Corrosion Science 51 (2009) 2014–2026

Contents lists available at ScienceDirect

Corrosion Science
journal homepage: www.elsevier.com/locate/corsci

Influence of environmental factors on corrosion of ship structures in marine

atmosphere
C. Guedes Soares a,*, Y. Garbatov a, A. Zayed a, G. Wang b
a
Centre for Marine Technology and Engineering, Technical University of Lisbon, Instituto Superior Técnico, Av. Rovisco Pais, 1049-001 Lisboa, Portugal
b
American Bureau of Shipping, Technology ABS Plaza, 16855 Northchase Dr., Houston, Texas 77060-6006, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper deals with the modelling of the effects of relative humidity, chlorides, and temperature on the
Received 20 January 2009 corrosion behaviour of ship steel structures subjected to marine atmospheres. A new corrosion wastage
Accepted 19 May 2009 model is proposed based on a reference non-linear time-dependent corrosion model that is modified by
Available online 27 May 2009
the effect of different environmental factors contained in the marine atmosphere. The model assesses the
corrosion degradation under stationary environmental conditions denoted as ‘‘short-term”. The long-
Keywords: term corrosion degradation is predicted by considering the succession of the various environmental con-
General corrosion
ditions that can be present in the marine atmosphere during a ship lifetime and adding the corrosion
Marine atmosphere
Ship structures
damage incurred during each of them. Corrosion records (depending only on time) are used to calibrate
Environmental factors the reference model while the effect of environmental factors in increasing or decreasing the corrosion
rate is based on data published by other authors. A numerical example of a representative application
of the new corrosion model is presented, to demonstrate how to apply the model.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction
Corrosion and corrosion-related problems are considered to be
the most important factors leading to age-related structural degra-
dation of ships and many other types of steel structures. Corrosion
has a harmful consequence from the point of view of safety and can
lead to thickness penetration, fatigue cracks, brittle fracture and
unstable failure. These failures can imply a risk of loss of human
lives and a risk of polluting the environment depending on the ship
type.
At the ship design stage corrosion additions will increase the re-
quired net thickness of structural hull members to compensate for
expected thickness reduction during the assumed ship life. After
commissioning, ships are surveyed at regular intervals to investi-
gate their condition, and any worn structural members are re-
placed, based on a wastage allowance. However, some ship
owners even prefer to build ships with higher plate thicknesses
to reduce the number of interruptions of ship operation during
the ship life for material replacement.
The decision about the preferred level of safety and conse-
quently of the plate thickness to be included in the ship is more of-
ten made on the basis of reliability studies that incorporate a
model of corrosion growth. Guedes Soares and Garbatov [1] pre-
sented a time-variant formulation to model the degrading effect
* Corresponding author. Tel.: +351 2184179607; fax: +351 218474015.
E-mail address: guedess@mar.ist.utl.pt (C. Guedes Soares).
that corrosion has on the reliability of ships’ hulls. They also con-
sidered one repair policy showing the effect of plate replacement
when its thickness reached 75% of the as built thickness. Wirsching
et al. [2] presented a reliability assessment relative to the ultimate
strength failure of a ship hull experiencing structural degradation
due to corrosion. Shama et al. [3] demonstrated the effects of dete-
rioration by corrosion on the reliability of double hull tanker plates
subjected to different loading conditions.
General corrosion, which is the most common form of corro-
sion, takes place over the entire surface of the metal and is the
one addressed here. The general practice to monitor the growth
of corrosion during ships life is to conduct thickness measurements
at regular inspections for maintenance and classifications accord-
ing to the requirements of IACS (International Association of Clas-
sification Societies) [4]. The ship hull periodical surveys are
classified as annual, intermediate and class renewal survey, and
other mandatory surveys may also be performed.
The thickness measurements indicate the amount of wastage as
a function of the age of structural elements and this data is used to
develop relations that predict how corrosion wastage grows with
time.
Typically linear relations have been used in traditional models.
However, more recently attention has been given to the detailed
aspects of steel corrosion and models that explain the various cor-
rosion mechanisms have been proposed mainly by Melchers [5],
indicating that the initial phases of are of a very high corrosion
growth, which tends to level off later.

0010-938X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2009.05.028
 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026
The ship owner associations and the Classification Societies col-
lect data of ship plating thickness measurements and some dat-
abases are available worldwide as reported by TSCF (Tanker
Structure Co-operative Forum) [6], Yamamoto and Ikegami [7]
and Wang et al. [8]. These databases show that the historical data
on corrosion wastage has a rapid growth in the initial years and
tends to level off later.
As a result of this type of evidence a new set of models exhibit-
ing this type of behaviour have been proposed by Guedes Soares
and Garbatov [9,10], Paik et al. [11] and Qin and Cui [12]. These
models have been calibrated with data from tankers [13] and from
bulk carriers [11,14] showing a very good adjustment.
All of these models do not represent the succession of corrosion
mechanisms described by Melchers [5], but they represent the
overall trend that can be seen both in Melchers model as well as
in ship historical data. Therefore, they are appropriate models to
be used by Classification Societies and ship owners to predict the
growth of corrosion in general and to plan inspections on that
basis.
However, it has been shown that sister ships can experience
different levels of corrosion, showing that the application of
such models represent average situations but can have signifi-
cant deviations when applied to one specific ship. Furthermore,
experience has shown that even in the same ship corrosion
rates vary significantly from location to location. The reason is
that the environmental conditions that are present in the differ-
ent ship spaces and which different ships are subjected are
different.
A first approach to solve this question has been adopted by sep-
arating the corrosion thickness measurements in areas of relatively
constant environmental conditions, such as deck, tanks, hull etc.
However, the definitive way of dealing with this problem is to rep-
resent explicitly the effect that the various environmental factors
have on corrosion.
Literature reviews by Melchers [5], Zayed et al. [15] and Pana-
yotova et al. [16] have identified the main corrosion mechanisms
that can be found in ship structures and the main environmental
factors that affect them.
Corrosion of deck plating is affected by green water, rain and ser-
vice water accumulation, in addition to the high humidity and the
aggressive marine atmosphere. Side shell plating is subjected to
three different environments. The upper part of it is subjected to
the aggressive atmospheric environment, rich in high chloride con-
tent, oxygen and other corrosive minerals, in addition to the seawa-
ter spray arising from the ship motion and wave effects. This region
is also subjected to high relative humidity. Salts may be detected,
which may result from salt spray blown by wind. The variations
in relative air humidity cause either the evaporation or condensa-
tion of water. The gases contained in the air may have carbon diox-
ide (CO2), hydrogen sulphide (H2S), sulphur dioxide (SO2), or
sulphur trioxide (SO3). They activate the thin layer of electrolyte
and accelerate the rate of corrosion even more, since this process
takes place in an oxygen-rich atmosphere. The effect of tempera-
ture and the different environmental factors on the atmospheric
corrosion behaviour was illustrated by various studies such as Am-
bler and Bain [17], Ailor [18], Grossman [19] and Vernon [20].
The lower part of the side shell of ship hull is completely im-
mersed in seawater. Water properties such as salinity, temperature,
oxygen content, pH level and chemical composition can vary
according to location and water depth. These properties actually af-
fect the corrosion degradation in that part of the ship as indicated,
for example, by Mercer and Lumbard [21], LaQue [22], Dexter [23]
and Melchers [5,24,25]. The flow of water contributes to accelerated
corrosion as studied by LaQue [26] and Melchers and Jeffrey [27].
Between the immersed and the atmospheric part of the side of
ship hulls, the side shell plating is subjected to a more aggressive
2015
environment. This zone is a very aerated area due to the turbulence
of the sea surface around the ship hull. These actions increase the
oxygen content of this layer, in addition to the wear effects arising
from the motion of the sea surface against the side shell and the
wet and dry effect making this region a highly corroded zone.
The extent of the maximum corrosion in the still waterline zone
is determined by the sea state conditions and the atmospheric
temperature. The maximum rate of attack in colder climates would
be smaller, while with more intensive wave action in warmer cli-
mates it would be greater [22].
Inside ship cargo tanks, different corrosive environments also
exist. The lower part is in contact with the cargo. The ceiling and
the upper part of the tank are in contact mainly with gaseous
atmospheres [28]. Between the previous parts, there is an area that
sometimes is in contact with cargo and others with gaseous atmo-
sphere depending on the level of cargo in the tank. Yasunaga et al.
[29] studied the corrosion degradation of the upper deck of cargo
oil tank of an oil tanker, while Katoh et al. [30] investigated local-
ized corrosion of cargo oil tank bottom plate of an oil tanker.
As these situations involve different types of environments that
will induce corresponding types of corrosion, the approach
adopted has been to develop a specific model for each. Gardiner
and Melchers [28] proposed a model for enclosed atmospheric cor-
rosion within ship spaces based on temperature, salt deposition
and time of wetness. Guedes Soares et al. [31] have suggested
how to deal with corrosion in the immersed part of ship hulls while
Guedes Soares et al. [32] proposed a model to represent the effect
of environmental factors inside cargo tanks. This paper comple-
ments those by dealing with atmospheric corrosion, and adopting
a similar general approach.
Corrosion involves the interaction between metal or alloy and
its environment and is affected by the properties of both the mate-
rial or alloy and the environment. Atmospheric corrosion occurs on
a steel surface in contact with the atmosphere and thus is not im-
mersed in water. However, for corrosion to occur, a thin wet film
needs to be created by the humidity in the air in combination with
impurities.
Atmospheric environments vary depending on the chemical
constituents. Some major chemical constituents of the atmo-
spheric marine environment are relatively constant worldwide.
However, the minor constituents vary from site to site and with
season and storms. The variations in the chemistry of open ocean
atmosphere tend to take place slowly and over horizontal and ver-
tical scales that are large in comparison with the dimensions of
most marine structures. Such gradual changes may produce an
equally gradual change in the corrosion rate of structural materials
with season and location, but they are unlikely to produce sharp
changes in either corrosion mechanism or rate.
A number of studies related to corrosion evaluation under mar-
ine atmospheric conditions have been reported in the specialized
literature. Grossman [19] showed the corrosion rate per day of
wetness and demonstrated the accelerating effect of temperature
on corrosion rate at six test sites. Sodium chloride is a chief con-
taminant, and salt spray is principally responsible for metal corro-
sion in ocean environments. Ambler and Bain [17] performed a
series of tests in Nigeria to illustrate these effects. Specimens of
steel were exposed to the atmosphere at distances from the ocean
ranging from a few yards to over 100 miles. The amount of sea salts
measured off the coast of Nigeria illustrate this relationship be-
tween the salinity and the corrosion rate. Ailor [18] studied the ef-
fect of SO2 levels on the corrosion rate of carbon steel, using data
from three Norwegian test sites. The data shows that as SO2 con-
centrations are increased, the corrosion rate, measured as weight
loss, increases. For atmospheric corrosion, the presence of mois-
ture, as determined by the time of wetness, is probably the most
important effect [33]. Gardiner and Melchers [28] investigated
2016 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026

the process of enclosed atmospheric corrosion within ship spaces. 1.3 0.13
They proposed that corrosion rates are influenced mainly by three 1.2 0.12
parameters; time of wetness, salt deposition and temperature.

Corrosion rate, ddn (t )/dt (mm/year)

1.1 0.11
Based on the available experimental data they showed that, as a 1.0 0.1

Corrosion Depth, dn (t ) (mm)

first approximation, the rate of atmospheric corrosion is linearly 0.9 0.09
proportional to both temperature and salt deposition.
0.8 0.08
The effect of temperature on the atmospheric corrosion behav-
0.7 Corrosion depth 0.07
iour was illustrated by Ailor [18]. Panayotova et al. [16] and Zayed
0.6 Corrosion rate 0.06
et al. [15] have described the factors governing marine atmo-
0.5 0.05
spheric corrosion phenomena on the structural steel components
level in open air and in marine environment. Paik et al. [34] re- 0.4 0.04
viewed different studies performed in the field of marine corrosion 0.3 0.03
assessment and modelling and the different environmental and 0.2 0.02
operational parameters related to it. 0.1 0.01
This work has the aim of studying the effect of some marine 0.0 0
atmospheric factors on corrosion. The work starts from reference 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
atmospheric conditions in which the corrosion of steel is described Time, t (years)
by the non-linear time-variant corrosion model of Guedes Soares
Fig. 1. Time-dependent standard corrosion model.
and Garbatov [9], whose parameters are fit to the measured data
of the progress of corrosion with time. This model only depends
on time and does not account for the influence of any of the atmo-
spheric variables. However, since it is fit to measured data over and performing a regression analysis in order to fit the proposed
many years and many different routes it represents the effect of model to this data using the least squares approach for example.
some time averaged environmental conditions which constitute Garbatov et al. [13] already fitted this model to two sets of corro-
the reference situation in this study. sion data for deck plates of ballast and cargo tanks of tankers pro-
It is important to note here that some of the corrosion data col- vided by ABS [36] and Wang et al. [8]. The first set includes 1168
lected from the literature [17,19–21,24,25,27,35], are based on measurements of deck plates from ballast tanks with original nom-
experimental tests of unloaded specimens and differ from the cor- inal thicknesses varying from 13.5 to 35 mm on ships with lengths
rosion of structural components subjected to time varying loading. between perpendiculars in the range of 163.5–401 m. The second
The model proposed here considers that changes in the atmo- set of data includes 4665 measurements of deck plates from the
spheric conditions and will increase or decrease the corrosion rate, cargo tank with original nominal thicknesses varying from 12.7
depending on its specific effect. Therefore, the approach adopted is to 35 mm on ships with lengths between perpendiculars in the
to develop a set of functions that will modify the reference model range of 163.5–401 m. The descriptors of the fitted model (see
by increasing or decreasing the corrosion rate depending on some Fig. 2) were as indicated in Table 1. Using regression analysis, the
governing variables in a similar way as done in Guedes Soares et al. standard deviation of corrosion wastage as a function of time
[31,32]. was fitted to the following equation [13]:
In the next section, an assessment will be made of the average Stdev ½dn ðtÞ ¼ 0:384 lnðtÞ  0:710 ð3Þ
environmental conditions in typical ship routes. These average
conditions will be the ones that will be considered to be related This model was developed to describe the time variation of the
with the reference model of Guedes Soares and Garbatov [9]. In thickness of ship plates and considers the only variables that are
the following section, a model for long-term corrosion is described normally available from ship surveys: thickness and time. There-
after which the effect of different environmental effects is analysed fore, this model has a shape that reflects the main trends in the
and models are proposed for each one. Finally, an example demon- long-term corrosion process and does not aim to describe in an
strates how this method can be applied. accurate way the various phases in which the corrosion process
goes through, in particular in the short term, as aimed by Melchers
[5]. Therefore, in some sense it is a different type of model.
2. Reference corrosion model

The time-dependent reference atmospheric corrosion behaviour

2.0
is represented by the non-linear time-variant model proposed by
Mean corrosion Depth, E [dn (t )] (mm)

Guedes Soares and Garbatov [9] (see Fig. 1): 1.8 Observed
( 1.6 Estimated
@dn ðtÞ d1
st eðtsc Þ=st t > sc
¼ ð1Þ 1.4
@t 0 t 6 sc
1.2
( 1.0
d1 ½1  eðtsc Þ=st  t > sc
dn ðtÞ ¼ ð2Þ 0.8
0 t 6 sc
0.6
This model is governed by several parameters that represent the
0.4
long-term non-linear description of corrosion deterioration under
0.2
average environmental conditions. st is the transition time, sc is
the coating life and d1 is the long-term thickness of the corrosion 0.0
0 4 8 12 16 20 24 28 32 36
wastage.
Time, t (years)
The reference corrosion in atmospheric condition and the gov-
erning parameters of the corrosion model are normally derived Fig. 2. Fitting of the time-dependent corrosion wastage of deck plates of ballast
based on the measurement data during a very long period of time tanks.
 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026
Table 1
Parameters of the corrosion model [13].
sc st
Ballast 10.54 17.54
Cargo 11.494 11.23
It is known that environmental conditions affect the rate of cor-
rosion and thus if one wants a model that instead of being applica-
ble to a whole class of ships, only applies to a subclass that has
some specific common features, then a more detailed model is re-
quired. Furthermore, this model can also answer the question of
how will the corrosion change if the vessels are subjected to other
routes with different environmental conditions. This model ini-
tially proposed in Guedes Soares et al. [31] and also adopted in
Guedes Soares et al. [32], is based on the postulate that the model
of Eq. (2) calibrated to data, as exemplified in Table 1, will be mod-
ified by multiplicative corrosion enhancement factors, which will
increase or decrease the corrosion rate depending on the environ-
mental conditions on which corrosion is developing.
To apply this model, it is necessary to determine which are the
average environmental conditions to which the model that was fit-
ted to the operational data refers to. In general, this can be done by
recording the environmental variables onboard and determining
their averages. However, as the data used to determine the values
in Table 1 does not have records of the environmental data during
ship operation, this will not be possible. So these conditions will
need to be estimated as described hereafter.
This corrosion data are collected from several tankers with dif-
ferent ages, geometric characteristics and navigating in various
routes through the world. This means that this corrosion data sets
represent marine corrosion through the ship life under long-term
mixed environmental conditions which change through the ship
route and season. From this it can be assumed that fitting the stan-
dard corrosion model to this data and its resulting parameters cor-
respond to long-term mean values of the different environmental
factors over the navigational area. So, it is important now to define
the most important environmental factors affecting the corrosion
process in marine atmosphere and the mean value corresponding
to each one. The mean values are considered thought this study
as the nominal or standard environmental conditions relating to
the standard corrosion model and its related standard parameters.
The most important environmental factors affecting corrosion un-
der atmospheric conditions, which will be considered through this
study, will include relative humidity, chlorides, and temperature.
The mean values of the environmental factors considered here
are estimated based on the published data. The chloride mean va-
lue is averaged over several coastal sites. The data corresponding to
sites near from the coast was only considered to reflect the marine
atmosphere accurately. There are several publications and experi-
mental works conducted to measure the chloride contents of the
marine atmosphere.
Table 2
Elected publications and sites for the estimation of chloride mean value.
Reference/location
Ambler and Bain [17], Nigeria (average)
Cook et al. [37], Gulf of México, Campeche,
Mx, 4 m from the Gulf
Cook et al. [37], Gulf of México, Kure Beach,
NC, 25 m from the Gulf
Cook et al. [37], Gulf of México, Kure Beach,
NC, 250 m from the Gulf
Mean
2017
Table 2 illustrates the references and test locations which are
selected for the estimation of the chloride mean value in this study.
d1 The chloride concentrations measured by Ambler and Bain [17] in
1.85 Nigeria are averaged and considered (see Table 2). The data pro-
1.91 vided by Cook et al. [37] for the Gulf of México in both of Campeche
(México) and Kure Beach (USA) at distances from the Gulf not more
than 250 m were the ones considered. Measurements made at
large distances from the sea were not considered to avoid inland
atmospheric characteristics and guarantee more accurate marine
atmospheric characteristics.
The long-term mean value of both of the temperature and rela-
tive humidity are estimated based on the charts provided by
NOAA/ESRL [38]. These charts represent the long-term mean value
distribution of both of temperature and relative humidity average
through the period from January to December. The corresponding
values were averaged all over the navigational area (see Table 3).
It is necessary to emphasize that using data from different
sources and pooling the data is a process that is liable to have large
uncertainties, as data has been produced by various authors in dif-
ferent conditions. However, at the same time as it is the best esti-
mate that can be produced at this stage with the available data it
also allows the explanation of the method proposed, which can
be improved when more accurate data becomes available.
3. Long-term corrosion model
The corrosion degradation given by Eqs. (1) and (2) give the
atmospheric corrosion rate and corrosion depth at nominal condi-
tions, which should be further corrected for the actual value of the
environmental variables that the referred ship is subjected to.
Neglecting the effect of the height, orientation and sunlight on cor-
rosion of ship hulls, the influence of atmospheric variables on cor-
rosion rate is modelled by multiplicative scaling factors that
increase or decrease the corrosion rate predicted by the reference
corrosion model with the average or nominal values of the envi-
ronmental variables. This influence is derived basically from exper-
imental field data and observations reported in the literature by
various authors.
The corrosion correction factors for each environmental factor
can be estimated from the models described in the following sec-
tion. Through this study, the effect of relative humidity, chlorides,
and temperature are analysed though the marine atmosphere.
Based on the analysis performed in this study, the effect of deviat-
ing sulphur dioxide concentration from the standard value shows
negligible effect on the corrosion in marine atmosphere. For this
reason, the effect of sulphur dioxide is not included and the pro-
posed model will only consider the corrections corresponding to
relative humidity, chlorides and temperature.
Therefore, the corrosion at any environmental conditions can be
estimated as:
@dc;k ðtÞ Q3
@t
¼ f ðRhr Þf ðClr Þf ðT r Þ @d@tn ðtÞ ¼ j¼1 f ðxj Þ
@dn ðtÞ
@t
Q3 ð4Þ
dc;k ðtÞ ¼ f ðRhr Þf ðClr Þf ðT r Þdn ðtÞ ¼ j¼1 f ðxj Þdn ðtÞ
@d ðtÞ
where, c;k
@t
and dc,k(t) represent the corrected corrosion rates and
corrosion thickness, respectively, through the kth interval of the
vessel route, @d@tn ðtÞ and dn(t) are the nominal corrosion rate and
Chlorides
(Cl) mg/m2/day
Table 3
131.8
Standard mean atmospheric conditions.
170
Factor Value Units
311
Chlorides 180.7 mg/m2/day
Temperature 16.2 (289.3) °C (K)a
110
Relative humidity 81.9 %
180.7 a
K = °C + 273.15.
2018 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026
corrosion thickness, respectively, and f(xj) are the correction factor
corresponding to effect of the environmental parameter xj.
This model assumes that there is no correlation among the ef-
fects, implying that the changes in each of them have cumulative
effects. However, in reality, the different environmental factors
considered here are dependent, but it is assumed that the effect
of the interaction will be small compared with the effect of the
individual changes of each parameter.
The ship encounters a wide range of environmental conditions
through its service life. This total time span may be regarded as a
large number of short intervals during which the environmental
conditions remain constant. The total wastage experienced is then
the sum of the corroded thicknesses in each of the periods.
The long-term corrosion wastage dL(t) is a weighted sum of the
various short-term corrosion wastages, accounting for the relative
duration of time the ship spends in each environment:
X
n X
n Y
m
dL ðtÞ ¼ pk dc;k ðtÞ ¼ pk fk ðxj Þdn ðtÞ
k¼1 k¼1 j¼1
where pk is a weighting factor representing the relative duration of
the kth short-term interval, fk(xj) is the corrosion wastage correction
factor for the kth interval and xj environmental factor, n is the total
number of the short-term corrosion intervals and m is the total
number of the environmental parameters affecting corrosion
wastage.
The sensitivity of the long-term corrosion wastage, dL(t), given
by Eq. (5), with respect to any environmental factor, xi, is given
as the partial derivative of dL(t) with respect to the particular envi-
ronmental factor, xi. Assuming that all environmental factors are
independent from each other, the sensitivity of the long-term cor-
rosion wastage, dL(t), can be calculated as:
" #
@dL ðtÞ Xn d Y
m 1
¼ p f ðx Þ fk ðxj Þ dn ðtÞ
@xi k¼1 k dx k i
i j¼1
This approach uses information from the physics of the basic corro-
sion process to determine the main trend of the time variation of
corrosion wastage as determined by the governing corrosion mech-
anism, as reflected in the model already presented by Guedes
Soares and Garbatov [9].
The common and systematic way of verification the proposed
corrosion models by using an actual corrosion measurement re-
sults is usually possible to obtain for the case of short-term models
dealing with the effect of one parameter as this can be monitored
in laboratory conditions.
However, in case of the models that predict the corrosion degrada-
tion on the long-term, it is not possible to validate them due to
unavailability of such kind of measurements. In order to validate
the long-term time-dependent model that predicts the corrosion
degradation of structural elements along the ship life, as proposed
in this study, it is necessary to have records both of the corrosion
wastage as well as of the relevant environmental parameters. Unfor-
tunately, this does not happen in practice and only structural ele-
ments thicknesses are measured during the ship inspection periods.
Unavailability of the proper corrosion measurements necessary
for the validation of the long-term time-dependent corrosion mod-
el should not detract from developing new models of such kind, as
they can serve as guides about the data that is necessary to start
collecting systematically in ships. In fact, there is a strong need
to have mathematical models capable of predicting the corrosion
degradation under different combination of service conditions so
that one is able to predict the expected relative differences be-
tween ships.
The reference corrosion model used in this study is based on fit-
ting the mathematical function of Eqs. (1) and (2) to measured cor-
rosion data sets gathered during thickness measurements of
tankers deck plates. The correction factors are based on corrosion
data or formulations developed by authors that studied the effect
of each parameter. The effect of the sequential application of the
effects should then add to the corrosion damage.
4. Modelling of the different factors affecting corrosion in the
atmospheric environment
It is normally not straightforward to develop a corrosion model
solely based on theory, because corrosion is a function of many
variables and uncertainties and ship structures operate in a com-
plex environment.
Properties of the marine atmosphere such as relative humidity,
temperature, sunlight, oxygen, chlorides, sulphur oxides and other
chemical composition can vary according to the area of navigation
and even though the different ship spaces. Sunlight influences the
ð5Þ
degree of wetness and affects the performance of coatings and
plastics. It may stimulate biological reactions, such as the develop-
ment of fungi. Ultraviolet (UV) light and photo-oxidation can cause
embrittlement and surface cracks in polymers; which can be
avoided by the addition of UV stabilizers (for example, carbon
black). Furthermore, tanks whose temperature increases from solar
radiation suffer more than others. Component orientation affects
the uniformity and thus the protectiveness of the corrosion oxides
[39]. Vera et al. [35] experimentally studied the weight loss of plain
carbon steel as a function of exposition time for steel samples ex-
posed at different angles to the horizontal level.
Cargo type, tank washing and inerting, corrosion protection
effectiveness and component location have a great effect on the
corrosion phenomena. In crude oil tanks, the most significant cor-
rosive component is the hydrogen sulphide which it contains.
ð6Þ Crudes high in sulphur also contribute to tank corrosion. The sul-
phur compounds present may react with water vapour and oxygen
to produce sulphuric acid which is corrosive to steel. The warm,
moist, salt-laden atmosphere which remains after hot, salt-water
washing is ideal for corrosion to occur. Cold water washing is re-
ported to result in less corrosion than hot water washing. An in-
ert-gas system can, depending on its type, application, and gas
quality, either aggravate corrosion conditions or minimize them.
By reducing the oxygen content of a tank, corrosion can be re-
duced. However, while reducing oxygen content to below 5%, inert
gas may also introduce corrosive elements into a tank. Sulphur
dioxide (SO2) and sulphur trioxide (SO3) contained in inert gas
can combine with the warm moist atmosphere in a tank to form
sulphuric acid which can cause accelerated corrosion of either bare
or coated tank surfaces. However, many studies have found inert
gas to have a beneficial effect in reducing tank corrosion, at least
in the top and upper most portions of the tank [40]. The internal
side of deck plating is subjected to aggressive atmospheric condi-
tions due to high humidity and water condensation.
The inside face of the plates will be exposed to aggressive envi-
ronments existing in cargo tanks. Significant corrosion of elements
in ballast tanks adjacent to heated cargo tanks or tanks with con-
sumables is also possible. Pitting corrosion occurs on horizontal
surfaces, because of frequent washing in cargo/ballast tanks. This
type of wear is quite serious especially in way of the flange edges
since the integrity of the structure may be put at risk in critical
connections [41].
An increased degree of local structural flexibility has been
claimed to increase corrosion rates as time progresses because of
serial increase in scale loss. This implies that the corrosion model
developed by statistical means will usually be different according
to the ship types and cargoes or structural member locations and
categories.
 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026
The model of this study will take into account the effect of rel-
ative humidity, chlorides and temperature only due to their signif-
icant effects on the atmospheric marine corrosion process. For each
of these parameters a correction factor is proposed. These factors
are proposed based on experimental and field observation data
from the literature. This data relates different values of each
parameter to the corresponding observed corrosion rate. From
the scatter of each data set, several mathematical models of
parameter variation are checked against it and the model which
shows the best fit is elected to represent the relation between
the parameter value and the changes in corrosion rate.
From each model corresponding to an environmental parame-
ter, the corrosion rate related to the reference value is calculated
to obtain the reference corrosion rate for that environmental
parameter. Then, for each model, the parameter value is norma-
lised by its reference value to obtain the parameter ratio and the
corresponding corrosion rate is normalised by its reference rate.
This normalised corrosion rate is considered here as the correction
factor which will be adapted to Eqs. (4) through (6).
4.1. Duration of exposure to wetness and relative humidity
Duration of wetness determines the duration of the electro-
chemical process or the percentage of time that the critical humid-
ity is exceeded [42]. The critical relative humidity is the humidity
below which water does not form on a clean metal surface and
thus electrochemical or wet corrosion will not occur.
Relative humidity (Rh) influences corrosion through its effect
on the duration of wetness and can be defined as the percentage
of the water vapour pressure in the atmosphere compared to that
which would saturate the atmosphere at the same temperature.
The relative humidity and temperature variations lead to cyclic
wet and dry periods, the so called wet–dry cycles. The wet–dry cy-
cles are a critical feature of atmospheric corrosion as the alternat-
ing wet and dry periods drastically change the rusting mechanisms
from those obtained in bulk aqueous corrosion. Indeed, earlier
studies have shown that during a wet–dry cycle, the atmospheric
corrosion of low alloy steel can be divided into three stages; wet-
ting stage, wet stage and drying stage. The corrosion rate and the
rust layer modifications are thus correlated to the number and fre-
quency of the wet–dry cycles. Hoerle et al. [43] gave detailed
description and modelling of the corrosion mechanics through
each of these stages and brief survey of related earlier studies.
Vernon [20] has established that a critical relative humidity ex-
ists below which corrosion is negligible. In pure air, rusting is min-
imal at relative humidity below 100%. In the presence of small
concentrations of impurities such as sulphur dioxide, serious rust-
ing can occur when the relative humidity of the air rises above the
critical one. Experimental results show that the critical relative
humidity is in general between 50% and 70% for steel, copper, nick-
el and zinc. It is very much influenced by the nature of the pollu-
tants or contaminants present in the atmosphere. Rusting may
occur when sulphur dioxide is present at about 65–75% of relative
humidity and when chlorides are present, as for the marine atmo-
sphere, it may happen at a relative humidity as low as 40% [44].
An example of rusting has been illustrated by Uhlig and Revie
[42]. Grossman [19] has determined the corrosion rate per day of
wetness and demonstrated the accelerating effect of temperature
on corrosion rate in six test sites.
Davis et al. [45] have shown data representing the dependence
of corrosion rate of steel in contact with particles of sea salts on
relative humidity. The corrosion rate as given by this data shows
a steady increase with increasing humidity. For the marine atmo-
sphere and based on the data provided by NOAA/ESRL [38], it
was observed that the relative humidity is varying almost within
60–100%. Based on this observation and the critical relative humid-
2019
ity for the marine atmosphere, as noted earlier by Chandler [44], it
is seen that the relative humidity through the marine atmosphere
is usually above the critical one of it. This means that the steel sur-
face under marine atmospheric conditions will usually be wet and
the electrochemical or wet corrosion will usually occur. Hence, for
marine atmospheric corrosion, the wet–dry process does not exist
and so it will not be included in the corrosion modelling of this
study.
Based on the range of variation of relative humidity through the
marine atmosphere, as noted earlier by NOAA/ESRL [38], and con-
sidering that this study is addressing ship structures, the data of
Davis et al. [45] with relative humidity within 60–100% is consid-
ered. This data is fitted to a linear function to give a relationship
between the relative humidity and the corrosion rate (see Fig. 3),
with R2 = 0.92. This relationship is considered to be specific for
marine atmospheric corrosion with relative humidity within 60–
100%. It was indicated by Graedel and Frankenthal [46] that it is
generally considered that atmospheric corrosion of carbon or
weathering steels begins at about 60% Rh with a very slow rate
and increases sharply at 80% 6 Rh < 100%, which is consistent with
the data and its fitting given by Davis et al. [45]. Hence, atmo-
spheric marine corrosion below 60% RH will be neglected in this
model, as follows:
o
d ¼ 0:3765½Rh  21:943; ½60% 6 Rh 6 100%
o ð7Þ
d ¼ 0:0 ½Rh < 60%
where do is the atmospheric corrosion rate in mg/dm2/day, Rh rep-
resents relative humidity (%) and R2 measures how successful the fit
is in explaining the variability of the data. Described in another way,
R2 is the square of the correlation between the response values and
the predicted response values. R2 can take any value between 0 and
1, with a value closer to 1 indicating a better fit.
The relative humidity ratio Rhr is defined as the ratio between
the actual relative humidity Rh and the nominal one Rhn,
Rh Rh
Rhr ¼ ¼ ð8Þ
Rhn 81:9
At the reference relative humidity Rhn, the reference corrosion
o
rate dn is estimated from Eq. (7), and equal to 8.9 mg/dm2/day.
By substitution of Rh equal to 81.9[Rhr] from Eq. (8) into Eq. (7)
o
and dividing all the equation by the reference corrosion rate dn and
the ranges of applicability of Eq. (7) by Rhn, one gets the following
relation:
f ðRhr Þ ¼ 3:467½Rhr   2:467; ½0:733 6 Rhr 6 1:221
ð9Þ
f ðRhr Þ ¼ 0:0 ½Rhr < 0:733
where, f(Rhr) is the corrosion rate correction factor for relative
humidity (corrosion rate at actual relative humidity/corrosion rate
12
d o = 0.3765 Rh - 21.943
Corrosion rate, d (mg/dm /day)
10
R 2 = 0.9199
2
8
o
6
4
2
0
50 55 60 65 70 75 80 85 90
Relative humidity, Rh (% )
Fig. 3. Dependence of corrosion rate of steel on relative humidity.
2020
at reference conditions). The correction below Rhr of 0.733 is equal
to zero, since the corrosion below the corresponding value
(Rh = 60%) is neglected as discussed earlier.
The sensitivity of the corrosion rate correction factor to the rel-
ative humidity f(Rhr) is given here as the first derivative of f(Rhr)
with respect to the relative humidity ratio Rhr. Since, the relation
between f(Rhr) and Rhr is linear, the sensitivity of f(Rhr) with re-
spect to Rhr is fixed and equal to the slope of this relation (3.467).
The relative humidity ratio is restricted within 0.73–1.22, which
is corresponding to a relative humidity between 60% and 100%. The
relative humidity through the marine atmosphere usually changes
through that range.
4.2. Chlorides
Sodium chloride is an important contaminant factor. Chlorides
in salt spray are principally responsible for metal corrosion in
ocean environments and various studies have proved that. For
example, Ambler and Bain [17] has illustrated the relationship be-
tween salinity and the corrosion rate based on the sea salts mea-
surement on the coast of Nigeria. Similar tests have been
performed at the Inco marine atmospheric corrosion test station
at Kure beach, North Carolina. It was observed that there was a
sharp decrease in the corrosivity in the atmosphere with respect
to a lower concentration of airborne salt 800 ft away from shore
in comparison with an 80 ft distance [22].
In order to propose a mathematical model of the corrosion rate
as a function of the atmospheric chloride concentration, both of the
corrosion data sets (at 80 and 800 ft from the ocean) from the tests
performed at Kure beach, are combined together. It was not possi-
ble to combine also the data given by Ambler and Bain [17] in Nige-
ria since the corrosion loss is expressed per 100 cm2 while the
corrosion data of Kure beach did not specify the area of corrosion
measurement and consider the duration only. Gardiner and Mel-
chers [28] approximated the rate of atmospheric corrosion as line-
arly proportional to salt deposition and showed that this is
consistent with available experimental data. For this study, the
combined data of Kure beach are fitted to a linear function (see
Fig. 4), to give the following linear relationship between the corro-
sion rate and chloride concentration with R2 = 0.55,
o Gardiner and Melchers [28] approximated the rate of atmo-
d ¼ 0:4184½Cl þ 69:062
o
where d is the atmospheric corrosion rate in mg/day of wetness, Cl
represents chlorides contents in mg/m2/day.
The chloride concentration ratio Clr is defined as the ratio be-
tween the actual chloride concentration Cl and the nominal one Cln,
Cl Cl
Clr ¼ ¼ ð11Þ
Cln 180:7
210
d o = 0.4184 Cl + 69.062
Weight loss, d (mg/day of wetness)
180 R 2 = 0.5479
150
120
90
o
60
30
0 mated from Eq. (13), and equal to 146.347 mg/day of wetness.
0 30 60 90
Chloride,Cl (mg/m /day)
Fig. 4. Fitted Kure beach data.
C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026
At the reference concentration of chlorides Cln, the reference corro-
o
sion rate dn is estimated from Eq. (10), and equal to 144.7 mg/day of
wetness.
By substitution of Cl equal to 180.7[Clr] from Eq. (11) into Eq.
(10) and dividing all the equation by the reference corrosion rate
o
dn , one gets the following relation:
f ðClr Þ ¼ 0:523½Clr  þ 0:477 ð12Þ
where, f(Clr) is the corrosion rate correction factor for chloride con-
centration (corrosion rate at actual chloride concentration/corro-
sion rate at reference conditions).
The sensitivity of the corrosion rate correction factor to chloride
content, f(Clr), with respect to the chloride content ratio Clr is given
here as the first derivative of f(Clr) with respect to Clr. Since, the
relation between f(Clr) and Clr is linear, the sensitivity of f(Clr) with
respect to Clr is fixed and equal to the slope of this relation (0.523).
4.3. Temperature
Temperature affects the relative humidity, the dew point, the
duration of wetness, and the kinetics of the corrosion process.
For corrosion in atmospheric conditions, the presence of moisture,
as determined by the duration of wetness, is probably the most
important factor [18]. From the corrosion under atmospheric con-
dition standpoint, dry, hot conditions are preferable to cooler,
moist conditions.
Surface temperature is a critical factor. When the surface tem-
perature increases, the corrosion rate will rise sharply to the point
at which evaporation of the electrolyte takes place. At this temper-
ature, the corrosion rate will decrease quickly [47].
Grossman [19] indicated corrosion rate data sets at six different
sites as a function of temperature, which is fitted to linear func-
tions. The corrosion data of Kure beach, NC, 24 m from the ocean
is the only data used in this analysis. It is the nearest location from
the ocean and shows higher fitting quality, which make it more
appropriate to describe corrosion under marine atmospheric con-
ditions. The corrosion data of Kure beach, NC, 250 m from the
ocean is rejected because it is more far from the ocean and shows
lower fitting quality, while the other data sets are also rejected be-
cause they are corresponding to inland atmospheric conditions.
ð10Þ
spheric corrosion as linearly proportional to temperature and
showed that this is consistent with available experimental data.
Grossman [19] used linear functions to fit the relationship between
the corrosion rate and temperature based on collected field data.
For this study, the corrosion data of Kure beach, NC, 24 m from
the ocean is fitted to a linear function to get a relationship describ-
ing the effect of temperature on corrosion rate (see Fig. 5). The fol-
lowing linear relationship between the corrosion rate and
temperature is obtained with R2 = 0.38.
o 
d ¼ 3:7493½T þ 85:689; T½ C
o ð13Þ
d ¼ 3:7493½T  938:43; T½K
where do is the atmospheric corrosion rate in mg/day of wetness, T
represents temperature.
The temperature ratio Tr is defined as the ratio between the ac-
tual temperature T and the nominal one Tn,
T r ¼ TTn ¼ T
16:2
; T; T n ½ C
ð14Þ
Kure beach, NC, 800 ft from ocean
T r ¼ TTn ¼ T
289:3
; T; T n ½K
Kure beach, NC, 80 ft from ocean o
Fitting line
At the nominal temperature Tn, the nominal corrosion rate dn is esti-
120 150 180 210 240
2 By substitution of T from Eq. (14) into Eq. (13) and dividing all
o
the equation by the reference corrosion rate, dn , one gets the fol-
lowing relation:
 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026 2021

250 250

Corrosion rate, d (mg/day of wetness)

d o = 3.7493 T + 85.689 d o = 3.7493 T - 938.43

Corrosion rate, d (mg/day of wetness)

R 2 = 0.3772 R 2 = 0.3772
200 200

150 150
o

o
100 100

50 50
Kure beach, NC, 24m from ocean Kure beach, NC, 24m from ocean
Linear fitting line Linear fitting line
0 0
0 5 10 15 20 25 30 275 280 285 290 295 300
o
Temperature, T ( C) Temperature, T (K)

Fig. 5. Fitting of Kure beach, NC, 24 m from the ocean data [19]. T in °C (left), T in K (right).

2 2

1.6 1.6
Correction Factor value

Correction Factor value

1.2 1.2

0.8 0.8

0.4 Relative humidity 0.4 Relative humidity

Chlorides Chlorides
Temperature Temperature
0 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Environmental variable ratio Environmental variable ratio

Fig. 6. Correction factors for the different environmental parameters. Tr is estimated based on T, Tn in °C (left), K (right).

Fig. 7. Selected routes through the Pacific Ocean.

2022 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026

Fig. 8. Pacific Ocean atmospheric temperature (left) and atmospheric relative humidity (right).


f ðT r Þ ¼ 0:415½T r  þ 0:585; T r is estimated based on T; T n in ½ C
f ðT r Þ ¼ 7:412½T r   6:412; T r is estimated based on T; T n in ½K
ð15Þ
where, f(Tr) is the corrosion rate correction factor for temperature
(corrosion rate at actual temperature/corrosion rate at reference
conditions).
The sensitivity of the corrosion rate correction factor to temper-
ature, f(Tr), with respect to the temperature ratio Tr is given as the
first derivative of f(Tr) with respect to Tr. Since, the relation be-
tween f(Tr) and Tr is linear, the sensitivity of f(Tr) with respect to
Tr is fixed and equal to the slope of this relation, which is equal
to 0.415 for Tr estimated based on T, Tn in °C and is equal to
7.412 for Tr estimated based on T, Tn in K. It is observed from Eq.
(15) that the temperature correction function with Tr calculated
based on T, Tn expressed in Kelvin will have larger slope and so lar-
ger sensitivity to the temperature ratio Tr, than the temperature
correction function with Tr calculated based on T, Tn expressed in
centigrade.
Fig. 6 shows the correction factors for the different environmen-
tal parameters considered in this study. It is noticed from Fig. 6
(left) that the corrosion rate correction factor is a little more sensi-
tive to chloride ratio than to temperature ratio estimated based on
T, Tn in °C, while the sensitivity to relative humidity ratio is much
higher than that to chloride and temperature ratios. From Fig. 6
(right) it is noticed that the sensitivity of the corrosion rate

Route 1 - RH Route 4 - RH Route 1 - Temp. Route 4 - Temp.

Route 2 - RH Route 5 - RH Route 2 - Temp. Route 5 - Temp.
Route 3 - RH Route 6 - RH Route 3 - Temp. Route 6 - Temp.
Route 1 - Corr. Route 4 - Corr. Route 1 - Corr. Route 4 - Corr.
Route 2 - Corr. Route 5 - Corr. Route 2 - Corr. Route 5 - Corr.
Route 3 - Corr. Route 6 - Corr. Route 3 - Corr. Route 6 - Corr.

90 4 30 3
25
85 3.5
20 2.5
80 3 15
Relative humidity, Rh (%)

Temperature, T ( C)

10 2
o

75 2.5
Correction
Correction

5
70 2 0 1.5
-5
65 1.5
-10 1
60 1 -15
-20 0.5
55 0.5
-25
50 0 -30 0
0 4 8 12 16 20 0 4 8 12 16 20
Points Points

Fig. 9. Actual relative humidity and corresponding corrections for each interval Fig. 10. Actual temperatures and corresponding corrections for each interval
through the routes. through the routes.
 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026 2023

Route 1 - (c-w) Route 4 - (c-w) function of relative humidity, where this study used a linear func-
Route 2 - (c-w) tion against very few field data points.
Route 5 - (c-w)
Route 3 - (c-w) Route 6 - (c-w)
5. Numerical example
Route 1 - wf Route 4 - wf
Route 2 - wf Route 5 - wf A numerical example is presented to demonstrate the applica-
Route 3 - wf Route 6 - wf tion of the proposed model in assessment of the influence of differ-
ent atmospheric marine environments and the long-term corrosion
0.18 0.45
deterioration of the deck plating above the ballast tanks of any
crude oil tanker during her service life in some particular routes
0.16 0.4
through the Pacific Ocean. The different atmospheric characteris-
0.14 0.35 tics related to temperature and relative humidity of the Pacific
Correction & Weighting (c-w)

Ocean are taken into account based on long-term annually average

0.12 0.3 data prepared by NOAA/ESRL (see Fig. 8) [38]. Due to insufficient

Weight factor (wf)

0.10
0.08
0.06
0.04
0.02
0.00
0 4
Fig. 11. Weighting factors and combined correction and weighting.
correction factor to the chlorides ratio is the lowest, while the sen-
sitivity to the temperature ratio estimated based on T, Tn in K is the
highest.
Although, the wide scatter and the low fitting quality (low R2) of
using linear function to fit the corrosion rate data sets as a function
of chlorides and temperature, linear functions is shown to be the
most consistent function describing the available data. Grossman
[19] used linear functions to fit the relationships between the cor-
rosion rate data sets and the temperature as indicated earlier.
Gardiner and Melchers [28] used a linear function to model the
relationship between the corrosion rate and both of temperature
and salt deposition and supported this with data. As more and bet-
ter quality data become available, higher quality fitting can be
done and more accurate correction formulations can be obtained.
The same also consider fitting of the corrosion rate data set as a
data for the chloride distribution, the atmospheric chloride is as-
0.25 sumed to be constant over the atmosphere of the Pacific Ocean
and equals to the standard mean value as indicated in Table 3.
0.2 The atmospheric marine environment affects the coating effec-
tiveness life. Coating performance is not independent of the sur-
0.15
rounding environmental characteristics and can vary with it. The
dependence between the coating effectiveness life and the sur-
0.1
rounding environment is away from the area of this study. The
0.05 coating protection is assumed here to have the same performance
in the different routes. The nominal or standard conditions are as
0 indicated by Table 3. The nominal corrosion deterioration through
8 12 16 20 the marine atmosphere is described by Eqs. (1) and (2). The param-
Points eters of these equations are estimated through regression analysis
of data provided by ABS [36] for the corrosion degradation of the
deck plating above the ballast tanks as indicated in Table 1 [13].
The corrections corresponding to non-standard conditions of rela-
tive humidity, chlorides or temperature are performed using Eqs.
(4) and (5), where the correction factors are estimated using Eqs.
(9), (12) and (15) (see Fig. 6).
The models indicated by Eqs. (4) through (6) do not account for
the correlation that exists between the various factors. However,
when using the values of the various parameters that are appropri-
ate for each ocean area some correlation between those effects is
already accounted for.
These parameters have a seasonal variability but since the aim
of this study is to determine the long-term effect over many years,
the climatological data used does not refer to monthly averages but
to yearly values, which absorb the seasonal variability.
Six ship routes in the Pacific Ocean have been selected to repre-
sent different marine environments (see Fig. 7). Each route is as-
sumed to consist of a number of short intervals during which the
environmental conditions are considered to remain constant. The

1.6 2.2
Nominal (mm)
Long term corrosion depth, (mm)

1.4 Route (1)

2.1
Route (2)
Corrosion depth, d (t ) (mm)

1.2 Route (3)

Route (4) 2.0
1.0 Route (5)
Route (6) 1.9
0.8

0.6 1.8

0.4 1.7

0.2
1.6
Route (1)

Route (2)

Route (3)

Route (4)

Route (5)

Route (6)

Long term
Nominal

0.0
0 4 8 12 16 20 24 28 32
Time, t (years)

Fig. 12. Long-term corrosion wastage for each route (left) and long-term corrosion depth for the different routes in the Pacific Ocean (right).
2024 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026

number of intervals corresponding to each route is as indicated in
Fig. 7. The relative humidity and temperature are estimated for
each interval through each route and the corresponding correction
factors are calculated from Eqs. (9) and (15), respectively (see Figs.
9 and 10). Since the chloride is assumed constant and equal to the
standard value, no correction is needed.
The short-term prediction of the corrosion wastage correspond-
ing to each time interval was estimated using Eq. (4). The long-
term corrosion wastage in the ship’s lifetime dL(t), corresponding
to each route, is estimated as a weighted sum of the various
short-term corrosion wastage corresponding to each interval. In
each such interval, there is a particular combination of different
environmental conditions. Eq. (5) is used to carry out the different
weighting factors to account for the relative exposure to that par-
ticular combination through each interval. The weighting factors
are assumed equal for all the intervals in the same route, which
is a consequence of the vessel spending almost the same time in
each segment of the route (see Fig. 11).
The long-term corrosion wastage corresponding to each route
though the ship life is indicated in Fig. 12 (left). The long-term

4.0 5.0
Temperature 4.5 Temperature
3.5
Relative humidity Relative humidity
4.0
3.0 Chlorides Chlorides
3.5
2.5

Sensitivity
3.0
Sensitivity

2.0 2.5

1.5 2.0
1.5
1.0
1.0
0.5
0.5
0.0 0.0
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Time, t (years) Time, t (years)

Route (1) Route (2)

5.0 5.0
Temperature Temperature
4.5 4.5
Relative humidity Relative humidity
4.0 4.0
Chlorides Chlorides
3.5 3.5
Sensitivity

3.0 3.0
Sensitivity

2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Time, t (years) Time, t (years)
Route (3) Route (4)
6.0 6.0
Temperature Temperature
5.0 Relative humidity 5.0 Relative humidity
Chlorides Chlorides
4.0 4.0
Sensitivity
Sensitivity

3.0 3.0

2.0 2.0

1.0 1.0

0.0 0.0
0 4 8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Time, t (years) Time, t (years)
Route (5) Route (6)
Fig. 13. Time-dependent sensitivity for various parameters through the different routes.

1.4
1.2
Corrosion depth, d (t ) (mm)
1.0
0.8
0.6
0.4
0.2
0.0
0 4
Fig. 14. Long-term and nominal corrosion wastage as a function of time (left), and long-term time-dependent sensitivity for various parameters (right).
corrosion depth (d1) corresponding to each route is indicated in
Fig. 12 (right). The other parameters of the corrosion model (sc,
st) will be similar for all routes and equal to the nominal values
indicated in Table 1. From Fig. 12, the route 6 shows the highest
corrosion degradation, while the route 1 shows the lowest one.
Fig. 13 represents the sensitivity of the long-term prediction of cor-
rosion wastage to the different environmental factors for each
route. This sensitivity is calculated as given by Eq. (6). From
Fig. 13 it is observed that the corrosion degradation is very sensi-
tive to relative humidity than chloride and temperature while
the sensitivity to chloride and temperature is similar for all routes
through the ship service life.
Fig. 14 (left) shows the application of the new long-term time-
dependent model for the corrosion deterioration specific for mar-
ine atmosphere of the Pacific Ocean. It is noticed from the figure
that the long-term corrosion deterioration trend through the Paci-
fic Ocean atmosphere is higher than the standard trend which is
corresponding to average conditions through the world oceans
atmosphere. The model takes into account the sensitivity effects
to various environmental factors on the corrosion degradation for
various routes studied as short-term predictions. Fig. 14 (right)
shows the corresponding sensitivity to the different environmental
factors considered.
6. Conclusions
The dependence only on time of some of the existing corrosion
models applicable to ship structures cannot distinguish between
ships that are subjected to harder or to more benign corrosion
environments.
The model proposed in this paper extends one of the existing
models by adding three variables that reflect the relative level of
relative humidity, chlorides, and temperature, which is relevant
to the rates of corrosion to be expected in marine atmosphere.
The inclusion of these variables in the model will allow more accu-
rate predictions of the expected corrosion levels and therefore a
better planning of the corrosion inspections along the life of the
ships with the consequent significant savings.
It was demonstrated that corrosion in marine atmosphere is
primarily influenced by moisture and is accentuated by contami-
nants such as sodium chloride. Duration of exposure to wetness
is a critical variable. The chemical composition of the water film
is important. For the Pacific Ocean atmosphere and with tempera-
ture ratio calculated based on T, Tn expressed in centigrade, the
corrosion degradation correction shows very high sensitivity to
the relative humidity ratio than that sensitivity to any of the other
C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026 2025
5.0
Nominal Temperature
4.5
Long term Relative humidity
4.0
Chlorides
3.5
3.0
Sensitivity
2.5
2.0
1.5
1.0
0.5
0.0
8 12 16 20 24 28 32 0 4 8 12 16 20 24 28 32
Time, t (years) Time, t (years)
ratios while the sensitivity to air temperature ratio is the lowest.
With temperature ratio calculated based on T, Tn expressed in Kel-
vin, the sensitivity to air temperature ratio is the highest, while the
sensitivity to the chlorides ratio is the lowest.
The example has shown how it is possible to quantify changes
in corrosion as a function of trading routes, by accumulating the
damage effects and the combined effects of the different environ-
mental variables though each route.
The proposed equations serve also as a guide to ship owners
and Classification Societies about which variables need to be mon-
itored to allow more accurate predictions of corrosion wastage in
marine atmosphere. It is necessary that monitoring programs are
put in place so as to produce the required data for validation of
the proposed model in the long-term. In fact the availability of
such data may allow some adjustment of the proposed equations
for the effect of each of the parameters.
Acknowledgement
This work has been funded by American Bureau of Shipping un-
der the project ‘‘Condition Assessment of Ageing Ship Structures”.
References
[1] C. Guedes Soares, Y. Garbatov, Reliability of maintained ship hulls subjected to
corrosion, Journal of Ship Research 40 (3) (1996) 235–243.
[2] P.H. Wirsching, J. Ferensic, A. Thayamballi, Reliability with respect to ultimate
strength of a corroding ship hull, Marine Structures 10 (7) (1998) 501–518.
[3] M.A. Shama, H.W. Leheta, Y. Abd El Nasser, A.S. Zayed, Reliability of double hull
tanker plates subjected to different loads with corrosion effects, Alexandria
Engineering Journal 41 (4) (2002) 587–597.
[4] IACS (International Association of Classification Societies), Unified
Requirements Group Z, International Association of Classification Societies,
2005.
[5] R.E. Melchers, Probabilistic models for corrosion in structural reliability
assessment – Part 2: models based on mechanics, Journal of Offshore
Mechanics and Arctic Engineering 125 (4) (2003) 272–280.
[6] TSCF (Tanker Structure Co-operative Forum), Condition Evaluation and
Maintenance of Tanker Structures, Witherby & Co., London, 1992.
[7] N. Yamamoto, K. Ikegami, A study on the degradation of coating and corrosion
of ship’s hull based on the probabilistic approach, Journal of Offshore
Mechanics and Arctic Engineering, ASME 120 (3) (1998) 121–128.
[8] G. Wang, J. Spencer, H. Sun, Assessment of corrosion risks to aging ships using
an experience database, in: Proceeding of 22nd International Conference on
Offshore Mechanics and Arctic Engineering, Cancun, Mexico, ASME, Paper
OMAE2003-37299, 2003.
[9] C. Guedes Soares, Y. Garbatov, Reliability of maintained, corrosion protected
plates subjected to non-linear corrosion and compressive loads, Marine
Structures 12 (6) (1999) 425–445.
[10] C. Guedes Soares, Y. Garbatov, Non-linear time dependent model of corrosion
for the reliability assessment of maintained structural component, in: S.
Lydersen, G. Hansen, H. Sandtorv (Eds.), Safety and Reliability, vol. 2, Balkema,
1998, pp. 929–936.
2026 C. Guedes Soares et al. / Corrosion Science 51 (2009) 2014–2026
[11] J.K. Paik, S.K. Kim, S. Lee, Y.E. Park, A probabilistic corrosion rate estimation
model for longitudinal strength members of bulk carriers, Journal of Ship and
Ocean Technology 2 (1) (1998) 58–70.
[12] S. Qin, W. Cui, Effect of corrosion models on the time-dependent reliability of
steel plated elements, Marine Structures 16 (1) (2003) 15–34.
[13] Y. Garbatov, C. Guedes Soares, G. Wang, Non-linear time dependent corrosion
wastage of deck plates of ballast and cargo tanks of tankers, Journal of Offshore
Mechanics and Arctic Engineering 129 (1) (2007) 48–55.
[14] Y. Garbatov, C. Guedes Soares, Corrosion wastage modeling of deteriorated
bulk carrier decks, International Shipbuilding Progress 55 (1–2) (2008) 109–
125.
[15] A. Zayed, Y. Garbatov, C. Guedes Soares, G. Wang, Environmental factors
affecting the time dependent corrosion wastage of marine structures, in: C.
Guedes Soares, Y. Garbatov, N. Fonseca (Eds.), Maritime Transportation and
Exploitation of Ocean and Coastal Resources, vol. 1, Taylor & Francis, 2005, pp.
589–598.
[16] M. Panayotova, Y. Garbatov, C. Guedes Soares, Factor influencing atmospheric
corrosion and corrosion in closed spaces of marine steel structures, in:
Proceedings of the 7th International Conference on Marine Science and
Technology (Black Sea 2004), Union of Scientists of Varna, Varna, Bulgaria,
2004, pp. 286–292.
[17] H.R. Ambler, A.A. Bain, Corrosion of metals in the Tropics, Journal of Applied
Chemistry 5 (1955) 437–467.
[18] A.H. Ailor, Atmospheric Corrosion, John Wiley & Sons, 1982.
[19] P.R. Grossman, Investigation of atmospheric exposure factors that determine
time of wetness of outdoor structures, in: S.K. Coburn (Ed.), Atmospheric
Factors Affecting the Corrosion of Engineering Materials, ASTM STP 646,
American Society for Testing and Materials, Philadelphia, 1978, pp. 5–16.
[20] W.J. Vernon, A laboratory study of the atmospheric corrosion of metals,
Transactions of Faraday Society 31 (1) (1935) 1668.
[21] A.D. Mercer, E.A. Lumbard, Corrosion of Mild Steel in Water, British Corrosion
Journal 30 (1) (1995) 43–55.
[22] F.L. LaQue, Marine Corrosion. Causes and Prevention, Wiley-Inter science,
1975.
[23] S.C. Dexter, Seawater, in: J.R. Davis, K.M. Mills, S.R. Lampman, T.B. Zorc (Eds.),
Metals Handbook, vol. 13 – Corrosion, ninth ed., ASM International Handbook
Committee, 1989.
[24] R.E. Melchers, Effect of immersion depth on marine corrosion of mild steel,
Corrosion (NACE) 61 (9) (2005) 895–906.
[25] R.E. Melchers, Effect of temperature on the marine immersion corrosion of
carbon steels, Corrosion (NACE) 58 (9) (2002) 768–782.
[26] F.L. LaQue, Behavior of metals and alloys in seawater, in: H.H. Uhlig (Ed.),
Corrosion Handbook, John Wiley & Sons, New York, 1948.
[27] R.E. Melchers, R. Jeffrey, Influence of water velocity on marine immersion
corrosion of mild steel, Corrosion (NACE) 60 (1) (2004) 84–94.
[28] C.P. Gardiner, R.E. Melchers, Enclosed atmospheric corrosion in ship spaces,
British Corrosion Journal 36 (4) (2001) 272–276.
[29] D.T. Yasunaga, K. Katoh, S. Imai, H. Miyuki, Y. Yamane, H. Ohyabu, M. Saito, M.
Yoshikawa, Y. Kobayashi, Y. Tomita, Study on Cargo Oil Tank Upper Deck
Corrosion of Oil Tanker, Transactions of the Society of Naval Architects and
Marine Engineers, vol. 3, 2003.
[30] K. Katoh, S. Imai, D.T. Yasunaga, H. Miyuki, Y. Yamane, H. Ohyabu, Y.
Kobayashi, M. Yoshikawa, Y. Tomita, Study on Localized Corrosion on Cargo
Oil Tank Bottom Plate of Oil Tanker, Transactions of the Society of Naval
Architects and Marine Engineers, vol. 3, 2003.
[31] C. Guedes Soares, Y. Garbatov, A. Zayed, G. Wang, Non-linear corrosion model
for immersed steel plates accounting for environmental factors, Transactions
of the Society of Naval Architects and Marine Engineers 113 (2005) 306–
322.
[32] C. Guedes Soares, Y. Garbatov, A. Zayed, G. Wang, Corrosion wastage model for
ship crude oil tanks, Corrosion Science 50 (11) (2008) 3095–3106.
[33] L.L. Shreir, Corrosion, vol. 1, John Wiley & Sons, 1963.
[34] J.K. Paik, F. Brennan, C.A. Carlsen, C. Daley, Y. Garbatov, L. Ivanov, C.M. Rizzo,
B.C. Simonsen, N. Yamamoto, H.Z. Zhuang, Condition Assessment of Aged
Ships, 16th International Ship and Offshore Structures Congress, ISSC-2006,
vol. 2, Committee V.6, Southampton, UK, 2006. pp. 265–315.
[35] R. Vera, B.M. Rosales, C. Tapia, Effect of the Exposure Angle in the Corrosion
Rate of Plain Carbon Steel in a Marine Atmosphere, Corrosion Science, 2003,
pp. 321–337.
[36] ABS, ABS Database of Corrosion Wastage for Oil Tankers, American Bureau of
Shipping, ABS RD 2002-07, 2002.
[37] D.C. Cook, A.C. Van Orden, J.J. Carpio, S.J. Oh, Atmospheric corrosion in the Gulf
of México, Hyperfine Interactions 113 (1–4) (1998) 319–329.
[38] NOAA/ESRL, National Oceanic & Atmospheric administration (US Department
of Commerce)/Earth System Research Laboratory (Physical Sciences Division),
2007. Available from: <http://www.noaa.gov/>.
[39] S.W. Dean, E.C. Rhea, Atmospheric Corrosion of Metals, American Society for
Testing and Materials, 1982.
[40] L.C. Herring, A.N. Titcomb, Investigation of Internal Corrosion and Corrosion
Control Alternatives in Commercial Tank Ships, Ship Structure Committee,
SSC-312, 1981.
[41] P. Caridis, Inspection, Repair and Maintenance of Ship Structures, first ed.,
Witherby, London, 2001.
[42] H.H. Uhlig, R. Revie, Corrosion and Corrosion Control, third ed., Wiley-Inter
science, USA, 1985.
[43] S. Hoerle, F. Mazaudier, P. Dillmann, G. Santarini, Advances in understanding
atmospheric corrosion of iron. II. Mechanistic modeling of wet–dry cycles,
Corrosion Science 46 (6) (2004) 1431–1465.
[44] K. Chandler, Marine and Offshore Corrosion, Butterworths, 1985.
[45] J.R. Davis, J.D. Destefani, H.F. Frissell, G.M. Crankovic, D.M. Jenkins, Metals
Handbook, vol. 13 – Corrosion, ninth ed., ASM International Handbook
Committee, 1987.
[46] T.E. Graedel, R.P. Frankenthal, Corrosion mechanisms for iron and low alloy
steels exposed to the atmosphere, Journal of the Electrochemical Society 137
(8) (1990) 2385–2394.
[47] S.L. Pohlman, General corrosion, in: J.R. Davis, J.D. Destefani, H.J. Frissell, G.M.
Crankovic, D.M. Jenkins (Eds.), Metals Handbook, vol. 13 – Corrosion, ninth ed.,
ASM International Handbook Committee, 1987.