International Journal of Pressure Vessels and Piping 208 (2024) 105135

Contents lists available at ScienceDirect

International Journal of Pressure Vessels and Piping

journal homepage: www.elsevier.com/locate/ijpvp

Buckling behaviour of steel dome cap design under external pressure

Mohd Shahrom Ismail a, *, Syed Mahathir Muhammad al-Attas b, Jamaluddin Mahmud c
a
Jabatan Kejuruteraan Mekanikal, Politeknik Sultan Salahuddin Abdul Aziz Shah, Shah Alam, 40150, Selangor, Malaysia
b
Quadrant 2 Technologies Sdn. Bhd., Kuala Lumpur, Malaysia
c
School of Mechanical Engineering, College of Engineering, Universiti Teknologi MARA, 40450, Shah Alam, Selangor, Malaysia

A R T I C L E I N F O A B S T R A C T

Keywords: This study aims to numerically investigate the buckling behaviour of externally pressurised dome caps. Following
Buckling the recommendations of the European Standard EN 1993-1-6, there was good agreement between the Finite
Dome cap Element (FE) model and the experimental results, with an average difference of less than 7 %. Several imper­
Geometric imperfection
fection approaches, such as (i) single load indentation (SLI) and (ii) axisymmetric outward bulge (AOB), and (iii)
Steel
PD 5500
eigenmode-affine (EM) were used to analyse the dome cap. The EM proved to be the worst imperfection for the
EN 1993-1-6 externally pressurised dome cap, yet the SLI is an attractive technique for simulating realistic imperfections. The
imperfection tolerance was designed according to EN 1993-1-6, and the SLI was used for the realistic-case
imperfection. From parametric study, a plot of the upper and lower bound curves serves as a design recom­
mendation, similar to the NASA SP-8032 guideline. Through a regression analysis, a first-cut estimation equation
of the imperfect dome cap buckling load with reliability of R2 = 0.97. A case study of shell buckling design on
pressurization damage to an oil storage tank’s spherical shell partition and its stress assessment are evaluated.
This study provides a practical solution for the evaluation of buckling of steel domes under external pressure.

1. Introduction The geometrical imperfection is defined in terms of its position,

dent-amplitude, shape, and size. The location of the imperfection has a
Externally pressurised domed caps are commonly used in many en­ strong influence on the load-carrying capacity of the structures [12–17].
gineering applications. Following their high load-carrying capacity, they The load-carrying capacity of the shell structure is referred to as the
are used as partitions or closed-end in pressure vessels, or as hatches to knockdown factor (KDF) (i.e., Pimp/Pperf). The buckling of domes with
cover the access ports of variously shaped pressure vessels in subsea various geometric configurations and initial geometric imperfections
applications, aerospace (nose-domed), and civilian applications [1,2]. has been extensively studied under uniform external pressure. For
To sustain the structural integrity, the material properties, pre-buckling example, the initial geometric imperfections in the forms of linear
deformations, and eccentric/concentric loading conditions play a vital eigenmode mode [18–21], local inward dimple [22,23],
role in internal pressurised domed caps [3]. In vacuum condition, the increased-radius imperfection and local flattening [24,25], measured
strength and stability of the domed caps depends strongly on their geometric imperfections (MGI) [19,20,26], localized reduced stiffness
design and geometric imperfections [1,4,5]. The domed caps are clas­ method (LRSM) [27,28], and variations of measured shell thickness [19,
sified as a complete, deep, and shallow shell structure [6]. Several 29]. The uncertainty analysis of dome caps with a variation of industrial
contemporary design standards such as the European Convention for design codes has been reported in Refs. [30,31]. Some of the formulae
Constructional Steelwork (ECCS) [7], Det Norske Veritas (DnV) [8], that are most widely used to estimate the buckling load of spherical caps
British Standard (PD 5500) [9], and American Bureau of Shipping (ABS) are Zoelly’s analytical linear buckling formula [32], NASA SP-8032 KDF
[10] are used to design the domes. [33], Energy Barrier Criterion (EBC) KDF [34], Wagner KDF [6] and a
The fabricated domed cap exhibits almost more or less pronounced closed-form lower-bound of KDF by Ismail et al. [18].
imperfections of geometry or material [11]. The geometrical imperfec­ The literature shows that there is still limited knowledge of the
tion strongly affects the load-carrying capacity of the shell due to the imperfection sensitivity of cone-cylinder transition in the open litera­
deviations in geometry, material behaviour and boundary conditions. ture. The objective of this paper is twofold. First, to numerically

* Corresponding author.
E-mail address: mohdshahrom@psa.edu.my (M.S. Ismail).

https://doi.org/10.1016/j.ijpvp.2024.105135
Received 2 December 2023; Received in revised form 4 January 2024; Accepted 10 January 2024
Available online 11 January 2024
0308-0161/© 2024 Elsevier Ltd. All rights reserved.
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

benchmark the externally pressurised dome caps following the recom­

mendations of the European Standard EN 1993-1-6. To analyse the
imperfection sensitivity of the dome, further investigation has been
conducted using different imperfection approaches - (i) single load
indentation (SLI) and (ii) axisymmetric outward bulge (AOB), and (iii)
eigenmode-affine (EM). Parametric study was also performed on
different dome cap configurations and compared with several experi­
mental results. Secondly, a case study of shell buckling design on pres­
surization damage to an oil storage tank’s spherical shell partition and
its stress assessment are analysed and discussed thoroughly. The present
work is entirely numerical using ABAQUS finite element (FE) code and it
complements the experimental results reported in Refs. [20,35].
This study is unique in that there have been fewer or no studies have
been conducted on this topic. In general, the results of this study are
highly useful and contribute to the body of knowledge on the effects of
structural imperfections in the preliminary stages of the design, fabri­
cation, and analysis of externally pressurised spherical shell structures.

2. Benchmark analysis

2.1. Design of dome cap according to PD 5500

The numerical results of buckling capacity of the pressurised dome

caps are validated with PD 5500 design code [9]. The available design
rules use the conventional working stress to determine each failure
mode. This approach is crucial for verification purposes at the pre­
liminary design stage. Together with the working stress and safety fac­
tors, the design rule also considers several uncertainties, namely
boundary conditions and loads, material hardening and structural im­
perfections of the tested spherical shell structures. Under external
pressure, the dome critical buckling, Pcr according to PD 5500 design Fig. 1. (a) Load and boundary condition of externally pressurised dome cap
code [9] is expressed by equation (1). and its (b) schematic diagram.
2E ( t )2
Pcr = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ (1) 2.3. Results of benchmark analysis
3(1 − ν ) R 2

Next, the plastic buckling pressure, Pyield, is given in equation (2) Fig. 3 (a) - (b) shows the results of the benchmark analysis with
(t) eleven (11) experimental data. The results show a case of hemispherical
PYield = 2σYield (2) (see Fig. 3 (a)) and shallow/flat (see Fig. 3 (b)) dome Caps. The ratio of
R
the buckling load calculated with PD 5500 and FE (i.e., the nonlinear
Finally, the collapse pressure of PD 5500 is estimated using equation (3) analysis GMNA) to the experimental results shows the accuracy of the
( )2 ( )2 ( )2 analysis. The nonlinear Riks analysis provides good agreement with the
1 1 1 experimental results compared to PD 5500 design code. The average
= + (3)
PPD 5500 0.3Pcr Pyield percent difference between the experimental results and the numerical
analysis is less than 7 % for all cases. Obviously, a more accurate esti­
mate of the buckling load was found by model D3 for the flat shell case,
2.2. Finite element model
with less than 1 % difference. The model D2 from FE overestimates the
buckling load by 10 % compared to the experimental results.
Eleven (11) laboratory scale dome caps under external pressure are
For the case of a hemispherical dome, models 1# - 5# significantly
benchmark with FE model (see Fig. 1 (a)) [20,35]. The FE model follows
overestimated the buckling load by 11 % on average. The result clearly
the recommendation by European Standard EN 1993-1-6 [36]. The
indicates that the numerical model is appropriate for the analysis.
dome caps are described in the following terms: spherical radius, R,
Nonetheless, an insignificant buckling load is recorded for the case of
spherical thickness, t, base radius, r, spherical height, H, and semi-vertex
identical shells, as shown by the 1# - 5# models. This indicates a
angle, φ, as shown in Fig. 1 (b). Table 1 outlined the geometry and
noticeable presence of structural imperfection in the experimental
material properties of benchmark models, which are based on the
result. The structural imperfection is suspected to be in the form of
experimental data [20,35]. The FE model assumes a fully clamped
dents, uneven thickness, strain hardening, load and boundary condition
boundary condition with an elastic-perfectly-plastic material. Fig. 2
eccentrics and many more. The severity of the imperfection is marked by
shows an example of a steel elastic-perfectly-plastic material model
the deviating ratio of the experimental buckling load to the calculated
taken from Ref. [35]. The nonlinear static Riks analysis is used in the
numerical analysis, PExp/PColl, which is less than or greater than 1.0.
benchmarking analysis. The nonlinear static Riks analysis was set to be
As been reported in Refs. [18,20,35], the buckling load calculated by
Nlgeom = on, maximum number of increments = 100, initial arc length
PD 5500 slightly underestimates the experimental results for the cor­
increment = 0.01, minimum arc length increment = 1E-15 and
responding cases of models D1 - D6 (with an average of 46 %) and 1# -
maximum arc length increment = 1E36. The S4R shell element is
5# (with an average of 10 %). The corresponding result shows that the
employed in the numerical model. The S4R shell element is described as
buckling load of the 5# model is closer to the experimental data with
a four-node shell element with six degrees of freedom. In this study, the
differences of 6 % (i.e., PPD 5500/PExp = 0.943) calculated. The largest
numerical model is assumed to be a perfect shell by using the geometric
difference was found for the D1 model, where the buckling load
and material nonlinear analysis (GMNA) approach.

2
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

Table 1
Geometry and material properties of benchmark models in referring to the experimental data.
ID R [mm] t [mm] R/t E [GPa] σy [MPa] v Remark Refs.

D1 1816.5 1 1816.5 207 303.5 0.28 Shallow dome [32]

D2 1605.7 1.02 1574.2
D3 878.6 1.03 853.01
D4 1166.7 1.76 662.9
D5 759.3 1.76 431.42
D6 563.4 1.76 320.11
1# 58.84 0.432 136.2 193 205 0.28 Hemispherical dome [16]
2# 58.77 0.422 139.27
3# 58.99 0.423 139.46
4# 58.77 0.406 144.75
5# 58.75 0.415 141.57

Fig. 2. An example of an elastic-perfectly-plastic material model taken from

material properties given in [35].

significantly deviates by 50 %. From the given results, a reasonable

agreement of the buckling load between the experiment and the PD
5500 design code was found for the case of a hemispherical shell [20].

3. Design of dome cap according to EN 1993-1-6

Generally, it is compulsory for the FE model to follow the recom­

mendation of European Standard EN 1993-1-6 [36] for verification
purposes. In this section, the FE model is designed according to the
recommendation stages of LBA, GMNA, and GMNIA.

3.1. Linear buckling analysis (LBA) and mesh sensitivity study

In this section, the based parameters used for the LBA model are set Fig. 3. Summary of benchmark results for (a) hemispherical and (b) shallow
to be dome radius, R = 500 mm, uniform wall thickness, t = 1 mm, dome caps.
spherical height, H = 6 mm, and base radius, r = 83 mm. The dome cap
is assumed to be made of a stainless steel with an elastic modulus = 207
GPa, a yield stress, σyield = 303.5 MPa, and a Poisson’s ratio, υ = 0.28. Table 2
The ABAQUS subspace solver was used to calculate the linear buckling Four types of boundary conditions applied at the equatorial (bottom) plane of
the dome cap, e.g. type 1 means fully clamped support. Note c = variable is set
load. The critical buckling load, Pcr of the LBA, is calculated by multi­
zero, f = variable is set free.
plying a subsequently applied pressure load of 1 MPa to the spherical
shell to the obtained eigenvalue. Type of BC U1 U2 U3 UR1 UR2 UR3

The LBA plots the external pressurised steel spherical shells against 1 c c c c c c
dimensionless ratios: spherical radius-to-wall thickness and (R/t)-ratio. 2 c c c f c c
3 c c f c c c
From the analysis, the wall thickness, t, is varied according to the range
4 c c f f c c
of dimensionless ratios from 100 <R/t < 1000. A fully constrained
boundary condition was considered in the analysis to mimic the appli­
cation of a dome cap as part of an industrial component along with three variations of boundary conditions are necessary to simulate any
other boundary conditions, as indicated in Table 2. The other three

3
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

insignificant joining of the dome and to determine the most appropriate and PPost-collapse. For the case of D6 model, it can be seen that the dome
condition for the FE analysis. The FE model uses a coordinate system cap reaches its first yield pressure, PYield, at 0.949 MPa, followed by the
type spherical (see Fig. 1 (a)). As shown in Table 2, the terms used are collapse pressure, PColl, at 1.202 MPa, and the post-collapse pressure,
the tangential direction, U1, radial direction, U2, normal direction, U3, PPost-Collapse, at 0.504 MPa (see Fig. 5). Fig. 6 shows the load-carrying
tangential rotational direction, UR1, normal rotational direction, UR2 capacity of model 1#. The shell was found to reach its yield point at
and tangential rotational direction, UR3. This study focuses on the effect PYield = 3.892 MPa before collapsing at PColl = 8.19 MPa and post-failure
of boundary conditions on (i) critical buckling load, (ii) number of at PPost-Collapse = 4.5777 MPa. In both models, there was a considerable
buckling waves (n), and, (iii) structural mode failure. drop in pressure, followed by a stable equilibrium curve, as can be seen
Table 3 presents a convergence analysis with the variance of the in Fig. 5 for model D6.
mesh density to analyse the sensitivity of an appropriate FE model. The numerical result indicates that D6 model suffered asymmetric
Generally, the finer the mesh, the more time the model requires to failure (see inserted (a)–(c) in Fig. 5). The shell experienced a pro­
complete the analysis. From the convergence analysis, 6897 elements nounced yield stress in the equatorial region before propagating to the
were found to be sufficient for the FE model. This is reflected in the fact apex region. There was a strong stress concentration in the apex region
that the percentage deviation from the convergence analysis was at the peak of the buckling load, before the shell failed completely in the
reduced to almost 0.04 %–0.26 %, which is due to the density of the FE post-collapse mode. Initially model 1# appeared to maintain a minimal
model. However, when the number of elements exceeded 10,000, an concentration of yield stress around the equatorial edge (see Fig. 6 (a)).
insignificant buckling load was produced. The results also showed a At the peak of the buckling load, the stress spreads out and concentrates
reliable mesh number with a reasonable time for completion. locally below the apex near the equator before entering the post-collapse
The LBA also includes the analytical calculation of critical buckling phase (Fig. 6 (b)–(c)) and before bulging out in an axisymmetric manner.
pressure. The critical buckling pressure, PZoelly, is given in equation (4) In general, dome failure is associated with the following elastic or
inelastic modes: (i) asymmetric bifurcation, (ii) axisymmetric collapse/
2
PZoelly = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ EH 2
(4) snap-through, or (iii) asymmetric collapse. With respect to the results
(3(1 − ν2 )) R2 reported in Refs. [20,35], there is no bifurcation between zero to the
The critical buckling pressure, PKarman & Tsien, is given in equation (5) collapse loads (Riks method). From the analysis, it appears that the shell
failed in the (ii) axisymmetric collapse/snap-through mode, as the
(Et2 ) bifurcation analysis predicts a much higher buckling pressure than the
PKarman & Tsien = 0.3625 (5)
R2 Riks method. Thus, indicates the dome buckle is governed by a plastic
The critical buckling pressure, PKoiter, is given in equation (6) failure. Therefore, it can be said that the numerical model is suitable for
( t )2 this analysis.
PKarman & Tsien = 0.359E (6) Fig. 7 presents the path location from points A to B to record the
R stress distribution of an externally pressurised domed shell. The point of
Table 4 shows the range of buckling load Pcr of the externally pres­ interest in this analysis is the stress distribution of the shell at three (3)
surised dome cap with the range of dimensionless ratios of 100 < R/t < modes of failure evolution: (i) yield point, (ii) collapse, and (iii) post-
1000 with four (4) variants of boundary conditions. It can be observed collapse. The points of interest were taken along the meridional line
that the Type 4 boundary condition provides a conservative estimate of (e.g., the total shell length from the edge along the curvature to the apex
the critical buckling load compared with the other boundary conditions. of the domed cap). The starting point was taken at the domed bottom
In contrast, boundary condition type 1 and 3 produced a much larger node and top node of the shell. A total of 40 nodes were selected for
buckling load for shell size R/t < 250. It is seen that the critical load analysis. The Von Mises failure criterion was employed as a reference in
decreased significantly for shell sizes R/t > 250. For analytical calcu­ this study. Several conclusions can be drawn from these results.
lation, a drastic drop of buckling load was found near to 70 % estimated First, for the shallow shell, when pressure was applied, the dome
by using Karman & Tsien and Koiter approaches for the range of 100 < experienced a stress distribution of more than 300 MPa when the load
R/t < 1000. reached the yield strength condition (i.e., 0.949 MPa), as shown in
Fig. 4 shows the buckling load, PLBA of LBA mode with different Fig. 8. Second, as the load increased until collapse (i.e., 1.202 MPa), the
boundary conditions exemplified by R/t = 1000. For the case of stress distribution gradually shifted from the bottom to the top and
spherical shell size R/t = 1000, the shell leads to axisymmetric failure concentrated at the centre of the dome. A high stress concentration at
with boundary condition type 1 and 2 as shown in Fig. 4 (a)–(b). the centre of the flat cap is expected because the shell has a low spherical
Conversely, for the same size, the shells failed asymmetrically with 2 height, H, and therefore deflects downward easily. In the post-collapse
lobes of modeshape, under boundary conditions type 3 and 4, as shown condition (i.e., at 0.504 MPa), the stress distribution shifted again
in Fig. 4 (c)–(d). from the centre to the top and bottom of the dome cap. At this point, the
dome cap is permanently deformed. Because the top and middle regions
3.2. Geometric and material nonlinear analysis (GMNA) of the shell are very large, they are prone to failure or damage owing to
stress concentration.
The FE model was assumed to be a perfect dome using the geometric Fig. 9 shows the stress distribution of a dome subjected to external
and material nonlinear analysis (GMNA) approach following the pressure in the case of a hemispherical shell. As the load increased to the
recommendation of European Standard EN 1993-1-6 [36]. The ABAQUS yield state (1.723 MPa), the stress distribution initially concentrated in
nonlinear static Riks analysis was employed in the study. Figs. 5 and 6 the lowest section of the dome. Interestingly, before reaching the
show a typical plot of the external pressure against the vertical deflec­ maximum stress, the shell experienced a significant drop in stress in the
tion of the dome apex using D6 (shallow dome) and 1# (hemispherical lower section. This indicates a redistribution of stresses along the
dome) models as examples. Both figures show the (i) initial yield pres­ meridional length of the shell, whereas the lower portion of the shell also
sure, PYield; (ii) collapse pressure, PColl; and (iii) post-collapse pressure acts as a hinge. Second, in the collapse condition (i.e., at 2.890 MPa), a
significant increase in the stresses and their distribution was observed,
Table 3 with the lower part of the shell being the most severe. After collapse (i.e.,
Convergence study of externally pressurised spherical shell for nominal model. 2.4198 MPa), there was a remarkable decrease in the stress in the upper
Element number 1768 4873 6897 10,962
and middle parts of the shell (i.e., much lower than the distribution in
the yield state). Again, the stress is strongly concentrated at the bottom
Pcr [MPa] 0.70027 0.69846 0.69817 0.69789
of the shell.

4
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

Table 4
Critical buckling load of externally pressurised dome cap for nominal model.
Pcr [MPa]

R/t 1 2 3 4 Equation (4) Equation (5) Equation (6)

100 53.600 26.463 54.245 21.015 24.898 7.560 7.431

250 3.732 3.059 3.113 2.190 3.984 1.210 1.189
400 1.156 1.093 0.929 0.791 1.556 0.472 0.464
500 0.698 0.699 0.555 0.489 0.996 0.302 0.297
1000 0.148 0.147 0.118 0.111 0.249 0.076 0.074

Fig. 4. Buckling load, PLBA of LBA mode with different boundary conditions.

Fig. 5. Plot of load versus deflection of externally pressurised spherical shell for
specimen D6 with analysis type GMNA. Fig. 6. Plot of load versus deflection of externally pressurised spherical shell for
specimen 1# with analysis type GMNA.

5
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

Fig. 7. The path location from point A to B to record the distribution of stress.

Fig. 8. The distribution of stress of an externally pressurised domed shell for Fig. 9. The distribution of stress of an externally pressurised domed shell for
D6 model. 1# model.

3.3. Geometric and material nonlinear imperfection analysis (GMNIA) By this time, shell model #1 was employed as a baseline model.

In this section, the FE analysis was performed to evaluate the 3.3.1. Single load indentation imperfection (SLI)
imperfection of a spherical shell subjected to external pressure. The Fig. 10 (a) illustrates the SLI approach. The displacement-controlled
Geometric and Material Nonlinear Imperfection Analysis (GMNIA) ac­ perturbation load was applied laterally to the mid-section of the dome
cording to the European Standard EN 1993-1-6 [36] was used for the cap until the required indentation depth was reached. Once the required
analysis. Two types of imperfections were used, namely (i) the single indentation depth is obtained, the magnitude of perturbation load is
load indentation imperfection (SLI) and (ii) axisymmetric outward bulge held constant to eliminate springback while external pressure is applied
imperfection (AOB). These imperfections were separately superimposed at an incremental rate of nonlinear analysis via the Riks method until the
on the perfect model, with the range of imperfection amplitude versus collapse load is reached. The dome cap is more sensitive to imperfection
wall thickness, wo/t, varying between 0.0 and 2.5. The imperfection when the load is applied at the mid-section in comparison to the other
sensitivity analysis was performed for the shell range 0.179 <H/r < 1. sections. The dome cap is in a fully clamped boundary condition during

6
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

Fig. 10. The (i) SLI, (ii) AOB, and (iii) EM imperfection techniques exemplified by model #1.

the application of the lateral load and external pressure. 3.3.4. Results on imperfection sensitivity of dome cap
Figs. 11–13 show the response of the dome cap to the external
3.3.2. Axisymmetric outward bulge imperfection (AOB) pressure with the SLI, AOB and EM imperfections in the central part of
Fig. 10 (b) demonstrates the execution of AOB, by taking the buck­ the shell. The results in Fig. 11 show the imperfection sensitivity of H/r
ling deformation of the perfect dome cap from the nonlinear static Riks = 1 dome cap (i.e., a hemispherical shell). The results confirm that EM
analysis as the imperfection shape. Once a suitable buckling deforma­ has the worst imperfection compared to SLI and AOB at 0.0 < wo/t <
tion is found, which in this case occurs at 12th, 14th and 16th increment 2.0, specifically for the case of n = 2 and 3. The profile of the dome
(i.e., peak load), the shape is retained and varied in the subsequent modeshape n = 2 with 20 lobes, is shown in the insert of Fig. 11. At wo/t
analysis with a specified imperfection amplitude (i.e., between 0.0 < = 1, EM was the worst imperfection, with 64 % drop of buckling strength
wo/t < 2.5). followed by 3.5 % for AOB and less than 1 % for the case of SLI. It can be
seen that in the range of wo/t = 1.5–2.5, the SLI performs significantly
3.3.3. Eigenmode-affine imperfection (EM) worse than the AOB when estimating the KDF, although EM is still the
For the case of eigenmode imperfection, the modes were varied from lowest. At wo/t = 2.5, the range of differences was found to be around 9
n = 1, n = 2, and n = 3, as shown in Fig. 10 (c)–(e). The use of selected %–76 %. The SLI and AOB cause an insignificant marginal decrease in
eigenmodes from LBA is necessary in order to determine the lowest the buckling load at wo/t = 0.0–2.0.
knockdown factor of the shell. A nonlinear static analysis using the In Fig. 12, the KDF is plotted against the ratio of imperfection
modified static Riks method was performed to determine the collapse amplitude and wall thickness (wo/t) for the case of H/r = 0.59. The
pressure of the spherical shell. Again, similar to the previous ap­ result confirms that EM, n = 2 for being the worst case of imperfection.
proaches, the bottom of the spherical shell is assumed to be in a fully This is followed by EM, (i.e., modeshape 3 and 1), AOB, and SLI. The
constrained state during the application of the external pressure. insert of dome modeshape n = 2 with 15 lobes in seen in Fig. 12. The SLI

7
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

was found to be slightly insensitive to imperfection at 0.0 < wo/t < 1.6.
After wo/t = 1.6, there is a significant decrease of 8 %. The AOB
experienced a steady decrease in KDF from 0.0 < wo/t < 2.5. The SLI and
AOB shared an identical magnitude of decrease in load-carrying ca­
pacity of 9 % at wo/t = 2.5. Simultaneously, at wo/t = 2.5 the EM
imperfection (i.e., modeshape 1–3) calculated the dome’s loss of
strength to be around of 73 %.
For the case of H/r = 0.179, as shown in Fig. 13, the EM imperfec­
tion, n = 2 is more conservative than the other for the externally loaded
dome cap. Interestingly, at wo/t = 1.8, the AOB somewhat being
considerably insensitive to the imperfection in comparison with AOB.
The EM imperfection predicts a bearing capacity loss of approximately
62 %, whereas the other imperfection techniques calculated in the range
of 17 %–28 %.
Fig. 14 shows the sensitivity of the dome cap to (i) SLI and (ii) AOB
with a configuration range of 0.59 < H/r < 1.0 for a more conservative
estimation of the KDF compared to H/r = 0.179. However, the H/r =
0.179 shell experiences a significant drop in the KDF for both imper­
fection approaches. This clearly demonstrates its high sensitivity to
Fig. 11. Dome cap KDF against imperfection amplitude to wall thickness ratio,
imperfections. For the case of H/r = 0.59–1.0, the KDF differences seem
wo/t for the case of (i) SLI, (ii) AOB, and (iii) EM with H/r = 1.
to be insignificant. Interestingly, for the case of EM imperfection, n = 2,
H/r = 0.179 found to be more conservative in estimating the KDF
compared to others.
It can be concluded that the eigenmode shape imperfection is the
worst imperfection for the analysis. However, this is not necessarily true
for the case of a realistic imperfection. In practice, most imperfections in
structures do not exhibit buckling modeshape (i.e., deformation). This is
confirmed as some researchers have suggested that the EM imperfection
method is not suitable for the worst case of geometric imperfection [37,
38]. Most of the time, eigenmode imperfection is often used only
cautiously to estimate the lowest buckling load [16,18].

4. Dome cap buckling load with variation of geometric

configurations and material parameters

The imperfection sensitivity of the externally pressurised dome cap is

further investigated by applying the SLI imperfection technique. In
accordance with the European Standard EN 1993-1-6 [36], the
nonlinear analysis based on the geometric and material nonlinear
imperfection analysis (GMNIA) was performed.

Fig. 12. Dome cap KDF against imperfection amplitude to wall thickness ratio, 4.1. Parametric study accordance to European Standard EN 1993-1-6
wo/t for the case of (i) SLI, (ii) AOB, and (iii) EM with H/r = 0.59.
The imperfect spherical shells subjected to external pressure are
tested with the following setup.

● the imperfections of the spherical shell (i.e. shape, dent and ampli­
tude) are calculated according to the European standard EN 1993-1-
6 [36], based on equation (7)
√̅̅̅̅̅
< listaend > Lgx = 4 Rt (7)

● the spherical shell is assumed to be completely clamped at the

equatorial edge
● the dent is created by applying the lateral displacement load in the
mid-section of dome
● the buckling load is recorded by considering the dimensionless ratio
of Pimp/Py to Pcr/Py according to PD 5500 design code [9].

The input values employed for the parametric study are listed in
Table 5. The parametric variance was set in 3 dimensionless-ratio
groups, namely, (i) R/t, (ii) H/R, and (iii) σYield. Therefore, a total of
30 numerical cases were run via FE nonlinear analysis in the range of
Fig. 13. Dome cap KDF against imperfection amplitude to wall thickness ratio, input parameters.
wo/t for the case of (i) SLI, (ii) AOB, and (iii) EM with H/r = 0.179.

8
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

Fig. 14. The sensitivity of dome cap for the (i) SLI, (ii) AOB, and (iii) EM.

Table 5
The input parameters used for the parametric study.
Variable Description Variation range

R/t Spherical radius to thickness ratio 136.2 <R/t < 412.28

H/r Dome height to spherical based radius 0.179 <H/r < 1.0
ratio
σYield Yield stress 200 MPa <σYield < 400
MPa

4.1.1. Effect of geometric and material parameter

Fig. 15 plots the distribution of dimensionless ratio of Pimp/Py with
the effect of 0.179 <H/r < 1.0 together with upper and lower bounds
curves. Several comments can be made according to the parametric
study. It is clear that the buckling loads (i.e., Pimp/Py) also increase with
a higher yield stress. The plot of the upper and lower bounds curves
serves as a design recommendation similar to the NASA SP-8032
guideline, which is relevant in industrial applications as a safe and
conservation approach.
Fig. 16 shows a plot of the dimensionless ratio of Pimp/Py (y-axis) Fig. 15. The upper and lower bounds of dome cap.
versus Pcr/Py (x-axis) for the case of (i) 136.2 < R/t < 412.28, (ii) 0.075
< H/R < 1.0, and (iii) 200 MPa<σYield < 400 MPa. Several comments ( )3 ( )2 ( )
are underline with the following, the interactive behaviour of the PImp
= 0.0281
Pcr
−
Pcr
+ 0.8688
Pcr
(8)
dimensionless-ratio of Pimp/Py versus Pcr/Py follows the format of the PD Py Py Py Py
5500 design code. The distribution of the results demonstrate that the Fig. 17 shows a comparison of the external buckling pressure
buckling load increases significantly before it levels off at 2.5 <H/r < calculated using equation (8) with a total of 64 samples of experimental
3.0. Afterward, a gentle increase of buckling load is seen at 3.5 <H/r < results. The selection of experimental data is set to fall in the range of (i)
4.0. A polynomial curve fitting method for the scatter plot of each case 250 < R/t < 1000, (ii) 0.075 < H/R < 1.0, and (iii) 200 MPa<σYield <
tested after regression analysis yielded an approximate equation that 400 MPa. The experimental results are taken from literature provided by
appears to be reliable with an R2 value of 0.97 as derived in equation (8). Zhang et al. [19], Blachut [35], Ismail [18], Zhu [20], and Kołodziej & J.

9
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

as a linear function. From the given results, it can be summarised that.

● the external buckling pressure calculated by equation (8) agreed well

with tested experimental results, except for several cases of Zhang
et al. [19], Blachut [35] and Ismail [18], denoted by the symbols (+),
(▴), and (x)
● a good correlation analysis between the predicted equation with
experimental results yield to be 0.9793
● finally, the proposed equation (8) can be useful for first-cut estima­
tions of the buckling load for imperfect spherical shells under
external pressure within the range of parameter given

5. Dome cap design: a case study

In this section, a shell buckling example is benchmark with a real

case study of failed vessel. The study focuses on pressurization damage
to an oil-storage tank’s spherical shell partition, which was investigated
previously by Jones [39]. The given parameters for the spherical shell
Fig. 16. A plot of dimensionless-ratio of Pimp/Py against Pcr/Py for the case of partition are: E = 193 GPa, v = 0.28, t = 4.75 mm and r = 3000 mm (see
(i) R/t, (ii) H/R, and (iii) σYield. Fig. 18 (a)). The oil-storage tank was made using 316 stainless steel
material thus gives the yield stress to be σyield = 270 MPa.

5.1. Knockdown factor guidelines and numerical model

The empirical formula based on NASA SP-8032 [33] format, is very

inclusive and flexible when dealing with different shell buckling prob­
lems. To date, the use of the NASA SP-8032 guideline is still relevant in
the aerospace, subsea or pressurised tank sectors at the preliminary
design and fabrication stages. Moreover, all aerospace or subsea regu­
latory agencies have adopted this procedure as a safe and conservative
approach to implementing the lower bound of the buckling knockdown
factor curve. This is necessary to keep the safety factor of the structure at
the highest level. The NASA SP-8032 guideline is calculated under the
following condition.

● according to NASA SP-8032 [33] the shell parameter, λ are calculate

using equation (9)
√̅̅̅ (φ)
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ R
λ = 4 12(1 − ν2 ) 2 sin (9)
t 2

Fig. 17. Buckling pressure of the experiment results versus proposed equation. ● the dome cap knockdown factor (KDF) is calculated by equation (10)
3.2
Marcinowski [26]. The black solid line in the figure represents the < listaend > KDFNASA = 0.14 + (10)
λ2
estimated buckling pressure as a function of critical buckling load over
yield pressure, (Pcr/Py) according to PD 5500 guideline and is expressed Besides NASA SP-8032, the dome cap KDF is also compared with
other available knockdown factor formulas, as indicated in Table 6.

Fig. 18. (a) Simplified drawing of the storage tank. Dimensions in mm and (b) idealisation of numerical model with boundary condition [39].

10
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

Table 6 Table 8
Knockdown factor formulas with its corresponding description. Summary of numerical results.
Formulas Ref. Analysis LBA Bifurcation GMNA

0.693 λ > 5 [30] Pressure load [MPa] 0.579 0.120 0.499

KDFEBC = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 2
5
(1 − ν)λ5
KDFWagner = 5.172λ− 1.464 + 0.1296 λ > 5.5 [6]
3.2 λ > 5.5 [14]
KDFIsmail = 0.07 + 2
λ

Jones [39] describes that the pressure inside the tank/vessel is the sum
of (a) compressed air under a pressure of 0.1–0.2 MPa forced by oil
filling process and (b) the oil density about 800 kg/cm2, giving to a
possible hydrostatic pressure of 0.032 MPa. The initial tank design re­
view and its operational procedure are discussed below.

● the combining pressure of (a) + (b) leads to a range of 0.132

MPa–0.232 MPa internal pressures built-up inside the tank
● in this analysis the minimum internal pressure built-up inside the
tank is used as a threshold value, P = 0.132 MPa
● using equation (9), the shell parameter, λ found to be λ = 24
● using shell parameter, λ, the KDF of partition is calculated following
formulas outline in Table 6
● the KDF is then multiply with Pcr calculated using equation (1) to
estimate the partition design pressure
Fig. 19. Comparison of design pressure calculated based on various design
● to support the analysis, the design pressure of the partition is also
guidelines against minimum internal pressure built-up inside the tank.
calculated using ECCS Class A, B and C [7], DnV [8], ABS [10], and
PD 5500 [9] guidelines
i. During normal vessel/tank operation, the spherical shell partition
is expected to withstand the hydrostatic pressure of 0.032 MPa
The numerical analyses that consist of LBA, bifurcation and GMNA
produced from crude oil. Therefore, under normal conditions, the
were performed. The LBA and GMNA approach have been clearly
spherical shell partition is well designed, in which point it will
explained in previous section. The idealisation of numerical model with
not fail by having a load of 0.032 MPa.
specific boundary condition for the spherical partition is given in Fig. 18
ii. However, under unusual conditions, the disposed air pressure
(b). This boundary condition allowed the dome cap to move inward or
increases to 0.132 MPa (i.e., minimum internal pressure built-
outward around the edge.
up). Following the condition, the design pressure predicted by
For the case of bifurcation analysis, to account for pre-buckling
NASA SP-8032 [33], Ismail et al. [18], Wagner [6], ECCS Class C
nonlinearity in the shell, the dome was first preloaded with a dead
[7] and ABS [10] guidelines somehow under design the spherical
load through non-linear static analysis, followed by the live load in the
shell partition. This implies that the spherical shell partition fails,
buckle step via a subspace solver. Since different magnitude of dead load
according to the stated design guidelines.
results in different bifurcation pressure, it was decided to estimate the
iii. At the same time, according to the PD 5500 [9] and ECCS Class A
effect of using different combination of dead load and live load for the
and B [7] guidelines, the spherical shell partition remains intact
dome and evaluate it convergence. Usually, the live load employed in
with a built-up pressure of 0.132 MPa. Nonetheless, it is also
the current analysis is a fraction of 10 % to the magnitude of a dead load.
suggest that the design load also include a value of safety factor
Table 7 computed bifurcation buckling pressure and its convergent so­
(e.g. 4/3 in compliance with ECCS [7])
lution for the spherical partition. The summary of numerical results is
iv. The bifurcation pressure (i.e., Pbif = 0.120 MPa) from numerical
given in Table 8.
result confirmed to be much lower than the threshold pressure
value of tank. Meanwhile, the LBA and GMNA demonstrate an
5.2. Comments on the design of oil-storage tank
overestimation of bearing load for the spherical shell partition
Fig. 19 shows the comparison of design pressure calculated based on
several contemporary design guidelines against minimum internal
5.2.1. Failure process of spherical shell partition
pressure inside the vessel that acts as a threshold. From the given results
The damage to the top of the tank indicates that the upper tank was
it may summarise that.
subjected to internal pressure. Through observation also explained why
the partition had flipped downward-a classic example of a spherical cap
buckling under external pressure. Presumably, the road tanker dis­
Table 7 charged its entire load, allowing the compressed air in the road tank to
Computed bifurcation buckling pressure and its convergent solution for the come into contact with the oil surface in the upper tank. At the same
spherical partition. time, the bore of the vent out pipe was much smaller than that of the
Case Applied load [MPa] Bifurcation [MPa] filling hose. The smaller outer pipe and larger filling hose created a
Dead load Live load condition in which the air would have entered the tank much faster than
it could have escaped. This condition results in a concentrated external
1 0.010 0.001 0.278 (36)
2 0.050 0.005 0.183 (34) pressure on the partition, flipping it downward.
3 0.100 0.010 0.123 (24)
4 0.110 0.011 0.120 (22) 5.2.2. Spherical shell partition stress assessment
5 0.120 0.012 0.121 (20) The stress state in the partition was calculated using the principal

11
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

stresses σ1 = σ2 = PR/2t and σ3 = 0. The assessment of spherical shell stress assessment are evaluated. Several remedies to overcome
stresses is as follows. the failure process are also outlined accordingly.

i. The value of the pressure in this case is the maximum range of To conclude, the results of this study are highly useful and contribute
pressure built-up, P = 0.232 MPa, thus giving the principal stress to the body of knowledge on the effects of structural imperfections in the
to be σ1 = σ2 ≤ 73 MPa. preliminary stages of the design, fabrication, and analysis of externally
ii. Since the oil-storage tank was made using 316 stainless steel pressurised dome cap structures.
material thus gives the yield stress to be σyield = 270 MPa.
Therefore, the stress state of a spherical shell partition can be CRediT authorship contribution statement
summarised by following the condition, σ1 = σ2 ≤ 73 MPa ≤270
MPa. Mohd Shahrom Ismail: Writing – review & editing, Writing –
iii. The following scenario prevails that buckling analysis assumes original draft, Visualization, Validation, Supervision, Project adminis­
that the cap remained elastic while the buckling instability tration, Methodology, Investigation, Funding acquisition, Formal anal­
initiated, thereby supporting the suggestion that partition was ysis, Conceptualization. Syed Mahathir Muhammad al-Attas:
well below yield when the failure initiated. Validation, Resources. Jamaluddin Mahmud: Resources, Funding
acquisition.
5.3. Remedies to the failure process

Finally, several suggestions and recommendations to improve the Declaration of competing interest
design are provided in this section. The following comments are high­
lights accordingly. The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
● it is suggested that the vent pipe should be redesigned to be slightly the work reported in this paper.
larger to compensate for the air pressure produced by the filling
process Data availability
● the redesign of vent pipe is necessary to avoid or minimize the
pressure build-up Data will be made available on request.
● since the failure initiated below yield, it is suggesting to slightly in­
crease the thickness of spherical shell partition up to 5 % to with­ References
stand the unusual conditions of pressure built-up
[1] P. Jasion, Stability analysis of shells of revolution under pressure conditions, Thin-
6. Conclusion Walled Struct. 47 (2009) 311–317, https://doi.org/10.1016/j.tws.2008.07.005.
[2] S.M. Tripathi, S. Anup, R. Muthukumar, Effect of geometrical parameters on mode
shape and critical buckling load of dished shells under external pressure, Thin-
The numerical results of the dome cap shell pressurised from the Walled Struct. 106 (2016) 218–227, https://doi.org/10.1016/j.tws.2016.02.011.
outside are presented in this paper. Several conclusions can be drawn [3] J. Błachut, Experimental perspective on the buckling of pressure vessel
components, Appl. Mech. Rev. 66 (2014) 1–24, https://doi.org/10.1115/
from the preceding analysis. 1.4026067.
[4] J. Błachut, K. Magnucki, Strength, stability, and optimization of pressure vessels:
i. In the benchmark study, the nonlinear Riks analysis provided review of selected problems, Appl. Mech. Rev. 61 (2008) 060801, https://doi.org/
10.1115/1.2978080. –1.
good agreement with the experimental results compared with the [5] J. Błachut, Combined stability of geometrically imperfect conical shells, Thin-
PD 5500 design code. The average percentage difference between Walled Struct. 67 (2013) 121–128.
the experimental results and numerical analysis was found to be [6] H.N.R. Wagner, C. Hühne, S. Niemann, Robust knockdown factors for the design of
spherical shells under external pressure: development and validation, Int. J. Mech.
less than 7 % for all cases. For PD 5500, the corresponding result Sci. 141 (2018) 58–77, https://doi.org/10.1016/j.ijmecsci.2018.03.029.
shows that the buckling load of the 5# model is closer to the [7] ECCS, Buckling of Steel Shells European Design Recommendations, Buckling
experimental data, with a difference of 6 % (i.e., PPD 5500/PExp = Shells, fifth ed., Brussels Eur. Conv. Constr. Steelwork., 2008.
[8] DnV, DNV RP-C202: Buckling Strength of Shells, Det Nor. Verit. AS., 2013, p. 27.
0.943).
[9] BS5500, BS5500: Specification of Unfired Fusion Welded Pressure Vessels, British
ii. The dome cap shell was designed according to EN 1993-1-6, Standards Institution, London, England, 1988. https://shop.bsigroup.com/uplo
which underwent the design processes of LBA, GMNA, and ad/Sector/Manufacturing/PD5500_2012_updatebrochure.pdf. (Accessed 16
GMNIA. Under the LBA, the buckling load Pcr of the externally October 2017).
[10] American Bureau of Shipping, Underwater vehicles, systems and hyperbaric
pressurised dome shell was calculated with Type 4 boundary facilities, in: Various Pagings, 2021, p. 77.
condition provides a conservative estimate of the critical buck­ [11] B.B. Pan, W.C. Cui, Y.S. Shen, T. Liu, Further study on the ultimate strength
ling load compared with the other boundary conditions. Under analysis of spherical pressure hulls, Mar. Struct. 23 (2010) 444–461, https://doi.
org/10.1016/j.marstruc.2010.11.001.
GMNA, the dome cap shell load-carrying capacity and stress [12] A. Spagnoli, Koiter circles in the buckling of axially compressed conical shells, Int.
distribution were analysed and discussed thoroughly. In the case J. Solids Struct. 40 (2003) 6095–6109, https://doi.org/10.1016/S0020-7683(03)
of GMNIA, the dome was tested under different types of imper­ 00369-X.
[13] S.G.P. Castro, R. Zimmermann, M.A. Arbelo, R. Khakimova, M.W. Hilburger,
fection approaches. The following result confirms that EM R. Degenhardt, Geometric imperfections and lower-bound methods used to
imperfection has the worst imperfection compared to others. calculate knock-down factors for axially compressed composite cylindrical shells,
iii. The imperfection sensitivity of the externally pressurised dome Thin-Walled Struct. 74 (2014) 118–132.
[14] C.Y. Song, J.G. Teng, J.M. Rotter, Imperfection sensitivity of thin elastic cylindrical
cap was further investigated by applying the SLI imperfection shells subject to partial axial compression, Int. J. Solids Struct. 41 (2004)
technique to various shell configurations and material parame­ 7155–7180, https://doi.org/10.1016/j.ijsolstr.2004.05.040.
ters. A plot of the upper and lower bound curves is plotted to [15] M. Shakouri, A. Spagnoli, M.A. Kouchakzadeh, Re-interpreting simultaneous
buckling modes of axially compressed isotropic conical shells, Thin-Walled Struct.
serve as a design recommendation, similar to the NASA SP-8032
84 (2014) 360–368, https://doi.org/10.1016/j.tws.2014.07.013.
guideline. Simultaneously, a polynomial curve fitting method for [16] M.S. Ismail, O. Ifayefunmi, S.H.S.M. Fadzullah, Buckling of imperfect cylinder-
the scatter plot of each case tested after regression analysis yiel­ cone-cylinder transition under axial compression, Thin-Walled Struct. 144 (2019)
ded an approximate equation with R2 = 0.97. 106250, https://doi.org/10.1016/j.tws.2019.106250.
[17] M.S. Ismail, O. Ifayefunmi, S.H.S.M. Fadzullah, M. Johar, Buckling of imperfect
iv. Finally, a case study of shell buckling design on pressurization cone-cylinder transition subjected to external pressure, Int. J. Press. Vessel. Pip.
damage to an oil storage tank’s spherical shell partition and its 187 (2020) 104173, https://doi.org/10.1016/j.ijpvp.2020.104173.

12
M.S. Ismail et al. International Journal of Pressure Vessels and Piping 208 (2024) 105135

[18] M.S. Ismail, J. Mahmud, A. Jailani, Buckling of an imperfect spherical shell [29] Y. Wang, J. Zhang, W. Tang, Buckling performances of spherical caps under
subjected to external pressure, Ocean. Eng. 275 (2023) 114118, https://doi.org/ uniform external pressure, J. Mar. Sci. Appl. 19 (2020) 96–100, https://doi.org/
10.1016/j.oceaneng.2023.114118. 10.1007/s11804-020-00125-7.
[19] J. Zhang, Y.Y. Wang, F. Wang, W. Tang, Buckling of stainless steel spherical caps [30] M.S. Ismail, J. Mahmud, Comparative evaluation of design codes for buckling
subjected to uniform external pressure, Ships Offshore Struct. 13 (2018) 779–785, assessment of a steel spherical shell, Lat. Am. J. Solid. Struct. 20 (2023) 1–11,
https://doi.org/10.1080/17445302.2018.1459358. https://doi.org/10.1590/1679-78257473.
[20] Y. Zhu, Y. Zhang, X. Zhao, J. Zhang, X. Xu, Elastic–plastic buckling of externally [31] L. Zhao, Y. Bai, Ultimate strength models for spherical shells under external
pressurised hemispherical heads, Ships Offshore Struct. 14 (2019) 829–838, pressure: a comparative study, Ships Offshore Struct. (2022) 1–12, https://doi.org/
https://doi.org/10.1080/17445302.2018.1564541. 10.1080/17445302.2022.2126115.
[21] S.R. Cho, T. Muttaqie, S.H. Lee, J. Paek, J.M. Sohn, Ultimate strength assessment of [32] R. Zoelly, Ueber ein Knickungsproblem an der Kugelschale Ober ein
steel-welded hemispheres under external hydrostatic pressure, J. Mar. Sci. Appl. 19 Knickungsproblein, 1915.
(2020) 615–633, https://doi.org/10.1007/s11804-020-00178-8. [33] V.I. Wingarten, P. Seide, NASA SP-8032: Buckling of Thin-Walled Doubly Curved
[22] J. Błachut, Locally flattened or dented domes under external pressure, Thin-Walled Shells, 1969, pp. 1–33.
Struct. 97 (2015) 44–52, https://doi.org/10.1016/j.tws.2015.08.022. [34] A.Y. Evkin, O.V. Lykhachova, Energy barrier as a criterion for stability estimation
[23] G.D. Galletly, J. Kruzelecki, D.G. Moffat, B. Warrington, Buckling of shallow of spherical shell under uniform external pressure, Int. J. Solids Struct. 118–119
torispherical domes subjected to external pressure — a comparison of experiment, (2017) 1339–1351, https://doi.org/10.1016/j.ijsolstr.2017.04.026.
theory, and design codes, J. Strain Anal. Eng. Des. 22 (1987) 163–175, https://doi. [35] J. Błachut, Buckling of shallow spherical caps subjected to external pressure,
org/10.1243/03093247V223163. J. Appl. Mech. Trans. ASME. 72 (2005) 803–806, https://doi.org/10.1115/
[24] J. Błachut, Buckling of composite domes with localised imperfections and 1.1993667.
subjected to external pressure, Compos. Struct. 153 (2016) 746–754, https://doi. [36] Eurocode 3, Eurocode 3 - Design of Steel Structures - Part 1-6: Strength and
org/10.1016/j.compstruct.2016.07.007. Stability of Shell Structures, 2007.
[25] J. Błachut, G.D. Galletly, Buckling strength of imperfect steel hemispheres, Thin- [37] S.G.P. Castro, R. Zimmermann, M.A. Arbelo, R. Degenhardt, The single
Walled Struct. 23 (1995) 1–20, https://doi.org/10.1016/0263-8231(95)00001-T. perturbation load approach compared with linear buckling mode-shaped ,
[26] S. Kołodziej, J. Marcinowski, Experimental and numerical analyses of the buckling geometric dimple and measured imperfections for the buckling of cylindrical
of steel, pressurized, spherical shells, Adv. Struct. Eng. 21 (2018) 2416–2432, shells, Thin-Walled Struct. (2013) 1–14.
https://doi.org/10.1177/1369433218774371. [38] M.A. Arbelo, R. Degenhardt, S.G.P. Castro, R. Zimmermann, Numerical
[27] H.N.R. Wagner, C. Hühne, J. Zhang, W. Tang, R. Khakimova, Geometric characterization of imperfection sensitive composite structures, Compos. Struct.
imperfection and lower-bound analysis of spherical shells under external pressure, 108 (2014) 295–303, https://doi.org/10.1016/j.compstruct.2013.09.041.
Thin-Walled Struct. 143 (2019) 106195, https://doi.org/10.1016/j. [39] D.R.H. Jones, Buckling failures of pressurised vessels-two case studies, Eng. Fail.
tws.2019.106195. Anal. 1 (1994) 155–167.
[28] H.N.R. Wagner, C. Hühne, J. Zhang, W. Tang, On the imperfection sensitivity and
design of spherical domes under external pressure, Int. J. Press. Vessel. Pip. 179
(2020) 104015, https://doi.org/10.1016/j.ijpvp.2019.104015.

13