1 s2.0 S0951832014001719 Main

Reliability Engineering and System Safety 132 (2014) 125145
Contents lists available at ScienceDirect
Reliability Engineering and System Safety

journal homepage: www.elsevier.com/locate/ress
Release of hazardous substances in ood events: Damage model

for horizontal cylindrical vessels
Gabriele Landucci a, Amos Necci b, Giacomo Antonioni b,
Alessandro Tugnoli b, Valerio Cozzani b,n
a
Dipartimento di Ingegneria Civile e Industriale Universit di Pisa, largo Lucio Lazzarino, 56126 Pisa, Italy
Dipartimento di Ingegneria Civile, Chimica, Ambientale e dei Materiali, Alma Mater Studiorum Universit di Bologna,
via Terracini n.28, 40131 Bologna, Italy
ar t ic l e i nf o
a b s t r a c t
Article history:
Received 1 August 2013
Received in revised form
14 June 2014
Accepted 17 July 2014
Available online 1 August 2014
Severe accidents may be triggered by the impact of oods on process and storage equipment containing
hazardous substances. The present study analyses the possible damage of horizontal cylindrical
equipment, either operating at atmospheric or at higher pressures. A mechanical damage model was
developed and validated by available literature data on past accidents. Simplied correlations were then
obtained to calculate the critical ooding conditions leading to vessel failure. A fragility model was
proposed for the straightforward assessment of equipment damage probability in the framework of the
quantitative risk assessment of NaTech scenarios triggered by oods. A case-study was discussed to test
the potentialities of the method.
& 2014 Elsevier Ltd. All rights reserved.
Keywords:
NaTech
Flood
Major accident hazard
Hazardous materials release
Storage tanks
Pressure vessels
1. Introduction
Global climate changes may signicantly inuence temperatures and precipitation patterns, increasing the frequency of severe
meteorological events which may lead to ooding [15]. Flooding
may dramatically impact on population among urban areas but it
can also affect industrial facilities. In the latter case, damages to
equipment and structures may lead to the release of hazardous
materials with potential escalation leading to domino effects
[612]. This type of accident scenario is indicated as a NaturalTechnological (NaTech) event. Specic studies [1319] and recent
publications concerning the analysis of industrial accidents
reported in available databases [11,20] highlighted that NaTech
accidents often generate severe consequences. Besides, the severe
events that affected several industrial sites and the Fukushima
power station after the Thoku earthquake and consequent
tsunami in Japan (April 2011) [13,21] showed the potential dramatic consequences of NaTech events.
Therefore, NaTech events may signicantly impact on the risk
prole of an industrial facility. However, the implementation of
NaTech scenarios in conventional Quantitative Risk Assessment
(QRA) studies is a critical task [16,2224]. In particular, the
Corresponding auth. Tel.: 39 051 2090240; fax: 39 051 2090247.

E-mail address: valerio.cozzani@unibo.it (V. Cozzani).
http://dx.doi.org/10.1016/j.ress.2014.07.016
0951-8320/& 2014 Elsevier Ltd. All rights reserved.
frequency assessment needs to take into account both parameters

related to the likelihood of the natural event and to the credibility
of equipment failures following the impact of the natural event. In
the case of NaTech scenarios caused by ooding, the rst type of
input data can be usually retrieved from national or local competent Authorities. Specic equipment fragility models must be
available in order to generate the data concerning the credibility
of equipment damage. Such models are aimed at the estimation of
equipment damage probability on the basis of the severity of the
natural event. Due to the features of a QRA study that usually
requires the assessment of a high number of scenarios, the use
of simplied models able to yield a conservative estimation of
equipment failure conditions is required to effectively support the
assessment of equipment vulnerability [2227].
Fragility curves and equipment vulnerability models are available for several equipment items in the case of earthquake [17,21].
In a previous study equipment vulnerability models were obtained
for atmospheric vertical cylindrical storage tanks in the case of
ood scenarios [8]. Specic fragility models for equipment vulnerability in the case of ood are thus not available for horizontal
cylindrical storage tanks. Past accident data analysis [7] evidenced
that these equipment items were often damaged in NaTech events
triggered by oods. However, since these vessels are usually
positioned on saddles or, more in general, are welded to supports
anchored to the ground, ooding damage occurs by different
mechanisms with respect to vertical cylindrical tanks. Actually,
126
G. Landucci et al. / Reliability Engineering and System Safety 132 (2014) 125145
horizontal cylindrical vessels are mostly damaged due to displacement caused by water drag and/or to oating [7,28].
The present study was devoted to the development of a model
for the vulnerability assessment of horizontal cylindrical process
and storage vessels involved in ood events. In order to evaluate
the resistance of the equipment items considered, a mechanical
model was developed. The model, validated with respect to the
available data obtained from past accident records, was applied to
derive simplied vulnerability functions to calculate vessel failure
probability in ood events. In order to explore the model features
and potentialities some case studies based on actual industrial
lay-outs were analyzed.
obtained from actual data available from industrial facilities and

from current design standards. In particular, pressurized vessels
(dened as vessels operating at pressures higher than 103.4 kPa
[31]) and atmospheric vessels were both included in the database.
The vessel database was used to obtain a dataset of failure
conditions with respect to ood intensity parameters (step 5).
In step 6, the dataset of failure conditions and vessel properties obtained in step 5 was used to extrapolate simplied
damage correlations for the calculation of vessel failure probability
(step 7).
2.1. Representation of vessel geometry (step 1)

2. Model
The approach proposed to assess the vulnerability of equipment items involved in ood events is schematized in Fig. 1. The
key features of the methodology were derived from a previous
study [8] and are briey summarized in the following, evidencing
the specic elements related to horizontal vessels.
As shown in Fig. 1, in the rst step of model development a
simplied representation of the vessel geometry and of its supports was introduced. In step 2, on the basis of this schematization, a mechanical model was developed, able to evaluate the
effects of oodwater impact on the vessel. In particular, in the case
of horizontal vessels, damages are caused by the displacement of
vessels due to water drag and/or to oating, leading to the rupture
of the connected pipelines and to a potential impact on the other
equipment items or structures. Hence, on the basis of available
data on past ooding events [29,30], the ood water impact was
schematized considering a credible range of values for ood water
depth and velocity (step 3). Reference data from past events were
then used for model validation (step 4). In step 5, a vessel database
was developed, including an extended set of representative
geometries of horizontal separators and vessel storage tanks,
Representation of vessel
geometry
Mechanical model set-up
Characterization of flood
impact vector
Model validation
Extended data set on

failure conditions
Development of
simplified correlations
Evaluation
of vessels vulnerability
Fig. 1. Schematization of the methodological approach adopted for the development of a vulnerability model for horizontal vessels involved in ood events.
In the present study, storage tanks and process vessels consisting of a horizontal cylindrical body with spherical edges were
considered. The vessels, operating at atmospheric pressure or
higher, were assumed as disposed on saddle-type supports, xed
to the ground. The references for the design and features of the
tanks considered in the present study are the API Standard 620
[32] and the ASME Pressure Vessel Code (Sec. VIII of the ASME
Boiler and Pressure Vessel Code [31]).
The schematic representation of these vessels is reported in
Fig. 2a, while the relevant mechanical features considered are
summarized in Table 1. As shown in Fig. 2a, one of the vessel
saddles is assumed to be xed to the ground with a bolt connection, while the other assumed to be only laid on the ground. This
conguration is frequently adopted in the process industry in
order to limit the stress due to steelwork thermal expansion [33].
Fig. 2. Schematization adopted to describe the impact of oodwater on horizontal

vessels: (a) denition of geometrical parameters and sketch of the vessel; (b) force
balance on the vessel and schematization of the wave impact; (c) schematization of
the base plate bolt connection to the ground.
Table 1
Main features of the vessels included in the database considered in the present study. For the denition of geometrical parameters (dimensions) see Fig. 2. Pd design pressure; ATM operating at atmospheric pressure.
ID
Capacity
3
Dimensions (mm)
(m )
l1
l2
l3
S1
S2
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
ATM
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
1.5
2
2
2
2
2
5
5
5
10
10
15
20
20
20
25
25
25
30
30
50
50
50
50
100
100
100
115
150
250
5
5
5
10
10
15
20
20
20
25
25
25
30
30
50
50
50
50
100
100
100
115
150
250
5
5
5
10
10
1600
1300
1000
1600
1200
1800
1500
2000
1900
1700
2200
2300
2400
1900
2100
2800
2700
2500
3200
3200
2800
2750
3200
3800
1600
1300
1000
1600
1200
1800
1500
2000
1900
1700
2200
2300
2400
1900
2100
2800
2700
2500
3200
3200
2800
2750
3200
3800
1600
1300
1000
1600
1200
3000
3500
6100
4500
7700
5500
9700
6000
7200
10,500
6000
7000
6500
11,100
13,200
8000
10,000
10,400
12,000
13,700
18,000
20,100
19,400
24,000
3000
3500
6100
4500
7700
5500
9700
6000
7200
10,500
6000
7000
6500
11,100
13,200
8000
10,000
10,400
12,000
13700
18,000
20,100
19,400
24,000
3000
3500
6100
4500
7700
5
5
5
5
5
5
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
6
11
9
7
11
8
12
10
13
12
11
14
15
16
12
14
18
17
16
21
21
18
18
21
24
14
11
9
14
11
800
650
500
800
600
900
750
1000
950
850
1100
1150
1200
950
1050
1400
1350
1250
1600
1600
1400
1375
1600
1900
800
650
500
800
600
900
750
1000
950
850
1100
1150
1200
950
1050
1400
1350
1250
1600
1600
1400
1375
1600
1900
800
650
500
800
600
980
780
680
980
780
1080
880
1180
1080
980
1280
1280
1380
1080
1180
1580
1480
1380
1780
1780
1580
1580
1780
1980
980
780
680
980
780
1080
880
1180
1080
980
1280
1280
1380
1080
1180
1580
1480
1380
1780
1780
1580
1580
1780
1980
980
780
680
980
780
2200
2850
5600
3700
7100
4600
8950
5000
6250
9650
4900
5850
5300
10,150
12,150
6600
8650
9150
10,400
12,100
16,600
18,725
17,800
22,100
2200
2850
5600
3700
7100
4600
8950
5000
6250
9650
4900
5850
5300
10,150
12,150
6600
8650
9150
10,400
12,100
16,600
18,725
17,800
22,100
2200
2850
5600
3700
7100
140
140
95
140
140
140
140
140
140
140
150
150
150
140
140
150
150
150
150
150
150
150
150
150
140
140
95
140
140
140
140
140
140
140
150
150
150
140
140
150
150
150
150
150
150
150
150
150
140
140
95
140
140
1240
900
780
1240
900
1420
1060
1600
1420
1240
1780
1780
1960
1420
1600
2200
2060
1960
2520
2520
2200
2200
2520
2800
1240
900
780
1240
900
1420
1060
1600
1420
1240
1780
1780
1960
1420
1600
2200
2060
1960
2520
2520
2200
2200
2520
2800
1240
900
780
1240
900
127
(MPa)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
Pd
128
Table 1 (continued )
ID
Capacity
3
Dimensions (mm)
(MPa)
(m )
l1
l2
l3
S1
S2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
2.5
15
20
20
20
25
25
25
30
30
50
50
50
50
100
100
100
115
150
250
5
5
5
10
10
15
20
20
20
25
25
25
30
30
50
50
50
50
100
100
100
115
150
250
1800
1500
2000
1900
1700
2200
2300
2400
1900
2100
2800
2700
2500
3200
3200
2800
2750
3200
3800
1600
1300
1000
1600
1200
1800
1500
2000
1900
1700
2200
2300
2400
1900
2100
2800
2700
2500
3200
3200
2800
2750
3200
3800
5500
9700
6000
7200
10,500
6000
7000
6500
11,100
13,200
8000
10,000
10,400
12,000
13,700
18,000
20,100
19,400
24,000
3000
3500
6100
4500
7700
5500
9700
6000
7200
10,500
6000
7000
6500
11,100
13,200
8000
10,000
10,400
12,000
13,700
18,000
20,100
19,400
24,000
16
13
17
16
15
19
20
21
16
18
24
23
22
27
27
24
24
27
32
17
14
11
17
13
19
16
22
20
18
24
25
26
20
23
30
29
27
34
34
30
29
34
40
900
750
1000
950
850
1100
1150
1200
950
1050
1400
1350
1250
1600
1600
1400
1375
1600
1900
800
650
500
800
600
900
750
1000
950
850
1100
1150
1200
950
1050
1400
1350
1250
1600
1600
1400
1375
1600
1900
1080
880
1180
1080
980
1280
1280
1380
1080
1180
1580
1480
1380
1780
1780
1580
1580
1780
1980
980
780
680
980
780
1080
880
1180
1080
980
1280
1280
1380
1080
1180
1580
1480
1380
1780
1780
1580
1580
1780
1980
4600
8950
5000
6250
9650
4900
5850
5300
10,150
12,150
6600
8650
9150
10,400
12,100
16,600
18,725
17,800
22,100
2200
2850
5600
3700
7100
4600
8950
5000
6250
9650
4900
5850
5300
10,150
12,150
6600
8650
9150
10,400
12,100
16,600
18,725
17,800
22,100
140
140
140
140
140
150
150
150
140
140
150
150
150
150
150
150
150
150
150
140
140
95
140
140
140
140
140
140
140
150
150
150
140
140
150
150
150
150
150
150
150
150
150
1420
1060
1600
1420
1240
1780
1780
1960
1420
1600
2200
2060
1960
2520
2520
2200
2200
2520
2800
1240
900
780
1240
900
1420
1060
1600
1420
1240
1780
1780
1960
1420
1600
2200
2060
1960
2520
2520
2200
2200
2520
2800
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
Pd
Fig. 2c shows in detail the schematization of the saddle base plate

bolt connection to the ground assumed in the present analysis.
2.2. Mechanical model set-up (step 2)
Flooding may cause damages to equipment or structures due to
the pressure associated to both water speed (vw) and water
effective depth (hw). Atmospheric vertical cylindrical vessels were
found to undergo failures in ood events mainly as a consequence
of buckling due to water external pressure [7,8,28].
Horizontal cylindrical vessels are likely to experience different
failure mechanisms in ood events. The analysis of past accidents
involving ooding in industrial facilities [7,28] evidenced that
failure caused by buckling was never experienced for horizontal
vessels. Actually, horizontal cylindrical vessels, either atmospheric
or pressurized, have a higher resistance to buckling with respect to
vertical cylindrical storage tanks used in industrial applications,
due to the higher thickness/diameter ratio (typically 2 orders of
magnitude, as evident from the database reported in Table 1 and
from the literature [8]). The analysis of past accidents evidenced
that this category of vessels is more prone to undergo failures due
to displacement in ood events. In particular, the rupture of the
saddle framework and of its connection to the ground was
experienced in several cases. This resulted in the displacement of
the vessel, with the consequent rupture of the vessel connections
and, in some cases, in the impact of the displaced vessel with
adjacent vessels or structures [34,35].
Therefore, the mechanical model was developed focusing on
the integrity of the saddle-type support, that is the element
required to fail in the case of vessel displacement caused by ood
water. In particular, the resistance of the saddle connection to the
ground under load conditions caused by oodwater was investigated. Table 2 summarizes the procedure applied for the evaluation of the framework connection resistance to the ood water
load.
The schematic representation of the oodwater impact on a
horizontal cylindrical vessel is reported in Fig. 2b. As shown in the
gure, the vessel is subject to the hydrostatic lift force (Fh in
Fig. 2b), which causes a vertical lift action, and, at the same time,
to a horizontal drag force caused by the ood wave (Fv in Fig. 2b).
Buoyancy (Fn in Fig. 2b) is the net force obtained considering the
opposite actions of hydrostatic lift and of overall weight force (Ft in
Fig. 2b), resulting from vessel weight and from the weight of the
uid inside the vessel:
Fn Fh Ft
Buoyancy may thus be expressed as a function of vessel geometry,

weight and lling level:
Fn V ext w g gW t V int l 1 V int v
where g is the gravity constant (9.81 m/s ), Wt is the tare weight

of the tank (i.e. the mass of the empty tank in kg), l is the liquid
average density, v is the average density of the vapor in the top
space of the vessel, w is the density of oodwater, Vint and Vext are
the inner and outer vessel volume respectively, and is the vessel
volumetric lling level dened as the fraction of liquid volume
respect to the total vessel inner volume Vint. If the value of vessel
weight, Wt, is not available, this parameter may be estimated as
follows, assuming the value of steel density s:
Wt s V ext V int
The parameter in Eq. (2) is the fraction of the vessel volume

wetted by ooding. Geometrical relationships summarized in
Table 2, which take into account the effective depth hw of ood
water, and the height of the saddles, l2, may be used for the
129
evaluation of . Further details on the calculation of are provided

in Section 2.3.
The drag force due to the oodwater wave kinetic energy (Fv)
may be calculated as follows [36,37]:
F v 12 w kw v2w Aext
where kw is the hydrodynamic coefcient [36,37] and Aext is the

projected area of the vessel external surface impacted by ooding
in a plane normal to water ow. In order to obtain a conservative
evaluation for the drag force, the water ow is assumed to impact
on the side of the vessel featured by a higher external surface (see
Fig. 2b). Table 2 summarizes the procedure for the evaluation of
Aext in Eq. (2), based on the vessel geometrical features and on
water depth at vessel location, hw.
Both the buoyancy, Fn, and the drag force, Fv acting on the
vessel generate a stress on the vessel support (e.g., the saddle
connected to the ground). Assuming that the vessel connection to
the saddle is a dap-joint (i.e. a connection of innite rigidity) the
forces acting on the vessel directly affect the bolt connection
between the saddle base plate and the ground, as shown in Fig. 2c.
As usual in engineering practice, it is assumed that a total number
nb of bolts having the same features characterize the connection.
The buoyancy Fn causes a normal stress on each of the nb bolts of
the connection [33,38]:
F n;b F n =nb
where Fn,b is the portion of normal force allocated to each bolt.

Beside the direct action of buoyancy on the connection, the
contribution of the bending moment Mb must be taken into
account. In fact, only one of the two vessel saddles is anchored
to the ground (see Fig. 2a and b), thus the buoyancy generates a
bending moment on the support. The bending moment Mb may be
expressed as follows:
M b 12 F n l3
where l3 is the distance between the anchored saddle and the

vessel edge (see Fig. 2a). In the absence of more specic data, the
position of the saddle can be conservatively assumed in correspondence of the end of the cylindrical part of the vessel, as shown
in Fig. 2a (thus, l3 L l1). The action of the bending moment is
distributed over the nb bolts of the plate according to their
distance from the neutral axis of the plate (xi) as follows:
F x;i
Mb
nb
xi
x2i
i1
where xi is the distance of the i-th bolt from the neutral axis of
the plate (see Fig. 2c).
Finally, the overall action of buoyancy on the i-th bolt (Ntot,i)
may be evaluated summing the contribution of the normal net
force (Eq. (5)) and of the bending moment (Eq. (7)):
N tot;i F n;b F x;i
It is worth to mention that buoyancy may also induce a bending

moment opposite to Mb, thus acting on the bolts of the connection as a
negative force (see the coordinate system in Fig. 2b) which results in
an attenuation of the effective Fx,i value. This effect was neglected in
the model, thus obtaining conservative results.
A similar procedure allows evaluating the action on each bolt of
the drag force, Fv, which affects the nb bolts of the connection as
follows [33,38]:
F v;b F v =nb
where Fv,b is the portion of shear force allocated to each bolt. Also
in this case, the bolts undergo a supplementary force due to the
torque caused by the vessel rotation induced by the drag force (see
130
Table 2
Summary of input parameters implemented in the present. study DB: derived from the database provided in Table 1, CALC: parameter calculated on the basis of input data.
Operation
ID
Item
(1) Selection of input parameters:

characterization of the equipment
and storage system
1.1
1.2
1.3
1.4
1.5
Pd
D
L
t
l1
1.6
l3

characterization of the equipment
framework

characterization of the ooding
scenario
(4) Evaluation of vessel resistance with

the mechanical model
1.7
1.8
1.9
1.10
min
max
Tst
1.11
Pst
1.12
1.13
Description/Denition
Value
Design pressure of the vessel

Storage vessel external diameter
Storage vessel length (end to end)
Minimum thickness of the vessel wall
Distance of the saddle from the vessel edge (Fig. 2a)
Distance between the anchored saddle and the opposite vessel end (See
Fig. 2a)
Operative lling level dened as the fraction of liquid volume respect to the
total vessel inner volume
Minimum operative lling level
Maximum operative lling level
Storage temperature (ambient temperature)
Storage pressure. Pst 0.01 barg for atmospheric tanks, Pst vapor pressure of
stored liquid evaluated at Tst for pressurized vessels.
Density of stored liquid phase
Density of vapor phase in the top space of the vessel v P st M w =ZRT st
where Z is the compressibility factor [54] and Mw the vapor molecular weight
[54]
Density of steel (vessel construction material)
Density of the reference substance used for the denition of CFL correlations
(see steps 5.3 and 5.4)
0.01a-2.5
DB
DB
DB
DB
MPa
m
m
m
m
DB
1.14
1.15
ref
1.16
Vext
1.17
Vint
1.18
Wt
2.1
l2
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
2.10
S1
S2
hc
Ares
nb
fd,N
fd,S
b,adm
b,adm
Saddle height parameter (Fig. 2a) which indicates the vessel axis height
respect to the ground anchorage point
Saddle width on the vessel axis (Fig. 2c)
Transversal saddle width (Fig. 2c)
Base parameter (Fig. 2a)
Resistance area of the bolt
Number of bolts in the base plate connection
Normal design stress
Tangential design stress
Normal admissible stress
Tangential admissible stress
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
hw0
hw
hmin
hwet
vw
w
kw
tr
Actual depth of ood water over the ground level

Effective ooding depth given by hw hw0 - hc
hmin l2 D=2 minimum ooding height able to wet the vessel surface
Wetting height, representing the height of the vessel wetted by ood
Velocity of ood water
Density of ood water
Flood hydrodynamic coefcient
Return period
3.9
Expected frequency of the ooding scenario
3.10
Geometrical parameter dened as follows: min [hmin; D]
3.11
Vessel submerged fraction:

h
i
p
2
min V Lext 8 D2 D4 arcsin 2D D 2 2 D 2D 2 ; 1
4.1
Fn
4.2
4.3
4.4
Fn,b
Mb
4.5
Fx,i
4.6
Ntot,i
Xi
4.7
Aext
4.8
4.9
4.10
4.11
Fv
Fv,b
Mt
yi
4.12
Fy,i
4.13
4.14
4.15
Stot,i
i
i
Vessel external volume (assuming spherical edges): V ext D4 L D D6

Vessel internal volume (assuming spherical edges):
2
V int D 42t L D D 62t

Vessel tare weight. If no data are available for the vessel under analysis this
simplied evaluation can be used Wt s (Vext-Vint)
Buoyancy due to the ooding (Fig. 2b):

Fn Fh Ft V ext w g g W t V int l 1 V int v
Portion of the normal force on each bolt: F n;b F n =nb
Bending moment due to lift action (Fig. 2b): M b 12 F n l3
Distance from the neutral axis on which Mb is applied
Resultant normal force on each bolt due to the bending moment:
2
b
F x;i xi U M b =ni
1 xi
Overall normal force acting on each single bolt: N tot;i F n;b F x;i
External surface of the vessel wetted by the ooding (Fig. 2b):

p
2
Aext L 2l1 8 D2 D4 arcsin 2D D 2 2 D 2D 2
Drag force due to ooding (Fig. 2b): Fv 12 w kw v2w Aext
Portion of the drag force on each bolt: F v;b F v =nb
Torque due to drag force (Fig. 2b):M t 12 F v l3
Distance from the neutral axis on which Mb is applied
Resultant shear force on each bolt due to the torque action:
2
b
F y;i yi U M t =ni
1 yi
Overall shear force acting on each single bolt: Stot;i F v;b F y;i
Overall normal stress acting on the i-th bolt: i N tot;i =Ares
Overall shear stress acting on the i-th bolt: i Stot;i =Ares
Units
Reference
[33]
[3133]
[3133]
[3133]
[33]
[33]
min max
0.01
0.9
300
CALC
Pa
5001500
kg/m3
[7]
CALCa
kg/m3
[54]
7800
kg/m3
[31]
1000
kg/m3
CALC
m3
CALC
m3
Given or
CALC
kg
DB
DB
DB
0.3b
157 c
410
560 c
396 c
373 c
264 c
m
m
m
mm2
MPa
MPa
MPa
MPa
04d
CALC
CALC
CALC
03.5d
1100
1.8
10500
10 3
10 2
CALC
m
m
m
m
m/s
kg/m3
CALC
y1
m
CALC
CALC
CALC
N
Nm
m
CALC
CALC
CALC
m2
CALC
CALC
CALC
N
N
Nm
m
CALC
CALC
CALC
CALC
N
MPa
MPa
[33]
[33]
[33]
[4042]
[33]
[4042]
[4042]
[4042]
[4042]
[4851]
[4851]
[36,37]
[36,37]
[52]
[52]
[55]
[55]
131
Table 2 (continued )
Operation
ID
Item
4.16
FC1
4.17
(5) Evaluation of vessel resistance with

correlations
(6) Evaluation of parameters

implemented in QRA studies
FC2
4.18
CFL
4.19
vw,c
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
CFL
vw,c
6.1a
6.1b
6.2
fLOC
Description/Denition
Value
Failure criterion FC1 for the i-th bolt:

Failure criterion FC2 for the i-th bolt:

2
i
b;adm
41
2 2
i
f
41
i
b;adm
i
f d;N
2
Units
Reference
[38,39]
[38,39]
CALC
CALC
m/s
CALC
CALC
CALC
CALC
CALC
CALC
CALC
CALC
m/s
CALC
d;S
Critical lling level for a given storage system and assigned storage conditions.
Repeat steps 4.1 to 4.18 in order to nd the minimum which allows satisfying
one of the two failure criteria (FC1 or FC2)
Critical velocity related to a specic vessel. Given an assigned ooding height
hw, repeat steps 4.8 to 4.18 in order to nd the minimum vw value which
allows satisfying one of the two failure criteria (FC1 or FC2)
First CFL correlation coefcient evaluated for reference substance: A K 1 Da
where K1, a are shown in Table 6
Second CFL correlation coefcient evaluated for reference substance:
B K 2 W t K 3 b where K2, K3, b are shown in Table 6:
Modied A coefcient considering a generic stored substance:

A' ref A = l v
Modied B coefcient considering a generic stored substance:

B' ref B v = l v
vw,c correlation factor: E K 4 Lc where K4, c are shown in Table 6

vw,c correlation exponent: F K 5 ln L=D K 6 where K5, K6 are shown in
Table 6
Critical lling level evaluated with correlations:CFL A'hw B'
Critical velocity evaluated with correlations: vw;c Ehw hmin F
Vessel failure probability evaluation based on critical velocity (see Fig. 7): 1
if vw Z vw,c
If vw ovw,c, vessel failure probability evaluation based on CFL (step 4.19 or 5.7,

see Fig. 7): CFL min = max min
Expected frequency of loss of containment (LOC) f LOC f
CALC
CALC
y1
Saturation pressure data were derived from [54].

Assumed for the case study in Section 3.4
c
Value selected for the present study. Other possible values may be inserted by the user according to Tables 3 and 4.
d
Range assumed for the preparation of failure charts reported in Appendix (Figs. A1A4).
b
Fig. 2b). The torque (Mt) is schematized in Fig. 2c and may be

evaluated as follows:
M t 12 F v l3
10
It is worth to notice that Eq. (9) may lead to conservative

results, since the friction associated to the slipping of the unanchored saddle may reduce the action of the net drag force on the
bolts of the anchored saddle.
The torque action of Mt is distributed over the i-th bolt
according to its distance from the center of the plate (yi) as
follows:
F y;i
Mt
nb
yi
11
y2i
i1
where yi is the distance of the i-th bolt from the base plate center
(see Fig. 2c).
Finally, the total shear force on the i-th bolt (Stot,i) is evaluated
summing the contribution of the shear force (Eq. (9)) and of the
torque (Eq. (11)):
Stot;i F v;b F y;i
12
The procedure reported in a reference technical standard for

bolt connection integrity verication [33,38,39] was adopted in
the present study in order to assess the failure conditions of the
vessel anchorage caused by ooding. The procedure is based on
the calculation of the total normal and shear forces acting on each
bolt (Ntot,i and Stot,i, calculated by Eqs. (8) and (12)) as a function of
assumed ood intensity parameters (water velocity, vw, and water
effective depth, hw) and of vessel geometry. The failure of the
connection is assumed if at least one of the following criteria is

veried for at least one of the nb bolts of the base plate (see
Fig. 2c):

2
2
i
i
41
13a
b;adm
b;adm

f d;N
2
i
f d;S
2
41
13b
where: b,adm and b,adm are respectively the normal and tangential
admissible stress; fd,N and fd,S are respectively the normal and
tangential design stress; i is the average normal stress and i is the
average shear stress on the i-th bolt derived as follows from the
loading conditions:
i N tot;i =Ares
14
i Stot;i =Ares
15
in which Ares is the resistant area of each bolt. Conditions

expressed by Eq. (13a) are derived from the application of the
maximum allowable stress criterion [38,39], while Eq. (13b)
represents the limit state criterion [38,39].
Clearly enough, the above conditions may be applied to any
bolt connection provided that specic data are available. In the
present study, a set of reference data were assumed to obtain
representative failure conditions for the vessel geometries considered in the vessel database summarized in Table 1. The features
assumed for the bolt connections considered are summarized in
Tables 3 and 4, and were derived from technical standards [40
43]. It is worth mentioning that bolt size may vary according to the
size of the saddle base plate, which in turn depends on the tank
geometry. Hence the Ares values reported in the Tables 3 and 4
132
Table 3
Standard bolt connection materials classes [4042] evidencing the materials of
interest for the present application (marked with X) [33]. For symbols denition,
see Table 2.
ID Class
type
Industrial
application
fd,N
(Mpa)
fd,S
(Mpa)
b,adm
(Mpa)
b,adm
(Mpa)
A
4.6
Ba 5.6
C
6.6
D
8.8
E 10.9
X
X
X
X
240
300
360
560
700
170
212
255
396
493
160
200
240
373
467
113
141
170
264
330
hmin l2 D=2
a
Selected for the development of mechanical model and correlations (see
Table 2).
Table 4
Standard bolt dimensions range [4042] evidencing the dimensions of interest for
the present application (marked with X) [33]. For symbols denition, see Table 2.
ID Industrial
application
Bolt
Diameter
(mm)
Ares
(mm2)
A
B
C
D
Ea
F
Ga
Ha
I
12
14
16
18
20
22
24
27
30
84
115
157
192
245
303
353
459
561
X
X
X
X
X
X
Typical diameter range (m) of

supported tanks [33]
o 1.2
1.22.4
42.4
were obtained from vessel diameter on the basis of literature

correlations [33].
2.3. Characterization of ood impact vector (step 3)
The elements needed for the characterization of the ood
impact vector are the ooding frequency and the expected ood
severity. The standard parameter for ood frequency evaluation is
the return period (tr) measured in years and given by hydrological
studies [4446], usually available from local competent authorities
[29,30,47,48]. The ooding frequency f can thus be estimated as
follows:
16
Since there are different types of ood events (e.g. oodplain

inundations with high water level, ash oods with high water
velocity, etc.), the possible modalities of ood impact (slow
submersion, moderate speed wave, high speed wave) must be
discriminated. As shown in Section 2.2, the ood severity can be
quantied by two parameters: water effective depth (hw) and
water speed (vw). The effective depth should take into account the
possible effect of protection measures, such as concrete supports
higher than the ground level to which the vessel saddles are xed.
Taking into account the schematization in Fig. 2a, if the height of
the supports (hc) is considered, the effective ood water height hw
may be calculated as follows:
hw hw0 hc
18
Hence, hmin is the minimum ooding height needed to wet the

surface of the horizontal vessel.
Finally, in the evaluation of vessel damage due to ood impact,
a further parameter is introduced in order to estimate the height
of the vessel effectively wetted by ooding, namely the ooding
wetting height:
hwet hw hmin
19
This latter parameter, representing the effective water depth, is

particularly signicant for the evaluation of vessels failure due
ood impact, as explained in the following sections.
2.4. Model validation (step 4)
a
Selected for the development of mechanical model and correlations (see
Table 2).
f 1=t r
values were derived analyzing the features of critical ooding

events [49,50], in order to obtain a worst case reference. Besides, a
minimum value of ood height is also introduced (namely, hmin),
that is dened as the minimum possible ooding height affecting
vessels mounted on saddles. This parameter depends on the type
of vessel and may be derived as follows (see Fig. 2a):
17
where hw0 is the actual depth of ood water. Clearly enough hw is

equal to hw0 if no protections are available.
On the basis of available data on past events, reasonable ranges
for credible values of water height and water speed recorded in
ood events were collected and are reported in Table 2. The higher
Past accident data analysis on NaTech scenarios involving

ooding of process equipment are scarce and not detailed [6
9,11]. In the accident description, the ood parameters (water
height and speed) needed to support model validation are usually
not mentioned or only provided in qualitative terms (e.g.: high
depth ood, low speed ood, severe ood, etc.). Nevertheless,
a signicant accident occurred in a park of propane tanks during
the 1993 Mississippi River basin oods close to Des Peres River,
Illinois. Fifty tanks, with nominal capacity of 30,000 gal (e.g., about
120 m3), supported by saddles, began to oat under the action of
the ood water [34,35]. The water speed is not reported, but a
water height hw of 2.4 m could be estimated [34]. Only small and
localized ash res occurred, with no catastrophic ruptures of
vessels. A number of 12,000 residents was forced to leave their
homes for twelve days [34,35].
The available accident data were taken into account for a
preliminary model validation. Input data used in the calculations
are summarized in Table 2. The typical vessel geometry adopted for
30,000 gal tanks is included in the database reported in Table 1.
However, since the design pressure of the ooded tanks is unknown,
three tank geometries corresponding to tanks having the same
diameter and length, but increasing thickness were considered:
#46 with design pressure 1.5 MPa; #70 with design pressure
2.0 MPa; #94 with design pressure 2.5 MPa (see Table 1 for further
details). The procedure described in Section 2.2 and summarized in
Table 2 was applied to assess possible vessel failure. The tanks were
considered conservatively as containing propane up to their maximum lling level (thus assuming the minimum buoyancy according
to Eq. (2)) and assuming a water depth of 2.4 m70.25 m (considering possible uncertainties in the actual reported oodwater height).
The model evidenced that connection failure conditions for the three
tank geometries assumed were veried independently of water
velocity vw. Thus, model results are in agreement with past accident
evidence, pointing out the criticalities connected with high-depth
ooding of storage vessels.
In order to extend model validation, generic threshold data on
damage caused to equipment items by oods derived in previous
studies [8,29,30] were considered. From available data, a ood
wave with a velocity of 2 m/s and a maximum height of 0.5 m
were considered as thresholds below which damages are not
expected for generic structures. Fig. 3 shows the fraction of vessels
contained in the database which fail under these reference ooding conditions. In order to consider reference conditions suitable
for both atmospheric and pressurized equipment, water was

assumed as the reference substance inside vessels and air was
assumed in the top space of the vessels.
As shown in the gure, most of the vessels do not fail even
when empty (failure fraction for empty vessels is around 15%).
When different lling levels are considered, the fraction of vessels
which fails due to the ood impact decreases to 5%. Large tanks (i.
e. having volumes higher than 100 m3) are not able to withstand
the high water speed conditions even considering a 100% lling
level, due to the high value of l3 (1215 m), which dramatically
increases the torque action on the support. However, it is worth
mentioning that the model only considers the worst possible
orientation of the vessel respect to the ood wave (e.g., the one
which maximizes the Aext in Eq. (4)), thus obtaining the maximum
value of drag force. Hence the results should be considered
conservative, thus evidencing a sufcient agreement with the
available literature data.
133
of lling levels between 1% and 90% was thus considered, as

shown in Table 2. Table 5 reports the reference values assumed for
the liquid and top space vapor density. Liquid density values are
based on available data concerning substances released in past
accidents [7]. The density of the vapor was calculated assuming
nitrogen blanketing in the top space of atmospheric vessels
and the vapor at saturation pressure for liqueed pressurized
storages. In both cases, Equation (1.13) in Table 2 was used for the
estimation of vapor density.
Figs. A1A4 in the Appendix show the failure conditions
obtained considering all the vessels in the database (see Table 1),
different stored uids and different operating conditions. Figs. A1
and A2 report the data obtained for atmospheric vessels, while
Figs. A3 and A4 report the data obtained for the pressurized
vessels considered in the database. Fig. 4 reports some examples of
failure plots derived from the failure tables reported in the
Appendix. The failure plots synthetically represent the results
obtained from the application of the model, and will be discussed
in detail in Section 3.
2.5. Dataset of failure conditions (step 5)

The vessel database reported in Table 1 was used to obtain a
dataset of failure conditions considering the set of reference
ooding conditions identied in step 3 (Section 2.3). A value of
1% of the reference lower bound, min, was assumed as the
minimum credible operational limit for the lling level. A range
% Failed tanks
16%
12%
8%
4%
0%
0%
5%
10%
15%
20%
Filling level (%)

Fig. 3. Results of the extended validation of the mechanical model: fraction of the
tanks failed as a consequence of the reference ood (hw 0.5m; vw 2 m/s) as a
function of vessel lling level. Stored uid density considered for validation is that
of water (l 1000 kg/m3), and air is considered in the top space of the vessel
(v 1.2 kg/m3). The evaluation was carried out for all the vessels reported in the
database (96 vessels, see Table 1).
2.6. Simplied correlations for vessel damage (step 6)

The analysis of the dataset of failure conditions obtained in step
5 allowed the identication of the critical parameters leading to
vessel failure. The plots reported in Fig. 4 highlight that the stored
uid has a strong impact on the failure region of the vessels. An
increase in the lling level and/or a higher density of the stored uid
result in an increased resistance of the vessel to the action of
buoyancy. Therefore, once the storage system is dened (i.e. dening
the geometry of the vessel and the substance stored) the lling level
is the only operating parameter which affects the vessel resistance
to buoyancy caused by a given set of ooding conditions. The critical
lling level (CFL) of a vessel may thus be dened as the minimum
value of able to ensure the tank resistance to buoyancy caused by a
ood having a given intensity [8]. This parameter can be evaluated
using the failure plots reported in Fig. 4, based on the failure model
described in Section 2.2.
Nevertheless, simplied correlations are also provided in the
following to allow a simplied straightforward evaluation of vessel
failure probability. The correlations are based on the analysis
of the CFL behavior respect to the water effective height (hw)
Table 5
Reference substances, associated to the correspondent type of vessel, and ood conditions considered in the present analysis. PRES pressurized vessel; ATM atmospheric
vessel.
Parameter
ID
Reference
substances
S1
S2
S3
S4
S5
Reference
ooding
conditions
S6
W1
W2
W3
W4
Reference parameters
Description
Liquid density
l (kg/m3)
Vapor density Typical type of

v (kg/m3)a
storage vessel
Flood water actual Expected occurrence

Flood water
frequency f (y 1)
speed vw (m/s) height hw0 (m)
LPG
Ammonia
Gasoline or diesel fuel
Liquid aromatics or
hydrocarbons
Water solutions with
contaminant(s)
Chlorine
High depth ooding
condition
High speed ooding
conditions
Intermediate severity
ooding conditions
Low severity ooding
conditions
500
600
750
900
20.7
7.8
1.2
1.2
PRES
PRES
ATM
ATM
1100
1.2
ATM
1400
1100
24.4
PRES
0.5
2.0 10 3
1100
0.5
2.0 10 3
1100
5.0 10 3
1100
0.5
0.5
3.3 10 2
See Table 2 for details on v evaluation.
134
5% S1
5% S6
50% S1
50% S6
3.5
3
Flooding velocity vw (m/s)
2.5
ID #5
V=10 m3
1.5
1
0.5
0
5% S1
5% S6
50% S1
50% S6
3.5
3
2.5
ID #20
V=100 m3
1.5
1
0.5
0
0.5
1.5
2.5
0.5
Wetting height (hwet) (m)
1.5
2.5
4
5% S1
5% S6
50% S1
50% S6
3.5
3
2.5
ID #53
V=10 m 3
1.5
1
0.5
0
5% S1
5% S6
50% S1
50% S6
3.5
3
2.5
ID #68
V=100 m3
1.5
1
0.5
0
0.5
1.5
2.5
0.5
1.5
2.5
Fig. 4. Example of failure plot for: (a) vessel #5; (b) vessel #20; (c) vessel #53; (d) vessel#68. Vessel data are reported in Table 1 at the correspondent ID number. Failure
plots were obtained considering two different lling levels (5 and 50%) and two different reference substances (S1,S6, see Table 5).
assuming a reference value, ref, for the density of the stored

substance. The effect of water speed is not taken into account to
assess the CFL, since the CFL signicantly affects only the resistance to buoyancy and has a limited inuence on the resistance to
the action of ood water drag force.
As shown by the examples reported in Fig. 5a, a linear empirical
correlation may be used to relate the CFL to water height, given
the vessel geometry and the stored uid density:
CFL Ahw B
20
where the parameters A and B are only a function of the vessel

geometry (hence, operating pressure and volume).
The data in Fig. 5a were obtained for the sample vessels
considered for Fig. 4 assuming a reference uid density (ref) of
the stored uid equal to 1000 kg/m3. Similar data were obtained
for all the vessels in the database (see Table 1), and are not
reported for the sake of brevity. The extended application of the
mechanical model allowed the calculation of the values of the A
and B parameters as a function of the geometrical features of the
vessel. Figs. 5b and 5c show an example of the behavior of the A
and B with respect to vessel diameter (D) and vessel tare
weight (Wt).
The data reported in the gures were obtained for vessels

having a specic design pressure (1.5 MPa). However, qualitatively
similar results were obtained for the other types of vessels. It is
worth mentioning that the CFL has a maximum value related to
the operating capacity of the vessels (i.e. the maximum CFL value
is equal to max). Thus, the following empirical correlations were
obtained for the A and B parameters with respect to vessel
features:
A K 1 Da
21
B K 2 W t K 3 b
22
The values calculated for the K1, K2, K3, a, and b parameters are
reported in Table 6 for each category of vessel considered. Two
sets of parameters were calculated:
SET A: best t parameters (dashed lines in Fig. 5);

SET B: envelope parameters allowing a conservative estimation
of CFL (solid lines in Fig. 5).
Clearly enough, the above set of parameters depends on the
value assumed for the reference density of the stored uid, ref. In
order to take into account the actual density of the stored uid, the
135
following changes may be introduced in Eq.(20):

CFL A'hw B'
23
The values of the A' and B' coefcients in Eq. (23) may be
calculated from the A and B parameters obtained using the
reference uid density ref:
A'
ref A
l v
24
B'
ref B v
l v
25
where l is the actual density of the stored liquid and v is the

density of the vapor phase inside the vessel.
The approach discussed above allows the calculation of a
critical lling level below which the vessel may fail due to
buoyancy. In order to account also for the action of the drag force,
the simplied model needs to be extended. Flood waves with high
water velocity, vw, may lead to vessel failure due to drag force even
in the case of limited water depth (see Section 2.4). Hence a critical
water velocity, vw,c, was dened as the vw value able to damage a
given vessel for an assigned value of wetting height (hwet see Eq.
(19)). As a matter of fact, in case of a food wave with a small hwet
value, thus unable to cause damages by buoyancy, vw,c represents
the critical ood velocity value which causes the minimum drag
force value required to damage the vessel.
Also in this case this parameter may be derived applying the
mechanical model, and simplied correlations based on hwet (see
Fig. 4) were obtained from failure plot analysis:
vw;c Ehwet
26
Fig. 5d and e show the behavior of E and F calculated for the

sample vessels in Fig. 4. Similar trends were obtained for all the
vessels in the database. Thus, also in this case it was possible to
obtain empirical correlations for the E and F parameters with
respect to vessel geometry:
E K 4 Lc
27

L
F K 5 ln
K6
D
28
where L is the vessel length and D is the vessel diameter. Again, for
the c, K4, K5, and K6 parameters a set of best t (SET A, dashed lines
in Fig.5) and of envelope (SET B, solid line in Fig.5) values were
calculated and are reported in Table 6.
3. Results and discussion

3.1. Analysis of vessel failure conditions
Fig. 5. Development of simplied correlations from the mechanical

model: assessment of (a) critical lling level (CFL); (b) coefcient A in Eq. (21),
(c) coefcient B in Eq. (22), (d) coefcient E in Eq. (27), and (e) coefcient F in
Eq. (28).
The application of the mechanical model developed in the

present study to the reference ooding conditions summarized in
Table 2 and to all the vessels considered in the database (Table 1)
allowed obtaining the failure charts reported in the Appendix (Figs.
A1 and A2 for atmospheric vessels, Figs. A3 and A4 for pressurized
vessels). The failure charts give an overview of reference conditions
leading to vessel failure in ood events, considering a large set of
vessels, several different ood conditions and assuming xed lling
level and uid density, thus considering the same operating
conditions for all the vessels. The failure charts may be used to
carry out a direct and straightforward preliminary assessment of
the credibility of vessel failure in different ooding conditions.
136
Table 6
Parameters for CFL and vw,c evaluation applying the simplied correlations ((Eqs. (20)-28)). SET A: best t parameters; SET B: envelope correlation parameters. Pd design
pressure (MPa); ATMatmospheric pressure.
Correlation type
SET A
SET B
Vessel type
K1
K2
K3
ATM
Pd 1.5
Pd 2.0
Pd 2.5
ATM
Pd 1.5
Pd 2.0
Pd 2.5
1.331
1.287
1.290
1.256
1.331
1.347
1.341
1.355
2.163
1.144
1.305
6.068
1.882
1.197
1.365
4.512
288.6
499.2
546.0
234.0
46.8
475.8
483.6
234.0
More detailed data may be obtained by the use of failure plots,

that require the application of the mechanical model to the actual
data of the vessel of interest. Fig. 4 shows an example of the failure
plots obtained for four representative vessels: two atmospheric
tanks and two pressurized tanks. For each design pressure, two
reference volumes were considered: 10 m3, representative of
medium scale vessels, and 100 m3, representative of large scale
storage. Table 1 reports the data of the vessels considered (see IDs
#5, #20, #53 and #68). The failure plots show the combinations of
ooding parameters leading to vessel failure due to the failure of
the support bolt connection. Two density values were assumed for
the stored liquid (i.e. the lowest and the highest values considered
credible in the present study, respectively associated to substances
S1 and S6 in Table 5). The effect of two different lling levels
(respectively 5% and 50%) was assessed. The plots in Fig. 4 report
the minimum value of water velocity, vw, able to cause the failure
of the connection at a given a value of wetting height (hwet). A
peculiar behavior of the curve is obtained, since when ood water
height exceeds a critical value the connection is predicted to fail
even in the case of low or negligible values of water speed. This
corresponds to a critical value of buoyancy (Fn in Eq. (1)), which,
according the failure criterion assumed (Eqs. (13a,b)), leads to the
rupture of the bolt connection even in the absence of drag force
(Fv). Therefore, for a given vessel type and substance, the critical
lling level (CFL, see Section 2.6) may be obtained from the failure
plots in Fig. 4 as a function of ooding effective height (hw). The
results are reported in Fig. 5a.
The failure plots allow understanding the key parameters
affecting the failure of the vessel related to the type of storage
system. In fact, the failure zone of low volume tanks (Fig. 4a and c)
is more extended with respect to that of larger tanks (Fig. 4b and d),
thus indicating a higher vulnerability. This is due to the lower
overall vessel weight (both considering inner uid and vessel shell),
which reduces the resistance to the hydrostatic lift force. This is
conrmed by the signicant change in the failure zone of a tank of
given volume if its operative pressure, and thus its shell thickness, is
increased (see the comparison of Fig. 4a and b with Fig. 4c and d).
For the same reason, the increase of the lling level and of the
stored uid density have a signicant effect on the vessel resistance.
Clearly enough, the effect of the increasing uid density is more
pronounced at higher values of (solid lines in Fig. 4), due to the
growing importance of this parameter at higher lling levels.
3.2. Sensitivity and uncertainty analysis
The failure plots in Fig. 4 were obtained assuming the values of
several parameters (see Table 2) related to construction materials
and bolt connection features on the basis of technical standards.
However, some variability may exist in these key-parameters.
Thus, a sensitivity analysis was undertaken to understand the
inuence of the geometry and mechanical properties of the
connection on model results. Tables 3 and 4 respectively
K4
K5
K6
9.910
0.037
0.399
3.195
0.037
0.399
0.990
0.952
0.966
0.951
0.990
0.995
0.976
0.999
0.260
0.112
0.109
0.263
0.252
0.129
0.120
0.239
0.718
0.341
summarize the standard range of bolt material mechanical properties (i.e. the resistance class) and of geometrical parameters. Both
tables also report the typical values of these parameters in general
industrial practice for the anchorage of process and storage
vessels.
The sensitivity analysis was carried out applying the mechanical model, considering all the possible combinations of bolts
geometry and construction material class. Moreover, each combination was modeled considering different numbers of bolts (ranging from 4 to 10). For the sake of simplicity, as for the extended
model validation (see Section 2.4) air was assumed in the top
space of the vessels, water was assumed as the stored liquid and a
xed lling level (e.g., 10%) was considered. Fig. 6a shows the
fraction of failed vessels in the database (Table 1) varying the bolt
connection features imposing in the model the reference ooding
condition used for model validation (see Section 2.4). Since this
condition represents a threshold limit below which no failure is
expected, as mentioned in Section 2.4, no change is expected in
results, even if bolt connection parameters are changed.
Actually, the results show a signicant variation (an increase of
the fraction of failed vessels) only when low resistance connections
are considered (e.g., bolt class lower than B and Ares o200 mm2).
However, such types of connections are usually not applied in
industrial facilities for vessel anchorage [33,41]. In Fig. 6a, the
connection material type selected for the present study is highlighted by an arrow. It is evident from the gure that the results
obtained are the same when higher resistance materials with failure
fractions lower than 25% for any type of bolt with Ares 4200 mm2
are used.
In order to evaluate model uncertainties, the statistical evaluation of model predictions was carried out determining the distribution of vessels failure fraction values considering the
reference ooding conditions (see Section 2.4). All the possible
input values were implemented in the model, thus the full range
of bolt connection materials, bolts geometries and bolts number
were used. Limited input values were then considered, only
selecting the values of the parameters commonly applied in
industrial facilities according to Tables 3 and 4 [33]. In the latter
case, only 4, 8 and 10 bolts were considered for the base
connections. Fig. 6b shows the distribution of vessel failure
fractions obtained applying both the model with all possible
combinations of input parameters and ltering the input parameters according to Tables 3 and 4. The results shown in Fig. 6b
were then interpolated considering a normal distribution and
obtaining the cumulative distribution functions (CDFs) shown
respectively in Fig. 6c and d for full range and ltered input
values. As shown in Fig. 6c, if all possible input values are
considered, values of failure fraction higher than 60% are obtained,
since the weaker connections lead to the failure even in presence
of the low severity ooding. However, if only the relevant
connections types are considered, the failure fraction is lower
than 50% even in the case of weaker connections. The third
137
Number of observations
50
All possible input values
40
"Filtered" input values
30
20
10
0
0%
10% 20% 30% 40% 50% 60% 70% 80%
Failure Fraction
CDF
1
Data
Cumulative Normal Distribution
CDF
1
0.75
0.75
0.5
0.5
0.25
0.25
0
0%
20%
40%
60%
80%
100%
0
0%
Data
20%
Failure Fraction
Cumulative Normal Distribution
40%
60%
80%
100%
Failure Fraction
Fig. 6. Results of the sensitivity and uncertainty analysis: (a) fraction of the tanks failed as a consequence of the validation reference ood (hw 0.5 m; vw 2 m/s)
considering different bolt connection materials (resistance class AE) and bolt sizes (Ares in mm2); (b) distribution of failure fraction values obtained applying the model with
different input types for all the vessels reported in the database (96 vessels, see Table 1); cumulated normal distribution of failure fraction values obtained applying the
model with all possible input values (c) and limiting input data (d). Density assumed for the stored uid is that of water (l 1000 kg/m3), and air is considered in the top
space of the vessel (v 1.2 kg/m3). The dashed box in panels (c) and (d) represents the third quartile (Q3) of model predictions.
quartile (Q3) of the vessel failure distribution (highlighted by the

dashed box in Fig. 6c and d) was taken into account in order to
quantitatively evaluate the robustness of the model. Q3 of failure
fractions is reduced from 36% to 22% when input data are ltered,
thus demonstrating the limited variability of model results when
realistic input values derived from current industrial practice are
applied.
It may be concluded that only few critical parameters are
determinant to assess the vessel resistance to a given ood
scenario (identied by a given ooding height and velocity
combination). In particular, given a vessel geometry, the key
parameters to assess vessel resistance to ood impact are the
uid density and the vessel lling level. Clearly enough, these are
the operating parameters of the storage system. These results are
in accordance with the outcomes of a previous study concerning
atmospheric vertical vessels [8].
3.3. Assessment of vessel damage probability
The results discussed above evidence that two key parameters
inuence vessel failure in a given ooding scenario: the vessel
lling level and the ood water velocity. Hence, in order
to estimate the failure probability of a vessel due to ood impact
(i.e. the vessel vulnerability to a ood), two threshold parameters
may be used as a reference: the critical water velocity, vw,c (see
Section 2.6), and the critical lling level, CFL (that may be
calculated by Eq. (4.18) in Table 2b, see Section 2.6). The rst
represents a threshold condition for velocity over which the drag
force generated by ood water is sufcient to cause the failure of
the bolt connection for a given ooding height. Thus, connection
rupture is predicted independently on vessel parameters other
from vessel shape and volume (e.g. vessel design pressure, stored
uid density or lling level have a negligible inuence, if any, on

the drag force caused by ood water). The second threshold
parameter identied, the CFL, represents the minimum value of
the lling level, , able to ensure the resistance of the bolted
connection to buoyancy. Thus, a specic approach, schematized in
Fig. 7, was developed in order to assess vessel failure probability
(). The procedure for the assessment of is outlined in the
following, while the detailed steps needed to carry out the specic
calculations required for the assessment of vessels damage probability are summarized in Table 2.
Given only the vessel geometry and the characteristic of the
ooding, the critical water velocity, vw,c, may be directly evaluated
and compared with the actual velocity of oodwater, vw, considered in the analysis. This may be done either with the detailed
model or with the simplied correlations. If vw is equal or greater
than vw,c the upper-bound failure condition due torque action
caused by the drag force is fullled. Thus a unit value of vessel
failure probability may assumed:
1 if vw Z vw;c
29
If oodwater velocity if lower than the critical threshold for

direct failure due to torque (vw ovw,c), also buoyancy should be
taken into account. According to Fig. 7, it is assumed that the CFL
delimitates the safe operating conditions given the features of
the storage system for a given ooding scenario. If the lling level
is lower than the CFL, the tank is in the unsafe zone since the
vessel support may not resist the ood water impact.
The lling level value may be determined on the basis of site
specic historical data or by implementing statistical distributions
which might be available for the site under analysis (as specied in
Fig. 7). Thus, after having built a cumulative distribution function
for the values of the lling level (CDF) the failure probability
138
may be dened as follows:

CDF CFL
This approach was applied to the evaluation of vertical atmospheric tanks vulnerability to ooding [8]. Either the complete
model discussed in Section 2.2 or the simplied correlations may
be applied to calculate the value of the CFL in Eq. (31).
Fig. 8a and b report parity plots obtained for the comparison of
the results obtained by the complete model and by the two sets of
simplied correlations. The data displayed in the gure were
obtained applying to all the vessels in the database (see Table 1)
a matrix of reference ood conditions (see Table 5). Data were
obtained assuming water as the stored uid (l 1000 kg/m3) and
air in the top space of the vessel (v 1.2 kg/m3). The vessels were
considered anchored (thus hw hw0).
Clearly enough, the simplied correlations allow a rapid and
straightforward assessment of vessel failure probability. On one
30
in which the CDF(CFL) is the value of the cumulative distribution

function evaluated for a lling level equal to the estimated CFL. In
the absence of any specic data for lling level distribution, a
linear distribution of possible lling levels between min (1%)
and max (90%) was assumed in the present study for the
calculation of . On the basis of the previous considerations, is
derived as the ratio between the unsafe operative conditions
with respect to all the possible operative conditions:
CFL min
max min
31
Evaluation of vessels
failure probability
Application of mechanical model or simplified correlations
Input data
=1
Yes
Flooding
conditions
Flood velocity
greater than
vw,c?
Vessel
geometry
Critical Filling Level (CFL) Calculation

SAFE
No
Filling level
distribution
calculated
by Eq. (30)
0.9
CFL
0.01
UNSAFE
Operating
conditions
Operative limits
Fig. 7. Schematization of the procedure for the determination of vessels failure probability as a consequence of a given ooding scenario.
100%
Correlation SET B
Correlation SET A
100%
80%
60%
40%
W1
W2
W3
W4
20%
0%
0%
20%
40% 60%
Model
60%
40%
W1
W2
W3
W4
20%
0%
0%
80% 100%
Overpredictions
80%
20%
40% 60%
Model
80% 100%
Underpredictions
Geometric Variance
(VG)
1000
100
10
1
0.1
0.5
10
Geometric Mean Bias (MG)

Fig. 8. Comparison among the prediction of vessel vulnerability (%) carried out by the deterministic model and correlation SET A (a), SET B (b); geometric mean bias (MG)
and geometric variance (VG) for correlations compared to deterministic model (c). Reference ooding conditions are reported in Table 5. Tank geometrical data are listed in
Table 1. Density assumed for the stored uid is that of water (l 1000 kg/m3), and air is considered in the top space of the vessel (v 1.2 kg/m3).
hand, as evident from Fig. 8a, the best t correlations (SET A),
provide results which are in good agreement with the detailed
model even if some of the prediction are not on the safe side (i.e.
failure probability is underestimated especially when values are
near to 1). On the other hand, Fig. 8b shows that the use of the
envelope correlations (SET B) allows for less accurate but always
conservative predictions of vessel failure probability.
The method proposed by Hanna et al. [51] was used to analyze
the performance of correlations in predicting the failure probability and to compare them with respect to the detailed model.
The method is based on the calculation of the geometric mean bias
(MG) and the geometric variance (VG) of the values of failure
probability predicted by the detailed model (mod) and by the
simplied correlations (corr):
"
#
mod
MG expln mod ln corr exp ln
corr
32
139
VG expln mod ln corr 2 exp
"
2 #
mod
ln
corr
33
The over-bars indicate that an average was performed over the

entire data set. Good model performances are achieved when both
MG and VG are close to unity. In order allow a systematic
performance assessment, VG values may be plotted versus the
corresponding MG values for each data set and may be compared
to the following reference curve:
lnV G lnMG2
34
As clearly appears from Eqs. (32) and (33), Eq. (34) represents
the relationship between VG and MG values in a correlation
having only a mean bias with respect to the detailed model results
(that is, a correlation for which the ratio of mod /corr is nearly
20m
River
V5-V10
V17-V19
G = GROUP OF VESSELS
G1 = V1-V4
G2 = V5-V10
G3 = V11-V16
G4 = V17-V19
V24-V26
G5 = V20-V23
G6 = V24-V26
V31
G7 = V27-V30
G8 = V31
V32-V35
V36-V39
G9 = V32-V35
G10 = V36-V39
Fig. 9. Layout of the facility analyzed in the case study. Tank data are reported in Table 7.
Fig. 10. Example of case study results: vessel failure probability (%) given the ooding conditions reported in Table 5. Panels (a) and (e): ooding W1; panels (b) and (f):
ooding W2; panels (c) and (g): ooding W3; panels (d) and (h): ooding W4. Panels (a) to (d): vessels anchored to the ground. Panels (e) to (h): vessels on a concrete base
(hc 0.3 m). Ggroup of vessels dened in Fig. 9.
2.10 10 3
NF
3.13 10 3
NF
NF
1.56 10 3
NF
2.10 10 3
NF
NF
1.68 10 3
NF
2.20 10 3
NF
NF
7.67 10 4
NF
1.68 10 3
NF
NF
NF
NF
1.16 10 3
NF
NF
6.21 10 4
NF
NF
NF
NF
1.67 10 3
1.84 10 3
2.69 10 3
9.24 10 4
3.11 10 4
2.03 10 3
3.17 10 4
1.67 10 3
1.07 10 5
3.11 10 4
1.60 10 3
8.57 10 4
2.54 10 3
6.70 10 4
8.13 10 5
1.91 10 3
1.07 10 4
1.60 10 3
NF
8.03 10 5
3.23 10 4
NF
2.43 10 3
8.66 10 4
5.51 10 5
1.86 10 3
NF
3.23 10 4
8.41 10 4
5.51 10 5
18
7.7
7.7
13.2
18
13.2
18
18
13.7
13.7
18
13
5
18
24
6
30
18
6
27
1.5
2.5
ATM
2.0
2.0
ATM
2.5
1.5
ATM
2.0
S2
S6
S5
S1
S1
S3
S1
S1
S4
S1
1.75 10 3
2.00 10 3
2.00 10 3
1.83 10 3
1.23 10 3
2.00 10 3
1.22 10 3
1.75 10 3
9.56 10 4
1.23 10 3
NF
NF
7.04 10 5
NF
NF
3.76 10 5
NF
NF
NF
NF
1.02 10 4
NF
1.33 10 4
NF
NF
4.65 10 5
NF
1.02 10 4
NF
NF
1.72 10 3
2.00 10 3
2.00 10 3
1.68 10 3
1.10 10 3
2.00 10 3
1.10 10 3
1.72 10 3
8.72 10 4
1.10 10 3
1.23 10 3
1.73 10 3
2.00 10 3
1.81 10 3
1.06 10 3
2.00 10 3
1.05 10 3
1.23 10 3
1.28 10 3
1.06 10 3
2.00 10 3
NF
1.90 10 4
NF
2.00 10 3
9.46 10 5
2.00 10 3
2.00 10 3
NF
NF
SET B
SET A
MOD
SET B
SET A
MOD
SET B
SET B
SET A
MOD
SET A
MOD
LOC frequency fLOC (y 1) W3

2.8
1.2
1.2
2.1
2.8
2.1
2.8
2.8
3.2
3.2
The value of f may be derived from site specic data available

from local authorities, or may be evaluated with specic models
present in the literature [29,30,47,48]. Table 5 reports the values of
f used in the case study, that were based on the actual values
obtained for an Italian site [52]. Tables 7 and 8 show the LOC
frequencies, fLOC, calculated for each tank on the basis of the
vulnerability assessment carried out applying both sets of simplied correlation and the deterministic model developed. The data
were obtained both for the case of vessels anchored to the ground
and for vessels having a concrete support. As evident from the
100
10
10
50
100
50
100
100
100
100
35
V1-V4
V5-V10
V11-V16
V17-V19
V20-V23
V24-V26
V-27-V30
V31
V32-V35
V36-V39
f LOC f
D
L
t
Pd
Stored
(m) (m) (mm) (MPa) Substance
ID
The methodology developed to assess vessel failure probability was

applied to the analysis of an industrial lay-out in order to provide data
on expected vessel failure probabilities, suitable for the use in a
quantitative risk assessment (QRA) framework. Fig. 9 shows the
storage area considered, in which several pressurized or atmospheric
horizontal tanks are present in a ood-prone zone. Two different
assumptions were considered for vessel supports: (i) all vessels
supports are xed on the ground (thus the effective water height hw
corresponds to the actual ood height hw0); (ii) all vessel supports are
xed to a concrete base having a height above ground, hc, equal to
0.3 m (thus hw is evaluated according to Eq. (17)). Table 7 summarizes
the features of the vessels analyzed and the densities of the substances
stored, while the reference ooding conditions used for model
assessment and reported in Table 5 were applied to assess vessel
resistance.
Fig. 10ad show the vulnerability values obtained for the tanks
following the developed methodology and applying the envelope
correlations (SET B) for vessels xed on the ground, while Fig. 10eh
show the results obtained considering the presence of the concrete
support. As shown in the gures, high failure probability values are
obtained for the rst three reference ooding conditions considered
(W1W3), while the low-severity ooding W4 does not affect most of
the vessels. In the case of high-speed ooding, vessel failure is mostly
due to the excessive drag force (W2 conditions). Buoyancy is responsible of vessel failure for high-depth oods (W1 and W3 conditions).
Low volume vessels, due to the reduced sizes, are more resistant
to ood action, since the momentums acting on the bolted
connections are lower. Actually, vessels V5 to V10 are those showing
the higher resistance to ood action. Moreover, such vessels also
have a lower inventory that may be release in the case of failure.
As shown in Fig. 10eh, a greater height of the supports (hc)
may be effective in protecting the vessels only if the actual water
depth hw0 (see Section 2.3) is lower than the overall vessel height.
If the expected frequency, f (y 1), of a ood event having a
given intensity (vw;hw) is known, the frequency of loss of containment (LOC) associated to the damage induced by ooding may be
calculated as follows:
Capacity
(m3)
3.4. Application to a case-study
ID Vessel
constant), but showing no systematic deviations, hence with a

good statistic performance.
Fig. 8c shows the chart with the results obtained in the application
of the method of Hanna et al. [51]. In the chart a point was obtained
both for SET A and B correlations for each ooding condition (e.g.,
W1W4). The reference curve dened by Eq. (34) is also reported. As
shown in Fig. 8c, for both correlation sets, the points fall above the
reference curve, hence showing no systematic deviations. It is worth to
mention that despite the high VG values, MG values are limited. In
fact, all the points associated to correlation SET A are inside the range
of MG between 0.5 and 2, thus demonstrating the good correlation
performance [51]. Correlation SET B is, as expected, more conservative,
but still with MG values close to 0.5.
Table 7
Case study denition and results: features of the vessels considered and expected failure frequency (y 1) due to each ooding condition dened in Table 5 considering the tanks anchored to the ground. MOD: frequency predicted
by the deterministic model; SET A and SET B refer to data predicted by the correspondent set of simplied correlations; NF: No predicted failure. Tank data were derived from the database reported in Table 1; for stored substance
features, see Table 5 at the correspondent ID.
140
141
Table 8
Expected failure frequency (y 1) due to each ooding condition dened in Table 5 implementing the concrete base (hc 0.3 m). MOD: frequency predicted by the
deterministic model; SET A and SET B refer to data predicted by the correspondent set of simplied correlations; NF: No predicted failure. For vessels features refer to Table 7
at the correspondent ID.
ID Vessel

MOD
V1-V4
V5-V10
V11-V16
V17-V19
V20-V23
V24-V26
V-27-V30
V31
V32-V35
V36-V39
SET A
4
8.98 10
1.66 10 3
2.00 10 3
1.37 10 3
7.48 10 4
1.74 10 3
7.26 10 4
8.98 10 4
9.99 10 4
7.48 10 4
SET B
3
1.40 10
2.00 10 3
2.00 10 3
1.26 10 3
7.83 10 4
1.77 10 3
7.85 10 4
1.40 10 3
5.85 10 4
7.83 10 4
3
1.42 10
2.00 10 3
2.00 10 3
1.39 10 3
8.96 10 4
1.82 10 3
8.89 10 4
1.42 10 3
6.70 10 4
8.96 10 4
MOD
MOD
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
SET A
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
SET B
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
tables, the simplied approach yields conservative values for fLOC,

in good agreement with those predicted by the complete model,
with maximum discrepancies that in general are lower than a
factor 2.
Calculated LOC frequency values for the reference ood scenarios considered range between 1 10 5 and 3 10 3 y 1. A
comparison was carried out with the frequencies of LOC events
due to internal failures available in the technical literature. In
particular, according to the Purple Book [53], data for pressurized vessels range between 5 10 7 y 1 (catastrophic and 10 min
release of entire inventory) and 1 10 5 y 1 (release from a
10 mm equivalent diameter). In the case of atmospheric storage
tanks with single containment, the conventional expected LOC
frequencies are quite higher: 5 10 6 y 1 (catastrophic and 10
minute release of entire inventory) and 1 10 4 y 1 (release from
a 10 mm equivalent diameter). These gures are more than one
order of magnitude lower than the site-specic ood-induced LOC
events calculated in the case-study. This conrms that in oodprone zones, NaTech scenarios triggered by oods may signicantly contribute to the overall risk of an industrial facility.
Nevertheless, depending on the type of expected ood scenario,
an increase in the height of the anchorage was evidenced as a
possible protection barrier. However, its application should take
into account both parameters related to the credible ooding
scenarios and the resistance of the vessel.
4. Conclusions
A model was developed to calculate the failure probability of
horizontal cylindrical vessels as a function of ood severity.
NF
NF
1.08 10 3
NF
NF
8.01 10 4
NF
NF
1.31 10 4
NF
SET A
SET B
4
7.94 10
NF
1.22 10 3
NF
NF
8.34 10 4
NF
7.94 10 4
NF
NF
4
8.58 10
NF
1.36 10 3
NF
NF
9.54 10 4
NF
8.58 10 4
NF
NF
MOD
SET A
SET B
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
NF
The modeling approach was validated against available literature

data and allowed the identication of the more critical parameters
affecting the vessel resistance to the ood. Several simplied
correlations were derived for the straightforward estimation of
vessel resistance. The application of the developed model and of
the simplied correlations to a case-study conrmed that NaTech
scenarios caused by oods may have an important inuence on
the risk due to major accidents caused by the release of hazardous
substances. The importance of an appropriate design of the vessel
support and basements was evidenced, highlighting the potential
importance of mitigation barriers in the prevention of NaTech
scenarios triggered by oods.
Acknowledgments
The authors gratefully acknowledge nancial support received
from the EU within the 7th FP iNTeg-Risk Project (CP-IP-213345-2).
Appendix
In the following, the results of the application of the failure
model are reported. The results are summarized in charts that
allow a quick overview of the vessels response to pre-determined
ood conditions (some examples are reported in Figs. A1A4). For
each vessel, identied by the vessel ID in Table 1, the failure to a
given ood condition (identied by the imposed ood velocity vw
in m/s and effective depth hw in m) is highlighted by a red color.
On the contrary, if the model does not predict a failure, the box
is white.
142
Fig. A1. Failure chart for the vessels considered in the present study assuming 50% lling level and stored liquid density of 1100 kg/m3 (water solution containing toxic
contaminant).
Fig. A2. Failure chart for the atmospheric vessels considered in the present study assuming 90% lling level and stored liquid density of 1100 kg/m3 (water solution
containing toxic contaminant).
Fig. A3. Failure chart for the pressurized vessels considered in the present study assuming 50% lling level and stored liquid density of 600 kg/m3 (ammonia).
143
144
Fig. A4. Failure chart for the pressurized vessels considered in the present study assuming 90% lling level and stored liquid density of 600 kg/m3 (ammonia).
References
[1] Knox JC. Large increases in ood magnitude in response to modest changes in
climate. Nature 1993;361:4302.
[2] Milly PCD, Wetherald RT, Dunne KA, Delworth TL. Increasing risk of great
oods in a changing climate. Nature 2002;415:5147.
[3] Min S-K, Zhang X, Zwiers FW, Hegerl GC. Human contribution to more-intense
precipitation extremes. Nature 2011;470:37881.
[4] Seidou O, Ramsay A, Nistor I. Climate change impacts on extreme oods I:
combining imperfect deterministic simulations and non-stationary frequency
analysis. Nat Hazards 2012;61:64759.
[5] Seidou O, Ramsay A, Nistor I. Climate change impacts on extreme oods II:
improving ood future peaks simulation using non-stationary frequency
analysis. Nat Hazards 2012;60:71526.
[6] Young S, Balluz L, Malilay J. Natural and technologic hazardous material
releases during and after natural disasters: a review. Sci Total Environ
2004;322:320.
[7] Cozzani V, Campedel M, Renni E, Krausmann E. Industrial accidents triggered
by ood events: analysis of past accidents. J Hazard Mater 2010;175:5019.
[8] Landucci G, Antonioni G, Tugnoli A, Cozzani V. Release of hazardous substances in ood events: damage model for atmospheric storage tanks. Reliab
Eng Syst Saf 2012;106:20016.
[9] Rasmussen K. Natural events and accidents with hazardous materials. J Hazard
Mater 1995;40:4354.
[10] Fendler R. Floods and safety of establishments and installations containing
hazardous substancesconclusions on a research project of the German
Umweltbundesamt. Nat Hazards 2008;46:25763.
[11] Krausmann E, Renni E, Campedel M, Cozzani V. Industrial accidents triggered
by earthquakes, oods and lightning: lessons learned from a database analysis.
Nat Hazards 2011;59:285300.
[12] Cozzani V, Krausmann E, Reniers G. Other causes of escalation. In: Reniers G,
Cozzani V, editors. Domino effects in the process industries Modelling,
prevention and managing. Amsterdam: Elsevier; 2013. p. 15474.
[13] Petrova EG, Krausmann E. From natural hazards to technological disasters. Nat
Hazards Earth Syst Sci 2011;11:30635.
[14] Krausmann E, Cruz AM. Natech disasters: when natural hazards trigger
technological accidents. Nat Hazards 2008;46:139263.
[15] Renni E, Krausmann E, Cozzani V. Industrial accidents triggered by lightning.
J Hazard Mater 2010;184:428.
[16] Cruz AM, Steinberg LJ, Vetere-Arellano L. Emerging issues for natech disaster
risk management in Europe. J Risk Res 2006;9:483501.
[17] Salzano E, Iervolino I, Fabbrocino G. Seismic risk of atmospheric storage tanks
in the framework of quantitative risk analysis. J Loss Prev Process Ind
2003;16:4039.
[18] Salzano E, Garcia Agreda A, Di Carluccio A, Fabbrocino G. Risk assessment and
early warning systems for industrial facilities in seismic zones. Reliab Eng Syst
Saf 2009;94:157784.
[19] Necci A, Antonioni G, Cozzani V, Krausmann E, Borghetti A, Nucci CA. A model
for process equipment damage probability assessment due to lightning. Reliab
Eng Syst Saf 2013;115:919.
[20] Krausmann E, Cozzani V, Salzano E, Renni E. Industrial accidents triggered
by natural hazards: an emerging risk issue. Nat Hazards Earth Syst Sci
2011;11:9219.
[21] Povinec PP, Hirose K, Aoyama M. Fukushima accident radioactivity impact on
the environment. Amsterdam: Elsevier; 2013.
[22] Campedel M, Cozzani V, Garcia-Agreda A, Salzano E. Extending the quantitative assessment of industrial risks to earthquake effects. Risk Anal
2008;28:123146.
[23] Antonioni G, Bonvicini S, Spadoni G, Cozzani V. Development of a frame work
for the risk assessment of Na-Tech accidental events. Reliab Eng Syst Saf
2009;94:144250.
[24] Antonioni G, Spadoni G, Cozzani V. A methodology for the quantitative risk
assessment of major accidents triggered by seismic events. J Hazard Mater
2007;147:4859.
[25] Landucci G, Gubinelli G, Antonioni G, Cozzani V. The assessment of the
damage probability of storage tanks in domino events triggered by re. Accid
Anal Prev 2009;41:120615.
[26] Cozzani V, Gubinelli G, Antonioni G, Spadoni G, Zanelli S. The assessment of
risk caused by domino effect in quantitative area risk analysis. J Hazard Mater
2005;127:1430.
[27] Landucci G, Molag M, Reinders J, Cozzani V. Experimental and analytical
investigation of thermal coating effectiveness for 3 m3 LPG tanks engulfed by
re. J Hazard Mater 2009;161:118292.
[28] Campedel M. Analysis of major industrial accidents triggered by natural
events reported in the principal available chemical accident databases. Report
EUR 23391 EN. Ispra (I): commission of the European Communities; 2008.
145
[29] New South Wales Government. Floodplain Development Manual: the management of ood liable land. Appendix L hydraulic and hazard categorization.
Sydney: Department of Infrastructure, Planning and Natural Resources; 2005.
[30] Rijkswaterstaat. Flood risks and safety in the Netherlands (Floris), Report
DWW-2006-014. Delft: Rijkswaterstaat, Dutch Ministry of Infrastructure and
Environment; 2005.
[31] American Society of Mechanical Engineers (ASME). Boiler and pressure vessel
committee. Boiler and pressure vessel code. New York: American Society of
Mechanical Engineers; 1989 (Section VIII. Div. 1).
[32] American Petroleum Institute (API). Design and construction of large, welded,
low-pressure storage tanksAPI standard 620. Washington D.C.: American
Petroleum Institute; 2002.
[33] Sinnott RK. Coulson and Richardson's Chemical Engineering, Vol. 6. Oxford:
Chemical Engineering Design: Butterworth-Heinemann; 1999.
[34] Army US. Corps of Engineers. The great ood of 1993 post-ood report: Upper
Mississippi River and lower Missouri River basins. St. Louis: U.S.: Army Corps
of Engineers; 1993.
[35] Gruntfest E, Pollack D. Warnings, mitigation, and litigation: lessons for
research from the l993 ood. Update Water Resour 1994;95:405.
[36] Tilton JN. Fluid and particle dynamics, Section 6. Perry's Chemical Engineers'
Handbook. 7th edition. New York: McGraw-Hill; 1999.
[37] Gudmestad OT, Moe G. Hydrodynamic coefcients for calculation of hydrodynamic loads on offshore truss structures. Mar Struct 1996;9:74558.
[38] European Committee For Standardization (CEN). European Standard EN 19931-8:2005. Eurocode 3: Design of steel structures. Part 18: Design of joints.
ICS 91.010.30. Brussels: CEN; 1993.
[39] Research Council on Structural Connections (RCSC). Specication for structural
joints using high-strength bolts. Chicago: RCSC; 2009.
[40] American Society for Testing Material (ASTM). Standard specication for
carbon and alloy steel externally threaded metric fasteners. West Conshohocken: Pa: ASTM; 2004 (ASTM standard F 568M-2004).
[41] International Organization for Standardization (ISO). ISO 7411:84 hexagon
bolts for high-strength structural bolting with large width across ats (thread
lengths according to ISO 888) product grade C property classes 8.8 and
10.9. Geneva: ISO; 1984.
[42] International Organization for Standardization (ISO). ISO 7413:1984 hexagon
nuts for structural bolting, style 1, hot-dip galvanized (oversize tapped)
product grades A and B property classes 5, 6 and 8. Geneva: ISO; 1984.
[43] Kulak GL, Fisher JW, Struik JHA. Guide to design criteria for bolted and riveted
joints. 2nd ed.. Chicago: American Institute of Steel Construction, Inc; 2001.
[44] Ramachandra Rao A, Hamed KH. Flood frequency analysis. London: CRC Press;
2000.
[45] Charlton R. Fundamentals of uvial geomorphology. Abingdon, Oxon: Routledge; 2008.
[46] Bryant E. Natural Hazards. New York: Cambridge University Press; 2005.
[47] Holmes R.R. Jr. Field methods for hydrologic and environmental studies. In:
Holmes RR Jr., Terrio PJ, Harris MA, Mills PC editors. Introduction to eld
methods for hydrologic and environmental studies. Open-File Report 01-50,
Urbana: US Department of Interior U.S. Geological Survey; 2001, p. 175.
Available at http://il.water.usgs.gov/pubs/ofr01-50_chapter1.pdf.
[48] Department of Regional Development and Environment Executive Secretariat
for Economic and Social Affairs Organization of American States. Primer on
natural hazard management in integrated regional development planning.
Washington, D.C.: Dept. of regional development and environment. Executive
secretariat for economic and social affairs. Organization of American states;
1991.
[49] Pistrika AK, Jonkman SN. Damage to residential buildings due to ooding of
New Orleans after hurricane Katrina. Nat Hazards 2010;54:41334.
[50] Bates PD, Dawson RJ, Hall JW, Horritt MS, Nicholls RJ, Wicks J, et al. Simplied
two-dimensional numerical modelling of coastal ooding and example
applications. Coast Eng 2005;52:793810.
[51] Hanna SR, Strimaitis DG, Chang JC. Hazard response modeling uncertainty
(a quantitative method). Report ESL-TR-01-28. Westford, MA, USA: Sigma
Research Corporation; 1993.
[52] Italian Ministry of the Interior. Decreto Ministeriale 14/02/1997. Direttive
tecniche per l'individuazione e la perimetrazione, da parte delle regioni, delle
aree a rischio idrogeologico. Rome: Italian Ministry of the Interior; 1994.
[53] Uijt de Haag PAM, Ale BJM. Guidelines for quantitative risk assessment (Purple
Book). The Hague: Committee for the Prevention of Disasters; 1999.
[54] Liley PE, Thomson GH, Friend DG, Daubert TE, Buck E. Physical and
chemical data. (In: Perry's Chemical Engineers' Handbook). 7th ed.. New York:
McGraw-Hill; 1999.
[55] Boyce MP. Transport and storage of uids, Section 10. (In: Perry's Chemical
Engineers' Handbook). 7th ed.. New York: McGraw-Hill; 1999.

1 s2.0 S0951832014001719 Main

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

1 s2.0 S0951832014001719 Main

Încărcat de

Drepturi de autor:

Formate disponibile

Reliability Engineering and System Safety 132 (2014) 125145

Contents lists available at ScienceDirect

Reliability Engineering and System Safety

Release of hazardous substances in ood events: Damage model

Corresponding auth. Tel.: 39 051 2090240; fax: 39 051 2090247.

frequency assessment needs to take into account both parameters

obtained from actual data available from industrial facilities and

2.1. Representation of vessel geometry (step 1)

Mechanical model set-up

Extended data set on

Fig. 2. Schematization adopted to describe the impact of oodwater on horizontal

Fig. 2c shows in detail the schematization of the saddle base plate

Buoyancy may thus be expressed as a function of vessel geometry,

where g is the gravity constant (9.81 m/s ), Wt is the tare weight

The parameter in Eq. (2) is the fraction of the vessel volume

evaluation of . Further details on the calculation of are provided

where kw is the hydrodynamic coefcient [36,37] and Aext is the

where Fn,b is the portion of normal force allocated to each bolt.

where l3 is the distance between the anchored saddle and the

It is worth to mention that buoyancy may also induce a bending

(1) Selection of input parameters:

(2) Selection of input parameters:

(3) Selection of input parameters:

(4) Evaluation of vessel resistance with

Design pressure of the vessel

Actual depth of ood water over the ground level

Expected frequency of the ooding scenario

Geometrical parameter dened as follows: min [hmin; D]

Vessel submerged fraction:

Vessel external volume (assuming spherical edges): V ext D4 L D D6

V int D 42t L D D 62t

Buoyancy due to the ooding (Fig. 2b):

(5) Evaluation of vessel resistance with

(6) Evaluation of parameters

Failure criterion FC1 for the i-th bolt:

Saturation pressure data were derived from [54].

Fig. 2b). The torque (Mt) is schematized in Fig. 2c and may be

It is worth to notice that Eq. (9) may lead to conservative

The procedure reported in a reference technical standard for

connection is assumed if at least one of the following criteria is

in which Ares is the resistant area of each bolt. Conditions

Typical diameter range (m) of

were obtained from vessel diameter on the basis of literature

Since there are different types of ood events (e.g. oodplain

Hence, hmin is the minimum ooding height needed to wet the

This latter parameter, representing the effective water depth, is

values were derived analyzing the features of critical ooding

where hw0 is the actual depth of ood water. Clearly enough hw is

Past accident data analysis on NaTech scenarios involving

for both atmospheric and pressurized equipment, water was

of lling levels between 1% and 90% was thus considered, as

2.5. Dataset of failure conditions (step 5)

Filling level (%)

2.6. Simplied correlations for vessel damage (step 6)

Vapor density Typical type of

Flood water actual Expected occurrence

See Table 2 for details on v evaluation.

Flooding velocity vw (m/s)

Flooding velocity vw (m/s)

Wetting height (hwet) (m)

Flooding velocity vw (m/s)

Flooding velocity vw (m/s)

Wetting height (hwet) (m)

VG expln mod ln corr 2 exp