Beruflich Dokumente
Kultur Dokumente
Corrosion Science
journal homepage: www.elsevier.com/locate/corsci
a r t i c l e i n f o a b s t r a c t
Article history: Life extension of ageing steel structures such as naval ships requires consideration of plate thickness
Received 4 January 2008 losses due to corrosion, particularly when protective measures such as paint coatings and sacrificial pro-
Accepted 20 August 2008 tection are not entirely effective. Traditionally, corrosion prediction models for ships take no account of
Available online 12 September 2008
the operational profile. Consequently, the corrosion models have an inherently high variability with poor
corrosion prediction capability. A new corrosion model for the prediction of corrosion loss in the seawater
Keywords: ballast tanks of naval vessels has been developed. The model incorporates previously available corrosion
A. Mild steel
models for immersion corrosion and atmospheric corrosion and takes account of operational and envi-
B. Modelling studies
B. Weight loss
ronmental variables. Experimental validation is presented for a trial on an operational naval vessel.
C. Atmospheric corrosion Ó 2008 Published by Elsevier Ltd.
C. Microbiological corrosion
Seawater ballast tanks are of particular interest, since they often historical data and on data collected from many ships operating
have the highest rates of corrosion due to high internal humidity under quite varied operational conditions, they have very limited
and the difficulty of access for maintenance. In order to validate capability to be used in predictive studies for any one particular
the proposed corrosion model, an experimental trial was con- vessel.
ducted using unpainted mild steel coupons exposed inside the sea- A different approach to corrosion loss modelling is to base the
water ballast tanks of an operational naval ship. Herein results are modelling effort to a greater or lesser extent on the actual pro-
presented for the modelling parameters involved as compared cesses involved in the corrosion process [20,21]. This means that
with observed coupon corrosion loss and measured environmental models must take some levels of account of the various factors
conditions. known to be important in the corrosion of steel in natural seawa-
ters, including oxygen concentration [22], marine growth and bac-
2. Models for estimating corrosion of steel structures teriological effects [23], pit formation and growth [24] and
diffusion of reactants through the rust layer [25–27]. A model that
The earliest corrosion models were developed entirely from has been developed along these lines is now available [20]. It has
empirical data sourced from thickness measurement surveys. Most mean seawater temperature as the main parameter [28] and has
data relates to bulk carriers and tankers and was obtained from been calibrated to corrosion data obtained for mild and low alloy
many ships and for various locations within ships [7]. Impetus steel coupons exposed to near-surface seawater conditions. Tem-
for this work may have been added by concerns over the ageing perature is particularly relevant for ships since the effect of opera-
of ships, particularly of bulk carriers, and the numerous marine tion at elevated temperatures has been implicated in a number of
accidents resulting in structural failure and serious environmental catastrophic failures of bulk carrier ships [29]. This model forms
consequences [8]. Because of the criticality of these matters, ship the main part of the model described herein for ballast tank
corrosion models increasingly are required to be of high quality corrosion.
and to produce accurate prediction of likely corrosion loss and its The other form of corrosion relevant to ship structures and to
extent [9]. ballast tanks is marine atmospheric corrosion. The model proposed
Earlier corrosion loss models presented corrosion loss in terms by Gardiner and Melchers [30] accounts for the chemical and phys-
of a constant ‘corrosion rate’, implying a constant rate of plate ical processes of ship corrosion in an enclosed atmospheric envi-
thickness reduction over time. The better efforts also included an ronment and, as is usual for atmospheric corrosion, depends on
estimate of the nominal or calculated variation [5,10]. This allowed temperature, time of wetness and salt deposition, but also on the
these models to be used in structural reliability analyses [11–14] voyage time [31]. This model forms the second part of the model
although typically these models exhibited a high degree of vari- for ballast tank corrosion.
ability and thus tended to produce poor estimates of the lifetime In developing corrosion loss models account must be taken of
risk of failure. the operational profile of the ship, as highlighted by Gardiner
A more refined approach, still empirically based, took into ac- and Melchers [3]. They considered the topside ballast tanks, double
count the physical phenomena of the corrosion process. Three bottom ballast tanks and cargo spaces in bulk carriers. Factors in-
stages were considered: (1) the effective life of the coating before clude tank status, effectiveness of protective coatings and presence
corrosion begins, (2) the generation of initial corrosion pitting or otherwise of sacrificial anodes. It is considered that a similar ap-
points, and (3) the subsequent progress of corrosion [15,16]. In proach is required for the ballast tanks in naval vessels.
some cases the latter was described by a non-linear function such
as
3. Models for immersion and enclosed atmospheric corrosion
dðtÞ ¼ a t b ð1Þ
3.1. Immersion corrosion
where d is the depth of pitting or corrosion loss as a function of time
t and both coefficients a and b are obtained from empirical fits to Immersion corrosion occurs when the seawater ballast tanks
the data. Sometimes they have been presented as random variables are fully ballasted, i.e. fully filled with seawater. Assuming a high
with probability distributions obtained to obtain a good fit to the degree of loss of coating protection, the corrosion process may be
empirical data. Corrosion models that account for the physical pro- represented using the phenomenological marine immersion corro-
cesses of corrosion, for paint life estimates and for a gradual break- sion model proposed by Melchers [20]. The mean value corrosion
down of coatings before the onset of non-linear corrosion also were loss model as a function of exposure time, and the parameters
proposed [14,17,18]. For example, the progression of corrosion was describing the model are shown Fig. 1.
proposed in the form [17] The immersion corrosion model consists of five phases [32] that
describe the phenomenon of corrosion progression. Phase 0 con-
dðtÞ ¼ d1 ð1 et=s tÞ ð2Þ
sists of very short-term kinetic and bacterial influences that may
where st is the transition time, d(t) is the thickness of corrosion be ignored in practical applications. Phase 1 represents the situa-
wastage at time t and d1 is the long-term thickness of corrosion tion when the rate of corrosion is controlled by the diffusion of
wastage. oxygen from the seawater. It may be modelled, closely, as a
In all cases the corrosion models require calibration to empirical straight line, implying a constant corrosion rate (r0). Phase 2 is
corrosion loss data to determine the values of parameters in the non-linear and controlled by the rate of oxygen diffusion through
model. For example, data for annual corrosion rates used by the (increasing thickness of the) rust layers, assumed, theoretically,
Yamamoto and Ikegami [15] were sourced from 27 very large bulk to be have uniform diffusion properties. Evidently, there is a
carriers with 7581 data points. The coefficient of variation (COV) smooth transition from Phase 1 to Phase 2. Phase 2 will proceed
ranged from 0.51 to 0.74 [3]. Paik et al. [14,19] applied data from until point AP is reached, at which time anaerobic conditions begin
7503 data points from 44 bulk carriers with a minimum COV of to occur under the rust layers next to the corroding steel. In prac-
0.66, however, most were above 1.0. These studies represent a tice there is likely to be a smooth transition around AP as not all
thorough presentation of corrosion loss data, however, due to the points on a corroding surface are precisely at the same stage [20].
large variations it has limited value for prediction and assessment Phase 3 is the theoretical start of predominantly anaerobic con-
of remaining life [3]. Moreover, since these models are based on trolled corrosion, associated with the metabolic activity of sulphate
3298 M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307
Table 1
Zero mean uncertainty estimates
Table 2
Values of constants for enclosed atmospheric corrosion model [30]
Study k1 k2 k3
1 0.0474 0.00611 0.00629
2 0.688 0.00359 0.028
3 0.146 0.00439 0.00663
4 0.127 0.00787 0.00953
Fig. 1. Marine immersion corrosion model for mild steel for near-surface ‘at sea’ 5 0.08 0.0014 0.00654
conditions [20]. 6 0.005 0.00241 0.00568
Fig. 2. Operational simulation program flowchart (based on Gardiner and Melchers [3]).
Fig. 3. Monthly average sea surface temperatures [35] as applied with the sea areas defined by British Maritime Technology [36].
Fig. 5. Simulated three-year corrosion predictions showing daily and cumulative corrosion loss for operations on the eastern Australian seaboard for (a) 100% immersion
corrosion, (b) 100% enclosed atmospheric corrosion and (c) a mixture of immersion and enclosed atmospheric corrosion based on simulated ballast tank conditions. Sea
surface temperatures (d) for this operational region are also shown and cumulative corrosion loss outputs also exhibit error bars representing variability of one standard
deviation.
Table 3
from Table 2 were used – see Table 3. It should be noted that these Values of constants used in the enclosed atmospheric corrosion model
values were derived from published experimental results with
k1 k2 k3
exposure times of not more than one year [30]. Typical ship ballast
tanks may corrode for longer than this since the minimum time be- Average value from Table 2 0.10523 0.004295 0.010445
tween inspection and repair for navy vessels typically is about 1.5
years.
Since the relative humidity in ballast tanks is typically above
70% for most of the time [30], it was assumed, conservatively, that of steel for 80% of the time. It was also assumed that the ship will
the wetting fraction (WF) was 0.8. This means that there will be be on a continued voyage, so that the voyage time becomes the
sufficient moisture for corrosion to proceed on the exposed areas time of exposure. Thus Eq. (7) simplifies to
M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307 3301
Corrosion lossðmmÞ ¼ 0:8tðk1 þ k2 ½Salt þ k3 TÞ ð11Þ It is highly liked that the frequency of ballasting would affect
the interaction of the two models, however, at present they are
where t is the exposure time in years.
seen as mutually exclusive. In practice, it is likely that there will
Finally, as a first approximation a salt concentration of 10 ppm
be some interaction between the corrosion losses in the ballasted
was used for all predictions. Using this value together with the
and the un-ballasted conditions, and this should be reflected in
other inputs produced an one-year corrosion rate of 0.12 mm/year
the models used. This has not been considered in the present work.
at 20 °C. Interestingly, this compares favourably with empirical
Also, in practice naval ships are not operated in one particular geo-
values given for tanker ships [37]. To estimate the importance of
graphical sea area as assumed above. Moreover, there will be var-
the salt concentration level, a sensitivity analysis was carried out.
iation in the operations and this will affect the ballasting profile. In
This showed that the one-year corrosion rate at 20 °C increased
order to obtain an estimate of these influences, a set of realistic
to 0.25 mm/year when the salt concentration was increased to
data for the geographical area of operation was applied. The data
50 ppm.
was derived from navy fleet activity schedules. Fig. 6 shows the
Fig. 5 also shows the daily and cumulative corrosion losses for
SST data using known geographical locations of the ship and
the simulated ballast conditions, i.e. using tank status data
Fig. 7 shows the resulting cumulative corrosion losses using these
(Fig. 4) and using outputs for both immersion and atmospheric cor-
temperature inputs.
rosion. The daily corrosion loss curve is seen to have two parts. The
first part represents the (approximately sinusoidal) change as a re-
sult of the changing seasonal temperatures as relevant for the en- 6. Example 2 – lifetime corrosion loss scenario
closed atmospheric corrosion model. The second part consists of
the spikes. These are for immersion corrosion under the ballast The objective of the ballast tank corrosion model lies in the pre-
condition and are the result of the changing phases with time in diction of how much corrosion is likely to occur in unprotected re-
the immersion corrosion model. It might be noted that the change gions of the seawater ballast tanks and in using that information to
in ballast condition does not affect the progression of the phase optimise protection and repair measures. An alternate application
transitions in the immersion corrosion model. is to use the prediction of corrosion likely to occur as a function of
operational conditions such that the operational profile of the ship
can be managed to achieve or exceed a desired minimum lifetime.
Fig. 8 shows the simulated corrosion loss outcomes for a ship oper-
ated in colder waters early in its life so as to reduce early corrosion,
followed by higher corrosion losses later, as it is operated in war-
mer waters.
7. Experimental validation
Fig. 7. Cumulative corrosion loss in sea water ballast tanks using actual geograph-
ical locations of a naval vessel. Fig. 8. Lifetime analysis using ballast tank corrosion model.
3302 M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307
7.1. Method and installation humidity and temperature within the tank were recorded through-
out the trial period.
A total of 64 coupons of mild steel composition similar to that of A series of four exposure periods with coupon recoveries at
the ship’s plating were sourced from the shipyard where the ship three months period were planned. At each time point the data
was built. The coupons were laser cut to a size 100 50 mm from was downloaded from the data logger, its batteries recharged
a 4 mm thickness D36 grade steel plate. The coupons edges were and one set of coupons recovered. Vessel operations dictated the
de-burred and holes were pattern-drilled for identification actual recovery time points. The first set of 16 coupons was recov-
purposes. ered after 94 days. Further sets of 16 coupons were recovered at
The coupons were prepared for exposure by sand-blast clean- 211, 294 and 541 days of exposure. The last set of coupons could
ing, measuring for size (length, width and thickness to 0.01 mm) not be collected close to the planned 12 months period due to
and weighing to the nearest 0.0001 g. The coupons were attached the extended operation of the ship during this time. Fig. 10 shows
to a test pod made from 150 mm diameter standard PVC pipe using the external condition of coupons after each of the exposure
a nylon nut and bolt and spaced from the pod using a short length periods.
of vinyl tubing. Two test pods were used, each suspended by ropes
from the internal stiffeners within the tank and sufficiently far 7.2. Results and observations
from plating and stiffeners so as to reduce the likelihood of damage
to the hull structure during ship operations (Fig. 9). After collection the coupons were cleaned and weighed accord-
A snorkel-like sensor unit was constructed to house the individ- ing to well-established corrosion testing principles. Cleaning was
ual sensors for air temperature, relative humidity and water level. performed using Clarke’s solution according to designation C.3.1
It was placed in the upper space of one of the compartments of the of ASTM G1 (1994) [39]. A summary of corrosion loss measure-
ballast tank used for the trial (Fig. 9). The temperature sensor uti- ments is shown in Table 4.
lised was a Type K thermocouple (chromel and alumel). The ballast Fig. 11 shows the recorded temperature, humidity and ballast
condition of the tank was recorded using a polysulphide drop float condition over the complete period of the trial. Although consider-
switch and the relative humidity was measured using a Vaisala able effort was made to protect the humidity sensor seawater
50Y HUMITTER Humidity and Temperature probe fitted with an splashing on to the capacitor eventually rendered it non-func-
INTERCAP capacitor-type humidity sensor. The humidity sensor tional. As noted, this was not unexpected. It is evident from the
has an operating range 0–100% RH (±2% RH at 20 °C) with a 0– trace shown in Fig. 11 that the effectiveness of the sensor de-
1 V DC output [38]. Because humidity sensors are known not to re- creased some time after the first ballasting period and that this be-
act well to a marine environment, it was not expected that the life came more serious subsequently. Hence the later humidity data
of the humidity sensor would be very long but might at least sur- should be disregarded.
vive a few weeks and thereby provide some useful information in
the early stages of the trial. 7.3. Comparison to model predictions
The moisture proof data logger box (360 200 150 mm) con-
tained a Datataker DT50 low voltage data acquisition unit with The results from the shipboard trial were compared with the
power supplied via re-chargeable 6 V battery. A 5–12 V DC con- output from the corrosion model. Fig. 12 shows a comparison be-
verter was also required in order to power the humidity sensor tween the predicted sea surface temperature (as used in the corro-
during data readings. The data logger box was installed in the com- sion model) and the shipboard temperature recorded inside the
partment immediately above the seawater ballast tank. Connec- ballast tank. Over the total experimental period, the (absolute)
tions to the sensor unit required a cable penetration through the mean daily difference between the two data sets is 1.3 °C, with
deck head. In order to minimize interference with the structure the overall mean difference being around 0.05 °C.
of the ship, a special temporary hatch cover with a through-deck Both temperature data sets, i.e. predicted sea surface tempera-
fitting was manufactured for this purpose. The ballast level, ture and the ballast tank observed temperature, were used in the
Fig. 9. Schematic cross section of the ballast tanks in the bow region showing installation of experimental equipment.
M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307 3303
Fig. 10. Test pods in-situ showing exteriors of coupons after exposure periods of (a) 94 days, (b) 211 days, (c) 294 days and (d) 541 days.
model to estimate the corrosion loss as a function of exposure variability from the ballast tank corrosion model at the times cor-
time. The estimates are shown in Fig. 13. Also shown in Fig. 13 responding to that of the observed corrosion loss.
are the four average corrosion losses determined from the corro- The observed corrosion losses and the components of the corro-
sion coupons. The error bars represent one standard deviation of sion model also can be compared, assuming these were acting
3304 M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307
Table 4
Average coupon thickness reduction and coefficients of variation
Fig. 13. Comparison of shipboard trial and simulated corrosion losses. Simulated
results use the ballast tank corrosion model with inputs of either the shipboard trial
water temperature or published sea surface temperature (with error bars of one
standard deviation).
Fig. 11. Environmental parameters recorded from sensors in seawater ballast tank.
Fig. 14. Comparison of shipboard trial and simulated corrosion losses with similar
output for the immersion model and atmospheric model acting alone (i.e. fully
ballasted or fully de-ballasted predictions). Also shown are the observed corrosion
losses, at corresponding times the variability of one standard deviation are shown
for the combined simulated outcome.
Fig. 12. Ballast tank temperature from the Type K thermocouple used during the the vessel in its operations. It follows that accurate predictions
shipboard trial and compared with the predicted sea surface temperature based on are desirable and therefore the assumptions made in the models
geographical area and date of ship operations. are of interest.
It was assumed that the marine immersion corrosion model
developed for coupon data is applicable to immersion corrosion
alone. Only the case of recorded tank temperatures is considered in seawater ballast tanks. Further work may be needed to confirm
here. The possibilities are (i) immersion only, corresponding to this. The model assumes ‘at sea’ conditions that are defined as fully
the ballast tank being continually filled and (ii) atmospheric, corre- oxygenated and for unpolluted fresh seawater with typical coastal
sponding to the ballast tank being continually empty but humid. seawater bacteriological content [28]. It needs to be established
Fig. 14 shows the results together with the combination model whether this is a valid assumption for typical ballast waters in na-
from Fig. 13 and the data points for the actually observed corrosion val vessels. For example, it would be expected that a ‘freshly’ bal-
losses. lasted tank would contain fully oxygenated seawater, however,
within the enclosed ballast tank environment this may be depleted
with time as a result of on-going corrosion. This could induce the
8. Discussion early onset of anaerobic bacteriological activity and hence corro-
sion. It is also assumed that the bacteriological component of sea-
Although a very small part of a naval vessel, the ballast tanks water would be similar to that experienced inside the tanks.
are known to dominate maintenance activity due to the highly However, due to the lack of natural sunlight within the tank of
aggressive internal environment and the criticality of this part of the bacteriological action could deviate from ‘at sea’ conditions.
M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307 3305
Further, the level of pollution within ballast tanks, even for naval cargo (e.g. iron ore or coal) higher in temperature than ambient.
vessels, could differ from ‘at sea’ conditions. This may affect the temperature inside nearby ballast tanks.
The major assumption herein regarding the application of the The float switch used to monitor the ballast condition during
atmospheric corrosion model was regarding the level of surface the shipboard trial provided two pieces of information, (i) the spe-
salt concentration. In the examples given above, experimentally cific time when the tank is ballasted and (ii) the overall ratio of
observed data was not used but generic data was employed. An time in the ballasted and de-ballasted condition. The first was used
example is the nominal salt concentration of 10 ppm used herein. to determine which of the two corrosion loss models (immersion
Evidently, the atmospheric corrosion loss estimation may be im- or atmospheric) to apply to estimate the daily corrosion losses.
proved by using experimentally determined surface salt concentra- For long-term corrosion loss prediction, however, the exact
tions. The modelling procedure, of course, is not affected. sequencing and timing of ballasting is unlikely to be known in de-
Under advanced corrosion conditions there is an apparent tail and the expected ratio of the ballasted to de-ballasted status of
anomaly between the immersion and the atmospheric corrosion the tank is more useful. In the present investigation the tank was
models when ballast conditions change. The immersion model ballasted for about 55 days during the trial, i.e. about 10% of the to-
may be in the anaerobic range but the atmospheric corrosion mod- tal time. Moreover, this may not be typical as the ballast tank
el used herein does not recognize such a state. This is a conse- instrumented was the largest on the ship and is usually the last
quence of the nature of the atmospheric corrosion model to be ballasted with seawater as significant amounts of fuel are
adopted. However, recent research has shown that anaerobic con- being consumed. It can be expected that for other ballast tanks
ditions are also likely to exist under advanced atmospheric corro- and for other vessels the ballasting ratio would be different. The ra-
sion conditions [40]. Moreover, it is known that alternating tio also would depend on the size of the tank and the type of ship.
oxidation and anaerobic conditions can exist and also that they It follows that where longer-term corrosion loss prediction is of
can cause high rates of corrosion. Evidently, there remains room interest, the exact ballast condition is unlikely to be needed, but
for improvement of the models to account for these factors. rather the ballasting to de-ballasting ratio would be used in mod-
Both the immersion and atmospheric models are based on data elling. In this regard, the validity of applying the immersion and
from coupon weight losses. Concern has been expressed about the the atmospheric corrosion models separately to account for the
validity of using these for continuous structures but the available dry–wet cycles has been assumed throughout as sufficiently accu-
evidence is that this is not an issue except for highly advanced rate. This aspect requires further research although the indications
states of corrosion when fatigue and stress issues start to affect from the modelling to date suggests that it us unlikely to be a ma-
corrosion [28]. These effects are not likely to be significant for jor issue.
well-maintained vessels with relatively low levels of corrosion Relative humidity, measured during the trial, is a parameter
loss. known to be important for estimating atmospheric corrosion loss.
The experimental trial extended over 541 days and produced a For interior ship spaces it has been estimated that above about 90%
total corrosion loss of 0.242 mm. This corresponds to an average RH the moisture content is sufficiently high for wetting of the sur-
corrosion rate of 0.163 mm/year. It compares reasonably well with face to occur and for wet corrosion to proceed [30]. The experimen-
the range of rates of general corrosion in the literature. For various tal observations (Fig. 11) showed that early in the trial the relative
structural details in crude oil tankers, they are in the range 0.08– humidity was close to 100% and then declined but stayed around
0.21 mm/year [12]. For ballast tanks, corrosion rates of between 70–75% for most of the trial. As noted, the later results must be
0.046 and 0.289 mm/year have been reported [6] with likely corro- treated with caution owing to the loss of calibration as the humid-
sion loss allowance rates in ships for segregated ballast tanks of be- ity sensor became affected by prolonged exposure to the saline
tween 0.04 and 0.10 mm/year for bottom shell plating and environment.
between 0.20 and 1.20 mm/year for longitudinal bulkhead web Whether coupons are realistic for representing the actual corro-
stiffeners [41]. The comparison between these values and those ob- sion losses inside ballast tanks also is of interest. There are at least
served for the naval vessel provides a reasonable degree of confi- three issues: (i) the effect of dynamic loading on corrosion loss, (ii)
dence that the observations are realistic. the effect of structural orientation and (iii) the effect of corrosion
As noted, temperature, the main environmental variable of protection measures.
interest, affects the progression of corrosion with time. For predic- Regarding the first, the corrosion coupons were exposed in the
tions using the model a database of geographically based sea sur- ballast tank as shown in Fig. 9, with the only possible external
face temperatures (SST) was used. However, more accurate loading on them being due to sloshing forces of seawater in the
temperature data obtained in-situ could be used, and it is of interest tank. In contrast, structural details that make up a ballast tank
to compare predictions based on generic temperature data and on would experience dynamic loading with associated flexing as a re-
measured data. Fig. 12 shows both the SST and the actual ballast sult of movement of the ship through the seaway. It is possible that
tank temperature recorded from the Type K thermocouple. Clearly the dynamic flexing could accelerate the corrosion process due to
there is a good agreement, adding confidence to the assumption in some, or all, of the corrosion product spalling from the metal sub-
the model that SST can be used to approximate the ballast tank strate. However, this is likely to be the case only when there is se-
temperature. Over the duration of the test, there is surprisingly lit- vere corrosion loss and strains are very high since it is known that
tle difference between the data sets. Some differences might have moderate strains have negligible effect on corrosion loss [25].
been expected since the SST data set combines temperatures spread In-situ plate thickness measurements (e.g. ultrasonics) are often
over large areas of ocean and then has these combined to produce used to estimate thickness losses but the resolution obtainable,
monthly averages. Evidently, even closer correspondence might while acceptable for period classification assessments, is inade-
be possible if SST data of a higher resolution were to be used, how- quate for on-going monitoring and for development of corrosion
ever, the results shown herein indicate that greater resolution is not loss models. Actual attachment of coupons to the surface of the
warranted for modelling purposes. The results indicate that with- ship structure so as to allow transfer of strain and hence stress to
out access to actual environmental data the use of the SST database the coupon might be possible in principle and has been attempted
is a rational means to predict temperatures in the ballast tanks of a in other studies, e.g. [42], though in most cases the coupling cannot
naval ship giving the geographical route of the vessel is known. directly transfer the strains of the ship due to flexibility in the cou-
However, this may not be correct in the case of commercial bulk pling. In the present study operational constraints did not permit
carriers with single or double hulls particularly if the holds contain this approach.
3306 M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307
For naval ships severe levels of rust leading to spalling is unli- Acknowledgments
kely to occur due to the maintenance scheme of these structures.
It may be concluded, therefore, that the use of coupons for estimat- The work reported herein was a collaboration between the Uni-
ing corrosion loss as used in the present shipboard trial is suffi- versity of Newcastle, Australia and the Defence Science and Tech-
ciently accurate to represent actual corrosion losses for ship nology Organisation, Department of Defence, Australia. The
structural details. authors appreciated the support provided by the Royal Australian
Much more conservative estimates of corrosion loss can, of Navy in permitting the shipboard field trial in the ballast tanks
course, be obtained from the corrosion loss models if only the early aboard an operational naval ship.
rates (r0) are considered. Both the corrosion loss models incorpo-
rated in the ballast tank corrosion model consider the initial rate References
of corrosion loss corresponding to the situation with little or no
corrosion product build-up. Thus Phase 1 of the immersion corro- [1] R.E. Melchers, The effect of corrosion on the structural reliability of steel
offshore structures, Corrosion Science 47 (10) (2005) 2391–2410.
sion model may be used to estimate the governing corrosion rate if [2] C. Gardiner, R.E. Melchers, Operational parameters affecting bulk carrier
rust layers are continually removed. The atmospheric corrosion corrosion, in: Proceedings of AME’98 Meeting the Needs of Industry, AME-CRC,
model already is based on empirical corrosion loss data in the first 1998.
[3] C.P. Gardiner, R.E. Melchers, Corrosion analysis of bulk carriers – Part I:
year of corrosion, i.e. before significant build up of rust layers. Operational parameters influencing corrosion rates, Marine Structures 16 (8)
The orientation of the structural details is likely to affect corro- (2003) 547–566.
sion progression. Generally structural members located in the low- [4] R.E. Melchers, Probabilistic model for marine corrosion of steel for structural
reliability assessment, Journal of Structural Engineering (ASCE) 129 (2003)
er regions of the tanks and those horizontally aligned where water
1484–1493.
is pooled on surfaces tend to exhibit higher corrosion rates [14]. [5] R. Loseth, G. Sekkeseater, S. Valsgard, Economics of high tensile steel in ship
These structural details may dominate corrosion progression in hulls, Marine Structures 7 (1994) 31–50.
commercial bulk carriers, however for naval ships the slenderness [6] J.K. Paik, A.K. Thayamballi, Y.I. Park, J.S. Hwang, A time-dependent corrosion
wastage model for seawater ballast tank structures of ships, Corrosion Science
and flare of the hull combined with smaller spaces for ballast tanks 46 (2) (2004) 471–486.
indicates that overall the plating details are more closely aligned to [7] J.K. Paik, J.M. Lee, J.S. Hwang, Y. Park Ii, A time-dependent corrosion wastage
the vertical plane and therefore the vertical orientation is appropri- model for the structures of single- and double-hull tankers and FSOs and
FPSOs, Marine Technology 40 (3) (2003) 201–217.
ate to represent the corrosion conditions in the naval ship ballast [8] P. Morris, Ships of Shame: Inquiry into Ship Safety, Parliament of the
tank. Obviously, this issue will require further consideration when Commonwealth of Australia, Australian Government Publishing Service,
dealing with commercial vessels, such as through using horizontal Canberra, 1992.
[9] R.E. Melchers, Probabilistic models of corrosion for reliability assessment and
and inclined coupons. However, the principles laid out in the pres- maintenance planning, in: Proceedings of the International Conference on
ent study remain valid. Offshore Mechanics and Arctic Engineering, vol. 2, 2001, pp. 27–47.
The effect of the protective coatings and cathodic protection [10] K. Okamoto, E. Sugita, Deterioration measurement of ship’s hull and
application to design, IHI Engineering Review 26 (1) (1993).
also is of interest. It is an important factor in simulating overall [11] L.D. Ivanov, Statistical evaluation of the ship’s hull cross section geometrical
long-term deterioration of a ship structure. As noted, coating ‘life’ characteristics as a function of her age, International Shipbuilding Progress 33
was not considered in the present work, mainly because the state (387) (1986) 198–203.
[12] W. Shi, In-service assessment of ship structures: effects of general corrosion on
of the art for coating life estimation is still largely empirical. Pub-
ultimate strength, Transactions – The Royal Institution of Naval Architects 135
lished data on coating life is scarce and the best practical approach (1992) 77–91.
appears to be close inspection and continual assessment of the [13] C. Guedes Soares, Y. Garbatov, Reliability of corrosion protected and
likelihood of coating breakdown on an actual structure [41]. There maintained ship hulls subjected to corrosion and fatigue, Journal of Ship
Research 43 (2) (1999) 65–78.
is little quantitative data for developing useful models of coating [14] J.K. Paik, S.K. Kim, S.K. Lee, Probabilistic corrosion rate estimation model for
breakdown [43]. Cathodic protection will limit the onset of im- longitudinal strength members of bulk carriers, Ocean Engineering 25 (10)
mersed corrosion at defects in coatings, but will not protect in (1998) 837–860.
[15] N. Yamamoto, K. Ikegami, A study on the degradation of coating and corrosion
the de-ballasted condition. These topics are obvious areas for more of ship’s hull based on the probabilistic approach, Journal of Offshore
research. Mechanics and Artic Engineering 120 (1998) 121–128.
[16] N. Yamamoto, Probabilistic corrosion model of ship structural members, in:
Proceedings of IMAS’97, Paper 1, Ships – The Ageing Process, 1997.
[17] C. Guedes Soares, Y. Garbatov, Reliability of maintained, corrosion protected
9. Conclusion plates subjected to non-linear corrosion and compressive loads, Marine
Structures 12 (6) (1999) 425–445.
[18] S. Qin, W. Cui, Effect of corrosion models on the time-dependant reliability of
Corrosion is known to influence the useful lifetime of ship struc- steel plated elements, Marine Structures 16 (2003) 15–34.
tures and to affect their lifetime operating costs. Herein a model is [19] J.K. Paik, A.K. Thayamballi, S.K. Kim, S.H. Yang, Ship hull ultimate strength
presented for the corrosion to be expected in ballast tanks of naval reliability considering corrosion, Journal of Ship Research 42 (2) (1998).
[20] R.E. Melchers, Modelling of marine immersion corrosion for mild and low-
vessels. The model takes direct account of the operational profile of alloy steels – Part 1: Phenomenological model, Corrosion (NACE) 59 (4) (2003)
the vessel and in particular the temperature variations as the ship 319–334.
operates in different geographical areas. The model also considers [21] R.E. Melchers, Probabilistic models for corrosion in structural reliability
assessment – Part 2: Models based on mechanics, Journal of Offshore
the variation of ballasting conditions. The model uses two previ- Mechanics and Artic Engineering 125 (2003) 272–280.
ously developed models, one for the corrosion of mild steel under [22] F.M. Reinhart, J.F. Jenkins, Corrosion of Materials in Surface Seawater after 12
marine immersion conditions and another for corrosion under and 18 Months of Exposure, Technical Note N-1213, Naval Civil Engineering
Laboratory, Port Hueneme, CA, 1972.
atmospheric conditions as previously calibrated for use in enclosed
[23] C.R. Southwell, J.D. Bultman, C.W. Hummer Jr., Estimating service life of steel
spaces. in seawater, in: M. Schumacher (Ed.), Seawater Corrosion Handbook, 1979, pp.
Results were presented of a shipboard trial for assessment of 374–387.
corrosion in the seawater ballast tanks of an operational naval ves- [24] Y. Kondo, Prediction method of corrosion fatigue crack initiation life based on
corrosion pit growth mechanism, Transactions of Japan Society Mechanical
sel. The results show a close correlation between the corrosion Engineers 53 (495) (1987) 1983–1987.
model simulations and the experimental observations. Addition- [25] U.R. Evans, The corrosion and oxidation of metals: scientific principles and
ally, the measured temperatures within the ballast tank showed practical applications, Edward Arnold Ltd., London, UK, 1966.
[26] B.B. Chernov, T.B. Pustovskikh, G.M. Chertskova, Evaluation of the maximum
good correlation with those predicted using geographically corrosion rate of metals in seawater, Protection of Metals 25 (4) (1989) 507–
sourced sea surface temperatures. 509.
M.T. Gudze, R.E. Melchers / Corrosion Science 50 (2008) 3296–3307 3307
[27] N.D. Tomashev, Theory of Corrosion and Protection of Metals, The MacMillan [36] N. Hogben, N.M.C. Dacunha, G.F. Olliver, Global Wave Statistics, British
Co., New York, 1996. Maritime Technology, 1986.
[28] R. Melchers, Recent progress in the modeling of corrosion of structural steel [37] B. Thygesen, Crude oil tanker cargo tank corrosion, Presented at the
immersed in seawaters, Journal of Infrastructure Systems (ASCE) 12 (3) (2006) International Technical Conference and Exhibition UK Corrosion 2002, 22–24
154–162. October, Cardiff, Wales, 2002.
[29] Hyper-accelerated corrosion, Marine Engineers Review, June 2001. [38] HUMMITER 50U/50Y(X) Integrated Humidity and Temperature Transmitter,
[30] C.P. Gardiner, R.E. Melchers, Enclosed atmospheric corrosion in ship spaces, Vaisala Instruments Catalogue, Ref. 1329en 1998–06, Helsinki, Finland,
British Corrosion Journal 36 (4) (2001) 272–276. 1998.
[31] C.P. Gardiner, R.E. Melchers, Corrosion of mild steel by coal and iron ore, [39] ASTM, ASTM G1-90: Standard Practice for Cleaning and Evaluating Corrosion
Corrosion Science 44 (2002) 2665–2673. Test Specimens, Annual Book of ASTM Standards, 1994.
[32] R.E. Melchers, T. Wells, Models for the anaerobic phases of marine immersion [40] R.E. Melchers, Transition from marine immersion to coastal atmospheric
corrosion, Corrosion Science 48 (7) (2006) 1791–1811. corrosion for structural steels, Corrosion (NACE) 63 (6) (2007) 500–514.
[33] R.E. Melchers, Mathematical modelling of the diffusion controlled phase in [41] Tanker Structure and Co-operative Forum (TSCF), Condition Evaluation and
marine immersion corrosion of mild steel, Corrosion Science 45 (2003) 923–940. Maintenance of Tanker Structures, Witherby and Co. Ltd., London, 1992.
[34] R.E. Melchers, Modelling of marine immersion corrosion for mild and low alloy [42] C.P. Gardiner, Corrosion Analysis of Bulk Carriers, PhD Thesis, Department of
steels – Part 2: Uncertainty estimation, Corrosion (NACE) 59 (4) (2003) 335–344. Civil Surveying and Environmental Engineering, University of Newcastle,
[35] National Oceanographic Data Centre (NODC), (Levitus) World Ocean Atlas Australia, 1999.
1998 (WOA98), NOAA-CIRES Climate Diagnostics Data Centre, Boulder, CO, [43] R.E. Melchers, X. Jiang, Estimation of models for durability of epoxy coatings in
USA, 1998. water ballast tanks, Ships and Offshore Structures 1 (1) (2006) 61–70.