisfactory
evaluation of buckling strength in LC6.
Figure 5-3: Ratio of the buckling loads of the first buckling mode of the stiffened structure to the one of the
unstiffened structure for different load cases and based on mesh size equal to 7cm
It is thus concluded that in cases in which beam theory is sufficient, the rings do not contribute
significantly to the buckling load of the first buckling mode, while in the case that beam theory is
insufficient, the stiffening rings lead to an increase in the buckling load. This can be explained by
considering that the use of beam theory does not take into account any possible changes of the cross-
sectional shape. As a result in the cases that beam theory is sufficient, there are no significant
changes of the cross-sectional shape and the rings have no important effect. On the other hand, in
the load case that the cross-sectional shape changes significantly, the beam theory is not valid and
the rings restrict those changes, thus leading to an increase in the buckling load factor. Of course, a
more dense arrangement of the stiffening rings may affect the buckling load of the other load cases,
too. But this restriction seems to be more obvious in LC6, in which wind pressure is directly applied to
the shell structure.
In Table 5-1, the first buckling mode shapes obtained for shell element size equal to 7cm for the six
load cases are presented. It is interesting to note that in LC6 the location of buckling of the first
buckling mode is modified due to the effect of the stiffening rings. Nevertheless, in both cases the
location of buckling does not correspond to the more compressed side (see Figure 4-8), as expected
according to beam theory for wind direction +X. Therefore, it is believed that the applied pressures to
the walls of the shell structure are of great importance. A similar observation was made by Greiner
and Derler (1995).
Table 5-1: Comparison between the 1st buckling mode shapes found for the unstiffened and the stiffened shell
structures
LC1: Axial
compression
LC2: Uniform
bending
LC3: Non-
uniform
bending
accompanied
by shear force
LC4: Attached
weights
LC6: Combined
loading
including wind
In order to better understand the predictions of LBA for Load Case 6, one may have to look into the
stress distribution in the structure based on a linear elastic analysis. To that effect, linear elastic
analysis results are presented for a load factor equal to 1.00 (for a load combination that could
correspond to serviceability limit state). LBA does not account for prebuckling deformations and is
based on the maximum stress for the evaluation of the buckling modes and loads. In Table 5-2, the
results obtained from linear elastic analysis under the application of vertical loads and wind loads are
presented. It is observed that the unstiffened shell deforms as a cantilever but also deformations of
the thin walls are encountered, leading to changes of the cross-sectional shape. As a result the
circumferential normal stresses have large compressive and tensile peaks. It is also worth mentioning
that large compressive stresses appear at the location, at which LBA predicted buckling for the 1st
buckling mode. The presence of stiffening rings reduces the deformation of the shell structure and the
changes of the cross-sectional shape are significantly mitigated. This has an effect on the
circumferential and longitudinal normal stress distribution, as shown in Table 5-2.
Table 5-2: Comparison between the deformed shapes and stress distributions found for the unstiffened and the
stiffened shell structures based on linear elastic analysis under the application of vertical loads and wind loading
(LC6)
3D View
Plan view
Circumferential
stress
distribution
Longitudinal
stress
distribution
It is interesting to make the same comparison for LC3, consisting of a concentrated lateral load
applied at the top of the structure, in which according to LBA the stiffening rings did not have a
significant effect on the buckling load. The results are presented in Table 5-3. It can be concluded
that neither the deformed shapes nor the normal stress distributions are significantly modified due to
the presence of stiffening rings. It is also shown that the large compressive longitudinal stresses
develop at the bottom, where LBA predicted buckling of the shell for the 1st buckling mode.
Table 5-3: Comparison between the deformed shapes and stress distributions found for the unstiffened and the
stiffened shell structures based on linear elastic analysis under the application of a concentrated lateral load at
the top (LC3)
3D View
Plan view
Circumferential
stress
distribution
Longitudinal
stress
distribution
In this section, load case LC6 is considered. The results for the stiffened structure are presented, by
making use of equilibrium paths, in Figure 5-4. It can be observed that failure is mainly governed by
plastification, as the collapse load factor found with MNA is much smaller than the one found with
GNA. The deformed shapes are depicted in Figure 5-5.
Figure 5-4: Equilibrium paths for the stiffened structure from different types of analysis based on LC6
Figure 5-5: 3D view and plan view of deformed shapes of the stiffened structure at failure from GMNA based on
LC6
The spread of plastification in the structure at failure is presented in Figure 5-6. It can be seen that
plastification is distributed at many locations throughout the structure. Apart from the shell, the
stiffening rings are also plastified.
Figure 5-6: Plastic and elastic regions of the stiffened structure at failure from GMNA based on LC6
A comparison between the structural response of the unstiffened and stiffened shell structures is
shown by the equilibrium paths in Figure 5-7 and a close view of them is presented in Figure 5-8. It is
noted that the elastic lateral stiffness of the structure under the application of this type of loading is
not significantly affected by the presence of the stiffening rings. This comes as a verification of the
view that the cantilever component of deformation, which dominates the initial response, is not
affected by the rings.
On the other hand, the bearing capacity of the structure is significantly enhanced by the stiffening
rings, as illustrated by the results obtained with LBA, GNA, MNA and GMNA. The largest increase is
observed for the cases that the material is assumed to be elastic. GNA of the stiffened structure
predicts a collapse load factor that is approximately 2.8 times larger than the one corresponding to
GNA of the unstiffened structure. The collapse load factor based on GMNA is approximately 1.35 times
larger in the case of stiffeners. Obviously, the limitation imposed by material yielding decreases the
effectiveness of the stiffening rings. It is therefore concluded that use of higher material yield strength
would enhance the beneficial effect of stiffening rings on the capacity.
Figure 5-7: Comparison of equilibrium paths of unstiffened and stiffened shell structure obtained for LC6
Figure 5-8: Close view of the comparison of equilibrium paths of unstiffened and stiffened shell structure obtained
for LC6
In LBA it was seen that the effectiveness of the stiffening rings spaced at a distance equal to 5m is
very small in load cases that do not induce significant changes in the cross-sectional shape. In these
cases, the simplified assumptions of beam theory can be applied, leading to satisfactory estimations of
the buckling load. This effectiveness is investigated in this section based on nonlinear analyses.
To that effect the results obtained with GMNA are presented and compared for LC6 and LC7 in Figure
5-9. It can be seen that, while in the case of pressure (wind loading) the stiffening rings increase the
capacity, in the case of a concentrated lateral load at the top, their effect on the overall strength is
negligible. This is attributed to the fact that the lateral load, when applied as concentrated at the top,
does not lead to significant changes of the cross-sectional shape and failure takes place at the bottom
support, exactly at the location predicted by the classical beam theory. In both cases, as already
stated previously, the lateral stiffness is not affected by the presence of stiffening circumferential
rings.
Figure 5-9: Comparison between equilibrium paths obtained with GMNA for stiffened and unstiffened structures
for load cases LC6 and LC7
Although the use of very stiff and elastic rings is not a very realistic option, it can be proved useful for
identifying the effectiveness of the specific ring cross-section (L120x120x10) that is primarily
employed in the present investigation. It is expected that very stiff and elastic rings prohibit cross-
sectional deformations to a very significant extent. The results obtained with LBA, GNA, MNA and
GMNA are presented in this section. The load case used is LC6.
The buckling load factors are presented for all investigated cases in Figure 5-10. The stiffened
structures reach buckling load factors that are larger than the one corresponding to the unstiffened
shell structure and the effect of the stiffening rings is obvious. The 1st buckling mode shape of the
rigidly stiffened structure is shown in Figure 5-11. It can be seen that the location at which local
buckling appears coincides with the one at which beam theory and analytical calculations ignoring
cross-sectional deformations would predict failure for +X-direction of the wind.
The analytical buckling load factor is approximately equal to 16. This value is quite close to the
numerically obtained value for the buckling load factor of the rigidly stiffened structure and is smaller
than it, as the latter is probably influenced by the presence of boundary conditions at the bottom of
the shell. The use of beam theory in combination with the buckling stress due to axial compression
(455MPa) leads to a satisfactory estimation of the buckling load factor in the case of rigidly stiffened
structure, as the rigid rings significantly restrict changes of the cross-sectional shape.
It is quite surprising though that the buckling load factor of the rigidly stiffened structure is smaller
than the one of the flexibly stiffened structure (L120X120X10 cross-section for the rings). Apart from
the values of the buckling load factors, the locations of buckling also differ between these two cases,
as shown in Figure 5-12. In the flexibly stiffened structure local buckling appears towards Y-direction
(cross-wind), while in the case of rigidly stiffened structure local buckling appears towards the +X-
direction (along-wind).
Figure 5-10: Comparison of the buckling load factors of unstiffened, stiffened and rigidly stiffened shell structure
obtained with LBA and based on LC6
Figure 5-12: First buckling mode shapes of the flexibly stiffened structure (left) and of the rigidly stiffened
structure (right)
Figure 5-13: Comparison of the equilibrium paths of unstiffened, stiffened and rigidly stiffened shell structures
obtained with GNA and based on LC6
The deformed shape of the rigidly stiffened structure is shown in Figure 5-14. It can be seen that the
deformation of the rigidly stiffened structure is associated with global bending (as cantilever) and local
buckling at the base. The location of buckling and the buckling waves are similar to the ones predicted
by LBA. Hence, prebuckling deformations (that are not taken into account by LBA) do not modify
significantly the overall response. The thin shell walls do not deform significantly under the application
of normal pressure, as such deformation is prevented by the rigid stiffening rings. Thus, the maximum
normal stresses develop at the bottom, at the compressed along-wind side.
Figure 5-14: Deformation and local buckling at the base of the rigidly stiffened structure for GNA and LC6
The deformed shape of the flexibly stiffened structure is presented in Figure 5-15. The deformed
shape is associated with global bending (as cantilever), as well as change of the cross-sectional
shape. It is shown that the application of normal pressure on the shell walls causes change of the
cross-sectional shape. The flexible stiffening rings reduce significantly this change in comparison to
the unstiffened case but do not fully eliminate it, as in the case of rigid stiffening rings. At failure there
is no obvious presence of significant local buckling deformations at a specific location, as in the case
of rigid stiffeners, except for a slight “elephant-foot” type of deformation at the cross-wind side, near
the bottom.
Figure 5-15: Deformation and close view of the base of the flexibly stiffened structure for GNA and LC6
It is hence observed that the two stiffened structures deform and fail in a different way, which may
explain the difference of the collapse load factors observed. Further analyses are required towards this
direction, in search of an optimum solution for the cross-section of the stiffening rings.
The results obtained from MNA are presented in Figure 5-16. It can be seen that all three structures
behave elastically at the beginning of the analyses. Then, yielding appears in all cases and the
stiffness decreases until the collapse load is reached. The unstiffened structure yields at the lowest
load factor, while the stiffened structure with rigid elastic rings at the highest. The overall capacities of
the two stiffened structures do not differ significantly, however the flexibly stiffened structure deforms
more at lower load levels.
The analytical evaluation of the collapse load factor according to the simplified assumptions of beam
theory and by ignoring buckling and geometrical nonlinearity has an upper and a lower bound. The
upper bound is obtained based on the full plastic resistance of the circular cross-section (numerically
predicted, as shown in Figure 5-17):
M=Mpl(1-n1.7), where Mpl is the full plastic moment of resistance and n is the ratio of the axial force to
the axial cross-sectional strength.
The lower bound is obtained by considering that failure takes place when the more stressed fiber
reaches the yield stress:
N/A+M/W=16kN/cm2
Figure 5-16: Comparison of the equilibrium paths of unstiffened, stiffened and rigidly stiffened shell structures
obtained with MNA and based on LC6
Figure 5-17: Plastification of the bottom cross-section of the stiffened shell structure with stiff and elastic rings at
failure
It is thus observed that the use of beam theory satisfactorily predicts the ultimate load factor (upper
bound) for both cases of very stiff as well as flexible rings. A similar conclusion is drawn for the load
factor associated with the first yield, which is close to the numerically obtained load factor, denoted by
a green star.
The results obtained from GMNA are presented by the equilibrium paths plotted in Figure 5-18. It can
be observed that the use of rigid and elastic rings does not increase the overall capacity in this specific
case in comparison to the flexible, elastoplastic rings. The rigidly stiffened shell structure behaves
elastically up to the collapse load. Then, yielding takes place and no further increase of the load is
possible.
Figure 5-18: Comparison of equilibrium paths of unstiffened, stiffened and rigidly stiffened shell structures
obtained with GMNA and based on LC6
The deformed shape of the rigidly stiffened structure is depicted in Figure 5-19. It can be seen that
the rigid rings prevent any changes of the cross-sectional shape. Therefore, the cantilever component
of deformation prevails, hence yielding and failure appear at the base and on the compressed side, as
predicted by beam theory (Figure 5-20).
Figure 5-19: 3D view and plan view of deformed shapes of the rigidly stiffened structure at failure
Figure 5-20: Plastic and elastic regions of the rigidly stiffened structure at failure
In this part, the results for the flexibly stiffened structure with initial imperfections are presented. The
stiffening of the structure is achieved by using angle L120X120X10 cross-sections placed at a distance
equal to 5m between each other (as described in previous sections). The initial imperfection patterns
are based on the 1st buckling mode shape due to compression only (LC1) or on the 1st buckling mode
shape due to wind pressure and weights (LC6).
The results are summarised in Figure 5-21 for imperfections according to the 1st buckling mode shape
due to LC1. It is observed that the effect of imperfections on the structural response is negligible and
that the equilibrium paths obtained from GMNA and GMNIA are almost identical. The deformed shapes
and plastification distributions of the stiffened structure at collapse load level for imperfection
magnitude equal to 0.5t are shown in Table 5-4. It can be seen that they are not very different
between each other. It should be also pointed out that in the case of initial imperfections due to LC1
buckling waves are observed at the midheight of the structure.
The results for the stiffened structure and 1st buckling mode shape due to LC6 are presented in Figure
5-22. It is shown that the insertion of initial imperfections leads to a decrease in the collapse load
factor. In the case of the maximum initial imperfection with amplitude 0.65t the decrease of the
collapse load factor due to initial imperfections is approximately equal to 13%. The deformed shapes
and distributions of plastification in the stiffened structure at collapse load level for imperfection
magnitude equal to 0.5t are shown in Table 5-4. The magnified local buckling appearing at the bottom
of the shell structure in the case of imperfections based on mode shape due to LC6 is clearly
illustrated.
Figure 5-21: Comparison between equilibrium paths obtained from GMNA and GMNIA for the stiffened structure
(L120X120X10) based on the 1st buckling mode shape due to LC1
Figure 5-22: Comparison between equilibrium paths obtained from GMNA and GMNIA for the stiffened structure
(L120X120X10) based on the 1st buckling mode shape due to LC6
Table 5-4: Comparison between stiffened chimney’s response under LC6 for imperfection magnitude equal to 0.5t
3D View
3D view of
elastic and
plastified
regions
Plan view of
elastic and
plastified
regions
In the present section, the effect of the stiffener spacing on the structural response is investigated by
making use of LBA, GNA, MNA and GMNA. The results were obtained in the context of the diploma
thesis of Nikoloudi (2015). The cases considered are the following:
No stiffeners
The spacing of 5m has already been presented but is also included herein for completeness.
The results obtained with LBA are presented in Table 5-5 with emphasis on the appropriate view of
buckling, the buckling load factor and the type of buckling.
3D View of
buckling
Type of Local buckling of Local buckling of Local buckling of Local buckling of Local buckling of
buckling shell at the top top stiffener intermediate shell at the shell at the
(Y-direction) stiffener bottom (Y- bottom (Y-
direction) direction)
It is observed that for the case without stiffeners buckling takes place at the top of the shell structure
towards Y-direction. If the number of stiffeners is relatively small, buckling takes place along a
circumferential stiffening ring. For a larger number of stiffeners, local buckling appears at the bottom
support towards Y-direction. The use of stiffening rings increases the buckling load factor.
The GNA results are presented for all cases in Figure 5-23. It can be seen that the stiffness is not
affected by the presence of stiffeners and by their density along the shell structure. On the other
hand, the capacity is affected.
Figure 5-23: Comparison between equilibrium paths obtained with GNA for various stiffener spacings
The MNA results are presented for all cases in Figure 5-24. It can be seen that initially all structures
behave in the same way independently of whether they have stiffeners or not. Plastification starts
earlier in the case that there are no stiffeners. The smaller the number of stiffeners used, the earlier
plastification appears in the cases of the stiffened structures. The capacities reached by the stiffened
structures with stiffeners at distances of 5m and 2.5m are almost the same.
Figure 5-24: Comparison between equilibrium paths obtained with MNA for various stiffener spacings
The GMNA results are presented for all cases in Figure 5-25. In the case that the nonlinearities are
combined, the larger the number of stiffeners used, the larger the ultimate capacity achieved. The
large number of stiffeners reduced the geometrical nonlinearity, and plastification developed at a
higher load level compared with structures with small number of stiffeners.
Figure 5-25: Comparison between equilibrium paths obtained with GMNA for various stiffener spacings
The deformed shapes of the structures investigated in the present section are depicted in Figure 5-26
for GMNA. It can be seen that the presence of many stiffening rings reduces the deformation of the
cross-section of the shell.
Figure 5-26: Comparison of 3-dimensional views of deformed shapes at failure from GMNA for various spacings of
stiffeners
5.8.5 Summary of results and comparison with the rigidly stiffened structure
In Table 5-6 the buckling and collapse load factors of section 5.8 are summarized and compared with
the case that the stiffening rings are rigid and remain elastic throughout the analysis. It is interesting
to note that the capacities in the case of rigid and elastic rings are smaller than the ones found for the
case of 2.5m spacing (and 5m spacing in LBA and GNA). This may be explained by the different
locations of failure observed in these cases. It also indicates that an optimum solution may be found
as far as the stiffness and spacing of the stiffeners are concerned.
Rigid and
elastic
Unstiffened 20m 10m 5m 2.5m
rings (5m
spacing)
LBA buckling
12.3 15.3 17.4 19.9 21.7 18.5
load factor
GNA collapse
6.7 9.6 14.5 18.7 18.5 15.9
load factor
MNA collapse
7.4 7.7 7.9 8.0 8.1 8.1
load factor
GMNA collapse
4.2 4.5 5.0 5.6 6.5 5.8
load factor
5.9 CONCLUSIONS
In Chapter 5, the structural response of a flexibly stiffened structure was investigated by making use
of advanced numerical procedures. The main flexibly stiffened structure that was used as a basis for
the analyses had circumferential stiffeners with angle L120X120X10 cross-section at a distance equal
to 5m between each other. The stiffened structure was compared with the unstiffened one and useful
conclusions were drawn. Parameters such as the distance between successive stiffeners, the initial
imperfections and the rigidity of the stiffeners were also examined.
The main conclusions that can be drawn from the investigation of the present chapter can be
summarized as:
The effect of the stiffening rings on the overall buckling capacity was important in the case of
wind pressure as they restricted to some extent possible changes of the cross-sectional shape,
smoothening stress peaks and local deformations, thus leading to increased capacity.
The use of nonlinear analyses verified the results obtained from LBA. The stiffening rings
increased the capacity of the structure in the cases that wind pressure was applied to the
shell surface, but did not have any effect for types of loading that did not modify significantly
the shape of the cross-section, regardless of whether elastic or elastoplastic materials were
used. For load cases including applied wind pressures, locations of failure appeared at many
locations in the structure and not necessarily at the location where beam theory predicted
failure.
The higher the material yield stress was, the largest the beneficial effect of the stiffening rings
on the capacity of the structure was.
The use of the stiffening rings did not modify the stiffness of the structures in all cases.
The use of very stiff and elastic rings totally restricted the cross-sectional shape’s changes and
failure appeared at the location predicted by beam theory, as expected. Nevertheless, the
capacity of the structure based on GMNA was not very different between the cases of very
stiff-elastic rings and L120X120X10 rings.
Using a more dense arrangement of stiffening rings increased the buckling and collapse load
factors. Nevertheless, it did not have any effect on the overall stiffness of the structures.
A comparison of the numerical results obtained for the case of 2.5m spacing with the ones
found for rigid and elastic rings showed that the latter did not necessarily lead to the largest
possible strengths. The buckling and collapse load factors for 2.5m spacing were larger than
the ones found for the rigidly stiffened structure.
6.1 INTRODUCTION
In all cases considered in the previous chapters, the shells had no openings. However, often a large
opening is required, in order to accommodate an inlet duct. This leads to reduced area and moment
of inertia of the chimney cross-section and has a significant impact on the strength and stiffness of
the structure. In many cases, the opening is so large that use of appropriate strengthening techniques
is necessary.
In the present chapter, the required inlet is accommodated by using either one single large opening or
an array of many smaller openings, as will be shown in the following sections. In both cases,
appropriate strengthening techniques are used and evaluated by making use of both linear and
nonlinear numerical procedures.
In this case, only a single large opening is used, having a height of 20m (between levels +5.00m and
+25.00m) and corresponding to an angle of 120o (one third of the circular cross-section). It is
assumed that the opening is located in the more compressed region along the wind direction +X,
which is expected to be one of the most unfavourable cases. A structure with such a large hole, if not
appropriately strengthened, has very small stiffness and strength to withstand service loads.
Therefore, stiffening and strengthening is absolutely necessary and only an appropriately stiffened
structure is considered in the present work. For simplification, the stiffened structure with one large
opening will be named “stiffened-LO”, while the shell structure that has no stiffeners and no opening
will be denoted as “unstiffened”, and will serve as basis for comparison.
The strengthening of the shell is achieved by making use of box-type cross-sections attached to the
shell, as shown in Figures 6-1 and 6-2. There are vertical stiffeners that are placed along the vertical
sides of the large opening and reach the base. Their cross-sectional area is selected sufficiently large
to replace the area of the original cross-section that is removed due to the opening. There are also
ring stiffeners that run along the lower and higher edge (+5.00m and +25.00m, respectively) of the
opening. Hence, stiffeners are placed along all four sides of the opening.
The areas of the vertical box-type cross-sections required for obtaining area and second-moment of
inertia of the cross-section equal to the ones of the cross-section of the unstiffened structure are
477cm2 and 1330cm2, respectively. The minimum required area for the vertical stiffeners specified by
CICIND recommendations is equal to 529cm2. Thus, the required cross-sections are quite large,
especially the one of 1330cm 2, which is required for achieving equivalent moment of inertia, and
would result in constructional difficulties, as well as problems in the force transfer between the shell
and the stiffeners. For this reason, box-type vertical stiffeners (RHS550x530x35) with cross-sectional
area equal to 570cm2 are selected, fulfilling the minimum requirement imposed by CICIND and being
larger than the one needed for replacing the area of the material that is removed due to the opening.
This area (570cm2) is much smaller, of course, than the one required for obtaining the second-
moment of inertia of the unstiffened structure. The ring stiffeners have a slightly smaller area than the
one used for the vertical ones (RHS550x500x35).
The applied loads were described in detail in Chapter 2 but are repeated here for completeness. The
vertical loads that are considered in the following analyses include the self-weight of the shell
structure and of the stiffeners, as well as the weights of the three platforms applied at three different
levels, of the damper and of the silencer. These vertical loads were also included in the analyses of
the structures without openings. In addition to them, the weight of the inlet duct is considered and
simulated through a total weight of 35.00 tons applied at the vertical edges of the breeching.
The horizontal loads are exactly the same as the ones used for the structures without openings and
the only obvious difference is that no wind pressures are applied in the area of the opening. For this
reason, the total shear force due to the wind loads is smaller in the case of stiffened-LO structure
when compared to the structures without openings.
A comparison between the results obtained with LBA and nonlinear analyses for the unstiffened
structure and the stiffened-LO structure is shown in Figure 6-3. From the LBA results it is concluded
that, despite the large hole, the use of the vertical and ring stiffeners leads to an increase of the
buckling strength, as compared with the unstiffened structure. The stiffeners restrict changes of the
cross-sectional shape due to the wind pressures that are applied directly to the thin-walled shells, and
this restriction results in strength increase.
The stiffness of the stiffened-LO structure is smaller than the one corresponding to the unstiffened
structure for all types of nonlinear analyses. This was anticipated because the effective moment of
inertia of the stiffened-LO structure over the height of the opening is much smaller than the one of
the unstiffened structure. The use of the stiffeners leads to a significant increase in the capacity in the
case of GNA, but in the case of elastoplastic material (MNA) the stiffened-LO structure has a smaller
strength in comparison to the unstiffened structure. In the most realistic case of GMNA the capacity
obtained for the stiffened-LO structure is similar to the one found for the unstiffened structure.
Figure 6-3: Comparison between equilibrium paths obtained with LBA and nonlinear analyses
Deformed shapes of the stiffened-LO structure at failure are presented in Figure 6-4. It can be seen
that the cross-sectional shape is slightly affected by the wind pressure only along the higher part of
the shell structure, where no ring stiffeners are provided. At the lower part, the cross-sectional shape
is almost undisturbed. The distribution of plastification in the structure is depicted in Figure 6-5. It can
be observed that plastification is accumulated close to the intersections between the lower ring and
the vertical stiffeners. The locations of plastification are different from the ones found in the case of
the unstiffened structure.
Figure 6-4: Deformed shape of the stiffened-LO structure at failure based on GMNA
Figure 6-5: Elastic and plastic regions of the stiffened-LO structure at failure based on GMNA
In conclusion, in order to recover the lost strength due to the large hole, very large areas are required
for the stiffening structural elements. Nevertheless, the resulting stiffness will be significantly smaller.
Using stiffeners that lead to a second-moment of inertia close to the one of the unstiffened structure
is prohibitive from a constructional point of view, which means that in most cases it will not be used in
practice.
Instead of providing one large opening, as presented in section 6.2, an array of many circular holes
can be used for achieving the same purpose. In that, the total area of the circular holes should be
larger than the minimum required for the satisfactory gas transfer. As in the case of the single large
opening, the area in which the circular holes are positioned has a height equal to 20m (between levels
+5.00m and +25.00m) and corresponds to an angle of 120o (one third of the circular cross-section).
Additionally, as in section 6.2, it is assumed that the area of the circular openings is located in the
more compressed region along the wind direction +X, which is expected to be one of the most
unfavourable cases. The openings are configured in 16 rows, each having 6 holes with diameter equal
to 80cm. The distance between two successive rows of holes is 125cm and a view of the geometry is
shown in Figure 6-6. This structure is denoted as “unstiffened-CH”. The loads used in the following
analyses are the same as the ones presented in subsection 6.2.1. No wind pressures are applied
between the circular holes, as the inlet duct protects this area from exposure to wind.
A detail of the meshing in the region of the holes is shown in Figure 6-7. It can be seen that a very
dense mesh of shell elements (equal to the minimum mesh size of 7cm used for the numerical
analyses of the unstiffened and stiffened structures without openings) is used in the regions between
the holes, in order to obtain a more accurate response of this highly stressed area.
Figure 6-8: Comparison between equilibrium paths obtained with LBA and nonlinear analyses
The stiffness of the unstiffened-CH structure is smaller than the one corresponding to the unstiffened
structure for all types of nonlinear analyses. The capacities of the unstiffened-CH structure obtained
with MNA, GNA and GMNA are also smaller than the corresponding capacities of the unstiffened
structure. Although the loads are not the same in the two cases (in the case of the holes the inlet
duct’s weight is added and the wind pressures are removed from the region of the holes), it is
observed that the holes reduce significantly the strength. It is interesting to observe that the response
up to buckling is almost linear and the capacity found with GNA is almost the same as the one
obtained with the use of LBA. Appropriate strengthening is required for increasing the capacity and
stiffness of the unstiffened-CH structure, which will be thoroughly investigated in the following
sections.
The deformed shape of the unstiffened-CH structure at failure according to GMNA is presented in
Figure 6-10 and can shed light on the behavior of the critical region of the holes. The structure’s
deformed shape consists of the cantilever’s global deformation, of a significant change of the cross-
sectional shape due to the wind pressure, and of local deformations in the region of the holes. The
first two types of deformation were observed in the case of the unstiffened structure, too. The third
type is due to the decreased cross-section around the holes, which forces the flow of stresses to
follow paths between the holes, leading to stress concentration and associated larger deformations in
these regions.
The distribution of plastification and the regions that remain elastic are depicted in Figure 6-11. It can
be seen that the material is plastified in the region of the holes, where LBA indicated that buckling
takes place. Contrary to the case of the unstiffened structure, the presence of holes weakens this
region and failures (buckling and plastification) are concentrated there.
Figure 6-10: Deformed shapes of the unstiffened-CH structure at failure based on GMNA
Figure 6-11: Elastic and plastic regions of the unstiffened-CH structure at failure based on GMNA
One way to increase the stiffness and capacity of the shell structure having an array of many rows of
circular holes is by using a larger thickness for the shell in the region of the holes. In this way the
material that is removed due to the holes is replaced in the weakened region and is expected to lead
to satisfactory recovery of the area and of the second-moment of inertia, because its distance from
the center of gravity is similar to the one of the removed material.
The structures that will be investigated in this section are presented in Figure 6-12. The unstiffened
and unstiffened-CH structures have already been presented in previous sections. The stiffened-CH
structure has larger thickness in the region of the holes than the one that the rest of the shell has
(1.3cm). The stiffened-CH-Rings structure with small vertical stiffeners (115x35) and the stiffened-CH-
Rings with large vertical stiffeners (1000x35) are the other two cases examined. In both cases, the
stiffening rings have a cross-section L120x120x10. In these two cases the thickness in the region of
the holes is larger, and there are stiffening rings and vertical stiffeners.
Figure 6-12: Views of the unstiffened structure and the appropriately strengthened structures with circular holes
It should be noted that apart from the geometrical characteristics, the loads applied to each one of
the structures are different. For instance, the unstiffened structure is loaded by wind loads applied to
the whole area of the chimney. On the other hand, the models with holes are not loaded by the wind
in the region of the holes but are loaded by the additional weights of the inlet duct. Nevertheless, a
quantitative and qualitative comparison between these 4 structures can be performed and lead to
useful conclusions, as the unstiffened structure is actually the structure according to which a
practicing engineer would calculate the ultimate strength and stiffness of the structure. Moreover, the
unstiffened structure without openings and stiffening is considered as the comparison basis because
CICIND provisions refer to such an idealized model.
During the first stage, LBA are performed in order to find the minimum thickness of the holes’ region
for which the buckling load factor is close to the one obtained for the unstiffened structure without
any holes (comparison basis). It is reminded that the buckling load factor of the unstiffened structure
is close to 12.3. The results are presented in Figure 6-13 in terms of the buckling load for various
values of the shell thickness (including loss due to corrosion) in the region of the openings.
Figure 6-13: Variation of the ratio of the buckling load (b.l.) of a stiffened structure with respect to the
unstiffened structure for various thicknesses of the holes’ region (LBA)
In the case of the stiffened-CH structure, a buckling load factor approximately equal to 12.1 and a
buckling mode shape similar to the one of the unstiffened chimney are found for a minimum thickness
close to 24mm. It is therefore concluded that for this thickness buckling does not take place in the
region of the openings but in the same region as in the unstiffened chimney.
In the case of the stiffened-CH-Rings-Small structure, the strengthening of the holes’ region requires a
thickness of a similar magnitude to the one found for the stiffened-CH structure, in order to obtain a
value of buckling load close to the one found for the unstiffened structure. This means that practically
the use of the stiffening rings does not contribute directly to an increase in the buckling strength of
the chimney and the buckling load factor of the chimney is not affected for values of thickness in the
critical region of the holes smaller than approximately 24mm. For larger thickness values, the use of
the stiffening rings leads to an increase in the buckling strength when compared with the case that no
rings are used.
It is also interesting to note that for a thickness equal to 24mm, in the case of the stiffened-CH-Rings-
Small structure buckling takes place in the region of the openings. This is attributed to the fact that
the stiffening rings offer overall stiffness to the shell, but they do not have an important direct effect
on the stiffening of the holes’ region. As a result, a larger value of thickness is required for the
stiffened-CH-Rings-Small structure than the one for the stiffened-CH structure, in order to avoid local
buckling in the region of the holes. This result is surprising, but it highlights the fact that using only
rings is not a sufficient stiffening method in the case of region with many holes. The thickness in the
region of the holes should be equal to 27mm in order to avoid buckling in the holes’ region. The
corresponding buckling strength is then equal to a value close to the buckling load factor of the
stiffened structure without openings.
In the case of the stiffened-CH-Rings-Large structure, the use of the large vertical stiffeners leads to a
better performance and a thickness approximately equal to 20mm is required for the region of the
openings in order to achieve a buckling strength equal to the unstiffened structure’s buckling factor.
For small values of thickness in the critical region (including the value of 20mm) local buckling is
observed in the holes’ region. For larger values of thickness (>24mm), buckling of the large thin
vertical stiffeners is critical and limits the buckling load factor that is smaller than the stiffened-CH-
Rings-Small structure’s.
In the following parametric study the thickness in the region of the holes is varied for all structures
with holes and the calculated collapse loads and elastic stiffnesses obtained by making use of GMNA
are compared with the corresponding collapse load and elastic stiffness found with GMNA for the
unstiffened structure without stiffening and openings. The results are summarized in Figure 6-14.
Figure 6-14: Variation of the ratio of the collapse load (c.l.) and elastic stiffness (e.s.) of a stiffened structure with
respect to the unstiffened structure for various thicknesses of the holes’ region (GMNA)
It can be seen that in the case of the stiffened-CH structure, a thickness approximately equal to
24mm is required in order to achieve a collapse load and an elastic stiffness close to the ones of the
unstiffened structure. A further increase of the thickness increases the elastic stiffness but does not
have a significant effect on the collapse load.
If the strengthening technique includes the use of larger thickness in the region of the holes combined
with stiffening rings and small vertical stiffeners, the results are modified. The stiffness increases with
an increase of the thickness of the holes’ region. But the most interesting conclusion is that for
thickness of the holes’ region smaller than 21mm, the effect of the stiffening thickness is similar to the
one observed in the case of the stiffened-CH structures. This practically means that if the region of
the holes is not appropriately strengthened with the use of larger thickness, the effect of the ring
stiffeners on the capacity of the chimney is not significant. Of course, the use of stiffening modifies
the stress field and for this reason the thickness required for reaching the collapse load and elastic
stiffness of the unstiffened structure is 21mm, a smaller value than the 24mm required in the case of
the stiffened-CH structure. It is also important to note that as the thickness increases, the collapse
loads of the stiffened-CH-Rings-Small structure reach values close to the one corresponding to the
structure without openings that is stiffened with rings.
The last case is exactly the same as the stiffened-CH-Rings-Small structure with the exception that
instead of small vertical stiffeners, large vertical stiffeners are used. Additionally, the stiffening rings of
the upper and bottom side of the holes’ region are considered as rigid because the construction is
expected to include a horizontal truss there, which is almost undeformable in its plane. Based on the
results, it can be concluded that these two additions lead to a significant enhancement of strength
and stiffness, which reach larger values than the ones found for the other strengthening methods.
In the last part of this research, the effect of imperfections is also considered in the analyses, in order
to obtain a more realistic structural response. The most cost-effective alternative from each type of
structure is considered:
Unstiffened structure (no openings and no stiffeners)
Stiffened-CH-Rings-Small structure with 24mm thickness in the region of the openings. The
value of 24mm corresponds to the minimum value that leads to satisfactory recovery of the
lost capacity, according to LBA, when compared with the unstiffened structure.
Stiffened-CH-Rings-Large structure with 13mm thickness in the region of the openings. The
value of 13mm proved to be quite sufficient for obtaining a capacity larger than the one found
for the unstiffened structure.
The results are presented by making use of equilibrium paths and by including the effect of initial
imperfections with magnitude equal to 0.65t=0.845cm (t is the thickness of the shell structure) and
are based only on the 1st buckling mode shape. The corresponding equilibrium paths are presented in
Figure 6-15.
It can be concluded that the capacity predicted for the unstiffened structure is very close to the result
obtained with CICIND. Of course, this conclusion is valid only for the specific case investigated, with
the pattern of imperfections considered, and cannot be generalised (see Chapter 4). It has already
been shown in the context of the present research work that the capacity found by CICIND is based
on beam theory assumptions that are far from reality, as buckling and plastification take place at
different locations from the ones predicted by it. Therefore, it is anticipated that, in many cases, the
lack of theoretical validity of the assumptions behind the CICIND predictions may lead to inaccuracies.
In the case of a single large opening that is appropriately strengthened (Stiffened-LO) in order to
obtain the lost capacity, the structural response is satisfactory in terms of the strength but not of the
stiffness, which is significantly smaller than the one found for the unstiffened structure. This is the
case because, as mentioned before, the material required in order to obtain the unstiffened structure’s
stiffness is very large and cannot be easily used in engineering practice.
The presence of many openings, combined with initial imperfections in this region, leads to significant
decrease of the stiffness and ultimate strength (Unstiffened-CH). In the weakened region, due to
stress concentration, buckling and plastification become evident. If the strength of CICIND is used as
a design a basis, a significantly unsafe prediction is made. Appropriate strengthening is required.
If the area of the openings is strengthened with the use of 24mm thickness (Stiffened-CH-24mm), a
very satisfactory recovery of stiffness and strength is observed. The obtained capacity is slightly larger
than the one predicted by CICIND and the structural response is similar to the one of the unstiffened
structure.
If apart from using a thickness equal to 24mm in the region of the openings, ring stiffeners and small
vertical stiffeners are also used (Stiffened-CH-Rings-Small-24mm) the structural response is not
significantly improved when compared with the stiffened-CH-24mm structure. One of the main
reasons is that the imperfection induced is based on the 1st buckling mode that is associated to
buckling in the region of the openings (it is reminded that for the stiffened-CH-Rings-Small structure a
thickness of 27mm is required in order to avoid buckling in the openings’ region). The use of stiffening
rings is meaningful, after avoidance of failure in the openings’ region is achieved.
Figure 6-15: Equilibrium paths obtained with GMNIA for various types of structures and comparison with the
result obtained with the provisions of CICIND
6.6 CONCLUSIONS
Several interesting conclusions can be drawn from the numerical investigations carried out for the
cases that a breeching is present in the chimney shell. The main observations can be summarized as:
The use of a large single opening for the accommodation of the inlet duct requires
strengthening with large cross-sections for replacing the removed material and achieving
similar areas and second-moment of inertia to the ones of the structure without an opening.
Especially, the requirement for similar second-moment of inertia is very difficult to fulfill.
An alternative way to construct openings is by making use of an array of many small circular
openings. This leads to smaller weakening of the cross-section and to smaller stress
concentrations. In such a case, potential failures are concentrated in the region of the holes.
The capacity and stiffness of the structure are reduced due to the presence of the holes.
There are many possible ways for strengthening the weakened structure. One of them that
appears to be effective and relatively easy to construct, is by increasing the thickness of the
holes’ region. For an appropriate value of thickness, failure is removed from the weakened
region of the holes and the structure behaves as if no holes exist.
The prediction of CICIND is based on beam theory assumptions that are not close to reality in
the case of large diameter chimneys subjected to wind pressures to their thin shell surfaces.
Nevertheless, the ultimate strength predicted by CICIND for the unstiffened structure is very
close to the result obtained with GMNIA. This conclusion, of course, cannot be generalized for
all cases of chimneys.
7.1 INTRODUCTION
The conclusions that were drawn from this research project have been thoroughly described in the
previous chapters. In the first part of this chapter, the main conclusions will be briefly repeated, with
emphasis on those that are usefull for practicing engineers who are active in the analysis and design
of steel chimneys. Afterwards, suggestions for required further research on the structural response of
large diameter chimneys in the future, are formulated.
The main conclusions that were presented in the previous chapters, are summarized below.
used. In the realistic case of applying the wind loading as a normal pressure to the shell surface,
beam theory cannot be applied, as the pertinent failure mechanisms are associated with 3D shell-type
response.
The use of nonlinear analyses verified the results obtained from LBA. The stiffening rings increased
the capacity of the structure in the cases that wind pressure was applied to the shell surface, but did
not have any effect for types of loading that did not modify significantly the shape of the cross-
section, regardless of whether elastic or elastoplastic materials was assumed. For load cases including
applied wind pressures, failure appeared at many locations and not necessarily according to beam
theory predictions. As far as the material was concerned, the higher the material yield stress was, the
largest the beneficial effect of the stiffening rings on the capacity of the structure was.
The use of very stiff and elastic rings totally restricted the cross-sectional shape’s changes and failure
appeared at the location predicted by beam theory, as expected. Nevertheless, the capacity of the
structure based on GMNA was not very different between the case of very stiff-elastic rings and
L120X120X10 rings. It was quite surprising though, that when compared with the case of rigid and
elastic rings, the case of flexible and densely arranged rings (2.5m), led to larger capacities.
On the other hand, in the case of direct application of wind pressure to the perimeter of the thin-
walled shell structure, the cross-sectional shape was significantly modified. For this reason, beam
theory was not sufficient, leading to unsafe results. The locations, at which buckling occurred, were
not the ones predicted by beam theory, appearing instead across the wind direction. Nevertheless, the
collapse load factor predicted by CICIND was relatively close to the numerical one obtained with
GMNA.
The research carried out in the context of the present work focused on a specific actual chimney, and
useful conclusions were drawn, as far as the effects of mesh size, stiffeners, breeching, initial
imperfections and other parameters are concerned. The investigation was partly based on analytical
evaluations and mainly on advanced numerical analyses. For validation of the numerical findings,
future research should include also experimental, as well as additional numerical and analytical work,
thus covering the entire range of possible tools that a structural engineer-researcher has at his
disposal nowadays.
The available experimental work related to the topic investigated in the present project is very limited.
An experimental programme should, therefore, take place, in order to validate numerical results,
including wind tunnel testing. Especially the case of breeching is of great interest and according to the
authors’ knowledge, there are no experimental tests showing the wind pressure distribution in such
cases.
Additionally, appropriate tests of scaled chimney specimens loaded to collapse should take place,
using various diameters, heights and thicknesses for the shell structures. The tests will lead to useful
conclusions as far as the structural response and the type of failure of such structures are concerned.
They can also be a basis for calibrating the numerical models that will be used in the next step.
After having been calibrated by comparison to experimental results, the numerical models can be
extensively used for performing parametric studies and investigating a wider range of cases. A large
number of numerical analyses can prove very useful for examining the way that such structures
behave and draw useful qualitative and quantitative conclusions. On the basis of such numerical
results, it will also be possible to propose criteria of validity of the beam theory for steel chimneys,
and identify cases in which shell modeling will be required, which is an issue of great practical
significance.
8 REFERENCES
1. ADINA R&D (www.adina.com), Theory and Modeling Guide - Volume I: ADINA Solids & Structures,
Report ARD 05-7, (2005).
2. Angelides, M., "Large Diameter Steel Chimneys", CICIND Report, Vol. 29, No. 2, pp. 113-116, 2013.
3. Angelides, M., "Shell Buckling Issues in Chimneys and Cooling Towers", CICIND Report, Vol. 30, No.
1, pp. 47-51, 2014.
6. Bailey, R. and Kulak, G.L., “Flexural and shear behaviour of large diameter steel tubes”, Report No.
119, Department of Civil Engineering, University of Alberta, Edmonton, 1984.
8. Dimopoulos, C.A. and Gantes, C.J., “Experimental Investigation of Buckling of Wind Turbine Tower
Cylindrical Shells with Opening and Stiffening under Bending”, Thin-Walled Structures, Vol. 54, pp.
140-155, 2012.
9. Dimopoulos, C.A. and Gantes, C.J., "Comparison of Stiffening Types of the Cutout in Tubular Wind
Turbine Towers", Journal of Constructional Steel Research, Vol. 83, pp. 62-74, 2013.
11. EN 1993-1-6, “Design of Steel Structures – Strength and Stability of Shell Structures”, 2004.
13. Flϋgge, W., 1932. Die Stabilitӓt der Kreisyzlinderchalen. Ingenieur-Arch. 3, 463–506.
14. Gantes, C.J. and Fragkopoulos, Κ.Α., “Strategy for Numerical Verification of Steel Structures at the
Ultimate Limit State”, Structure and Infrastructure Engineering, Vol. 6, Issue 1-2, pp. 225 - 255, 2010.
15. Greiner R. & Derler, P., "Effect of Imperfections on Wind-Loaded Cylindrical Shells", Thin-Walled
Structures, Vol. 23(1-4), pp. 271-281, 1995.
16. Nikoloudi, I., “Design of Steel Industrial Chimneys Against Buckling”, Diploma Thesis, School of
Civil Engineering, National Technical University of Athens, 2015.
17. Reddy, B.D., Calladine, C., “Classical Buckling of a Thin-walled Tube Subjected to Bending Moment
and Internal Pressure”, Int. J. Mech. Sci., Vol. 20, pp. 641–650, 1978.
18. Rotter, J.M., Sadowski, A.J., Chen, L., "Nonlinear Stability of thin Elastic Cylinders of Different
Length under Global Bending", International Journal of Solids and Structures, Vol. 51, pp. 2826-2839,
2014.
19. Seide, P., Weingarten, V.I., “On the Buckling of Circular Cylindrical Shells Under Pure Bending”,
ASME J. Appl. Mech., Vol. 28, No. 1, pp. 112–116, 1961.
20. Teng, J.G. & Rotter, J.M., editors, "Buckling of Thin Metal Shells", Taylor & Francis, pp. 286-333,
2004.
21. Timoshenko, S., “Theory of Elastic Stability”, McGraw-Hill Book Company Inc., New York, 1932.