ABS TECHNICAL PAPERS 2004

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

Lyuben Ivanov, Ge Wang, Ah Kuan Seah - American Bureau of Shipping
Presented at the Pacific 2004 International Maritime Conference (PACIFIC 2004) Reprinted with kind permission of PACIFIC 2004 , Sydney, Australia, February 3-5, 2004

ABSTRACT Probabilistic approach has long been recognized as a preferred tool for the handling of the uncertainty of ships strength calculations. By factoring in corrosion wastage over time, such an approach can be expanded to account for time-dependency of structural strength. In ABS, analytical techniques had been developed to express the geometric properties of the hull girder and its components in the form of time-dependent distribution laws. Other analytical techniques were also developed to express the design wave-induced bending moment in probabilistic terms for any given life span. This paper shows, using examples, how these loads and geometric properties, all represented in the form of time-dependent distribution laws, could now be utilized to express hull girder bending stresses in the form of time-dependent distribution laws as well. Then, the probability of exceeding given permissible total stresses was calculated and suggestions for a reliability measure based on this probability of exceedence were made. They can be used as a quantitative decisionmaking assistance at time of initial design or steel renewal. I. INTRODUCTION

Broadly speaking, corrosion wastage could be treated as a random phenomenon that affects the dimensions of plates, stiffeners and girders, just as the latters geometric properties could due to manufacturing and construction tolerances, etc. As such, all geometric properties of the hull girder and its components at any point in time in service due to corrosion could be treated in a probabilistic manner. This would permit the uncertainty related to the geometric properties be interacted with the uncertainty in predicting the loads acting on the hull structure. The former may be numerically evaluated by either analytical or approximate method [4], [6]. The latter is normally determined by separately calculating the wave-induced bending moment and the still water bending moment, and then combining them to obtain the total hull girder bending moment. Having assessed the uncertainty in the hull girder total bending moment and the uncertainty in the hull girder geometric properties, calculation of the total hull girder stresses may be performed in probabilistic terms at any ships age. Thus, the probability of exceeding the permissible total bending stresses given in the Classification Rules can be calculated and the potential risks for hull girder failure can be predicted. While the rational and the power of the probabilistic approach have been well recognized, there remain many challenges before it can be applied to everyday design and decision-making. This paper reports one aspect of the current efforts by ABS in addressing such challenges. The focus has been on using analytical formulations, rather than numerical algorithms, to the largest extent so that almost no recourse is needed for numerical solutions. Using a bulk carrier example, main
Evaluating Corrosion Wastage and Structural Safety of Aging Ships 11

ABS TECHNICAL PAPERS 2004

features of this methodology are illustrated. This includes: 1) derivation of time-dependent distribution laws of total hull girder bending moment by combining that of still water and waveinduced bending moments; 2) derivation of time-dependent distribution laws of hull girder geometric properties (in the example, deck section modulus); 3) derivation of time-dependent distribution laws of hull girder stresses with reference to specific time span; 4) a proposed measure of reliability II. CORROSION WASTAGE

Corrosion wastage depends on many factors such as ships operational life, quality of steel, methods of corrosion protection, operation conditions, etc. The combination of all these factors can vary substantially from case to case. There is an increasing interest in building up phenomenological models for the probabilistic description of the corrosion [14], [15] based on improved understanding of specific corrosion mechanisms. However, there still remain considerable uncertainties even in elegant theoretical models proposed. They are confined to some specific situations and cannot be applied to the general fleet. Practically, all corrosion models have to be calibrated with gauging results. There are limited corrosion wastage database reported in public domain. The publications of Tanker Structure Cooperative Forum (TSCF) provide the industry with experience gained in tanker industry, and have been regarded as the most reliable source (see [21], [22]). In the last 10 years following TSCF publications, some collective studies on corrosion wastage were reported (see [23], [24]), which have improved and broadened the knowledge on corrosion wastage in commercial ships. These databases provide valuable data for development of analytical methods and make it possible to quantitatively discuss general corrosion in tankers. There exist other types of corrosion, such as grooving/necking, pitting, etc. Unfortunately, there is virtually no collective data for these corrosion types, and the discussions remain, and will continue to remain for some years, mainly qualitative. In a reliability-based approach, it is still acceptable to use simplistic corrosion progress model. Although the existing uncertainties in its prediction are too many, this approach could still contribute to a better understanding of an aging ships strength. A significant number of detail records exist showing that the overall corrosion wear prevails in most cases [3]. This type of evenly distributed corrosion wear over each individual steel plate is considered in this paper. The mathematical model used in the calculations is shown in Fig. 1. (T) [mm/year]

Corrosion rate

Phase I T1

II T2

Phase III Ships age T

Fig. 1 Assumed phases of the corrosion wear

12

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

ABS TECHNICAL PAPERS 2004

The first phase (T=0-T1) is when the protective coating is intact and there is no corrosion wastage of the structure. The second phase (T=T1-T2) is a gradual acceleration of corrosion as the coating breaks down. The final phase (T > T2) is when the coating has completely failed, and the corrosion rate reaches its maximum. The corrosion wastage at any time, (T), is

( T ) = ( T ) dT
0

(1)

Data for the corrosion wear used in the paper are taken from [17]. However, the methodology is applicable to any set of data for evenly distributed corrosion wear.
III. STILL WATER BENDING MOMENT

The early publications on statistical analysis of still water bending moments [5], [7], [10], [12], [13] and [18] showed that MSW obeys the normal (Gaussian) distribution. Within the ships life, the p.d.f. of the still water bending moment depends on the ships type and loading patterns. As an example, the p.d.f. of a 25000 DWT bulk carrier for loaded conditions (hogging) is shown in Fig. 2 (p.o.e. stands for probability of exceedence). Strictly speaking, the p.d.f. of the still water bending moments should be truncated. However, it is quite possible that in practice, the design values of MSW may be exceeded. This may sometimes be revealed by careful analysis of the ships cargo plans (e.g., 10-30% exceedence relative to the design still water bending moments as reported in [19]). In this paper, the distribution law of the still water bending moment is assumed the same through the ships life due to lack of information about its change with ships age. However, the methodology is general one and can be applied to any distribution law of the still water bending moment.

4.5E-05 4.0E-05 3.5E-05 3.0E-05 p.d.f. 2.5E-05 2.0E-05 1.5E-05 1.0E-05 5.0E-06 0.0E+00 15000 25000 35000 MSW [t.m] 45000 55000 65000 75000 85000 MSW,mean = 53820 t.m MSW,st.dev. = 10350 t.m MSW,design = 69000 t.m with p.o.e. = 7.12%

p.d.f. of M(SW)

M(SW, mean)

M(SW, design)

Fig. 2 Probability density function of MSW (hogging) for the 25000 DWT bulk carrier

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

13

ABS TECHNICAL PAPERS 2004

IV.

WAVE-INDUCED HULL GIRDER BENDING MOMENT

The wave-induced bending moment is calculated following the ABS Rules [1]:
1.5 300 L M w = 19.37 10.75 100

2 L C B .10 3 [t.m]

(2)

For the 25000 DWT bulk carrier Mw is calculated as = 94795 t.m. It is widely accepted that the wave-induced bending moments follow the Weibull distribution (e.g., [16]): c.d.f.
p.d.f.

( .M W ) FM W ( M W ) = 1 e

(3)
c

fM W ( M W ) = c M W
c

c 1

( .M W ) e

(4)

The coefficient c is close to 1. For the purposes of the report, it is acceptable to assume c = 1. Thus, the Weibull distribution becomes exponential distribution (see Fig. 3):
FM W M W = 1 e

.M W

f M W ( MW ) = e

.M W

(5)

To calculate the hull girder stresses in probabilistic terms at any given year or within any given time period, one should represent the wave-induced bending moment as direct function of any given duration of the ships operational life (see [8] and [9]). In the following examples in this paper two cases will be examined: i) ii) planned operational life of 20 years; extension of the ships life with 10 years after 15 years in operation.

Conventionally, the probability of wave-induced bending moment exceeding the Classification Rulespecified design wave-induced bending moment within 20 years is of order of 10-8.
1.0E+00 1.0E-01 1.0E-02 1.0E-03 [-] 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 0 20000 40000 60000 80000 100000 wave-induced bending moment [t.m]

Fig. 3 P.o.e. of MW for operational life To=20 years with coefficient = 194321.10-9

14

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

ABS TECHNICAL PAPERS 2004

For consistency, the absolute value of the design wave-induced bending moment in the second case is determined in such a way that the p.o.e. is equal to 10-8. For the first case, the coefficient is 194321.10-9 while for the second case = 201633.10-9 (see [9]). The design wave-induced bending moments for various operational life is plotted in Fig. 4 all with p.o.e. = 1.10-8. One can observe the nonlinear relationship between the wave-induced bending moments vs. ships life.
100000 96000 92000 [t.m] 88000 84000 ship's lifetime [years] 80000 0 5 10 15 20 25 30 35 40 M(W, design) for To = var M(W, design) for To = 20 years Design MW for To = 20 years: M0, T o=20 = 94795 [t.m] with p.o.e. = 1.10 Design MW for To = 10 years: M0, T o=10 = 91357 [t.m] with p.o.e. = 1.10
-8 -8

Fig. 4 Design values of MW for different ships operational life, M0,To, having the same probability of exceedence (p.o.e. = 1.10-8) for the 25000DWT bulk carrier
V. TOTAL HULL GIRDER BENDING MOMENT

As per the Rules of the classification societies, the design still water bending moments for hogging and sagging should be added to the wave-induced bending moments to obtain the total bending moments for hogging and sagging. Dividing the total bending moment by the hull girder section modulus yields the total bending stresses. These stresses should be smaller than the permissible stresses given in the Classification societies Rules. The probabilistic approach discussed in this paper uses the same procedure except that the calculations are performed in probabilistic terms and accounts are taken of the ships age. Based on the principles of the composition of the distribution laws of the constituent variables (in this case the still water bending moment, MSW, and the wave-induced bending moment for hogging, MW), the following formulae for the cumulative distribution function (c.d.f.) and the probability density function (p.d.f.) of the total bending moment, Mt, were derived (see [11] and [20]):
FM t ( M t ) = FM
0

SW

(M t M W)f M
M t M W

(M W ) d (M W )

(6)

FM

SW

(Mt MW) =

fM

SW

( M SW ) d M SW

(7)

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

15

ABS TECHNICAL PAPERS 2004

f Mt (Mt ) = f M

SW

(Mt MW)f M

(M W ) d (M W )

(8)

fM where: f Mt (M t ) FM t ( M t ) fM
W

SW

(M ) =
SW

1
M SW

2 1 M SW M SW exp 2 M SW 2

(9)

= probability density function of Mt = cumulative distribution function of Mt = probability density function of MW calculated with Eq. (5) = mean value of MSW

(M W )

M SW

M SW = standard deviation of MSW

As an example, the probability density function of Mt for the 25000 DWT bulk carrier with operational life To = 20 years is shown in Fig. 5 and the probability of exceedence is shown in Fig. 6. Although the p.d.f. of Mt shown in Fig. 5 looks very much like Gaussian distribution, it is not Gaussian as can be seen in Fig. 6. The p.d.f. and the c.d.f. of the total design hull girder bending moment, Mt, for the second case (i.e., To = 10 years) are also calculated by Eqs. (6) - (9). The p.d.f. of Mt for time periods of 10 and 20 years are shown in Fig. 7.
4.0E-05 3.5E-05 3.0E-05 2.5E-05 p.d.f. 2.0E-05 1.5E-05 1.0E-05 5.0E-06 0.0E+00 0 50000 p.d.f. of M(total) design M(total) 100000 150000 200000 mean value of M(total) st.dev. of M(total) Mt,mean 58689 t.m Mt,st.dev 11343 t.m Mt,design = 163795 t.m with p.o.e. = 1.036.10
-7

Mtotal [t.m]

Fig. 5 P.d.f. of the total bending moment Mt for operational life To = 20 years

16

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

ABS TECHNICAL PAPERS 2004

1.0E+00 1.0E-01 1.0E-02 1.0E-03 [-] 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 0 50000 100000 150000 200000 total bending moment [t.m]

Fig. 6 Probability of exceedence of Mt for operational life To=20 years

1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 1.0E-15 0 50000 To = 10 years 100000

p.d.f.

Mt [t.m] 150000 To = 20 years 200000

Fig. 7 P.d.f. of Mt for To = 10 and 20 years operational life To overcome the non-Gaussian characteristics of Mt, as is necessary for analytical calculation purpose, the p.d.f. and c.d.f. of Mt for To= 10 and 20 years operational life were substituted with the following equations: For 20 years operational life, the p.d.f. of Mt was presented by the equation a + c.M t + e.M 2 t f M t ( M t ) = exp 3 1 + b.M t + d.M 2 + f.M t t (10)

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

17

ABS TECHNICAL PAPERS 2004

For 10 years operational life, the c.d.f. of Mt was presented by the Weibull equation

d M t + c.ln 21/ d b (11) F M t M t = a 1 exp c Much analytical work and curve-fit exercises were performed to examine the types of equation that would best fit the p.d.f. and c.d.f. of Mt. Eqs. (10) and (11) were selected from many candidates based on accuracy and convenience for use.

( )

One should emphasize here that the distribution laws of the design total bending moment refer to the whole time periods under consideration (i.e., for To = 20 and 10 years)
VI. HULL GIRDER SECTION MODULUS

Extensive studies had been made in ABS in probabilistic assessment of hull girder geometric properties [6]. In essence, the variability of the probabilistic distributions of hull girder geometric properties over time can be represented by a three-dimensional surface. As an example, the calculations are performed for the deck section modulus of the 25000 DWT bulk carrier. The envelope of the p.d.f. of SMdeck is shown in Fig. 8. The p.d.f. of SMdeck should be truncated due to physical reasons (SM cannot be infinite). It requires calculating a coefficient CA that should be used to correct the ordinates of the p.d.f. [6]. In this paper, the upper limit of SMdeck is assumed 1.05 of SMdeck, nominal while the lower limit is assumed 0.8 of SMdeck, nominal. Having built the envelope, one can calculate the probability that SMdeck could be between any limits during any given time interval with the formula

30 20 10 0 1 0

S M /S

M (n o

0.9

20 0.8

m)

Fig. 8 Envelope of the p.d.f of SMdeck for the 25000 DWT bulk carrier
18 Evaluating Corrosion Wastage and Structural Safety of Aging Ships

sh ip 's

ag e[ ye ar

10

s]

p.d.f.

ABS TECHNICAL PAPERS 2004

P SM1 SM SM 2 ;T 1 T T2 where: SM1 T1 CA,T PT(SM) = = = =

1 = T2 T 1

SM 2 dT C p SM d SM ( ) ( ) A,T SM T T1 1 T2

(12)

lower limit of SM SM2 = upper limit of SM lower limit of T T2 = upper limit of T coefficient for correction of the ordinates of pT(SM) at any year T Gaussian probability density function of SM at any year T

In physical terms, Eq. (12) represents the ratio between the volume below the envelope within the given limits and the total volume below the envelope. The cross section of the envelope at any ships age gives the probability density function of SMdeck for this year. In the paper, the envelope of SMdeck is used in two ways: i) To calculate the probability of SMdeck exceeding a given value within the planned operational life (see the darker surface in Fig. 9) or any given time interval; ii) To obtain the probability density function of SMdeck at any given year (see Fig. 10).

30

20

p.d.f.

10

0 0.8

sh

ip

10

's

0.9

ag

e[

ye

ar

s]

20

SM

( /S M

no

m)

Fig. 9 Illustration of the procedure to calculate the probability that SMdeck will be greater than a given value As an example, two p.d.f. of SMdeck are shown in Fig. 11 for operational life To=10 and 20 years. It is interesting to note that the distribution law of SMdeck for the whole time period does not obey Gaussian distribution although at each year it is Gaussian. One possible explanation of this fact is that the envelope of the p.d.f. is not smooth because of the knuckle in the envelope (see Fig. 8) at the end of the coating longevity.

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

19

ABS TECHNICAL PAPERS 2004

To overcome the non-Gaussian characteristics of SMdeck necessary for calculation in probabilistic terms of the total hull girder bending stresses, the p.d.f. of SMdeck were replaced by equations in the following way:
40 35 30 25 20 15 10
SM(T)/SM(nom)

5 0 0.70 year 25

1.05

1.00 year 0

0.95 year 5

0.90 year 10

0.85 year 15

0.80

0.75 year 20

Fig. 10 P.d.f. of SMdeck at several years of the ships operational life

2.0E-04

1.5E-04
p.d.f.

1.0E-04

5.0E-05 SM [m.cm ] 0.0E+00 75000 80000 85000 90000 95000 100000 p.d.f. of SM(deck) for time period T = 0 - 20 years p.d.f. of SM(deck) for time period T = 15 - 25 years would-be Gaussian distribution for time period T = 0 - 20 years would-be Gaussian distribution for time period T = 15 - 25 years
2

Fig. 11 P.d.f. of SMdeck for To = 10 and 20 years operational life

20

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

ABS TECHNICAL PAPERS 2004

For 20 years operational life, the cumulative distribution function of SMdeck was first analytically presented by the Weibull equation. Then it was differentiated and the following equation of the probability density function of SMdeck was derived:

f SM

deck

(SM deck )

e e 1 SM SM deck + d.ln 21/ e c + d.ln 21/ e c b.e deck = exp d d d

(13)

For 10 years operational life, the probability density function of Mt was represented by the equation

2 3 4 5 ln f SM SM deck = a + b.SM deck + c.SM deck + d.SM deck + e.SM deck + f.SM deck deck

(14)

As for Mt, much analytical work and curve-fit exercises were performed to examine the types of equation that would best fit the p.d.f. of SMdeck. Eqs. (13) and (14) were selected from many candidates based on accuracy and convenience for use.
VII. CALCULATION OF THE TOTAL BENDING STRESSES FOR AS-BUILT AND AGING SHIPS

Technically, the calculations could be performed using p.d.f. or c.d.f of SMdeck with reference to a given year or to a given time period. Since the p.d.f. or c.d.f. of Mt is with reference to given time periods (To=10 and 20 years), for the same time reference should be used for calculation of the total bending stresses. For purposes of comparison, two sets of calculations of the total hull girder bending stresses were performed for each time period of 10 and 20 years: A. The distribution law of the design total hull girder bending moment refers to the whole time period while the distribution law of SMdeck at the beginning of the time period is used. This case is close to the present practice for calculating the hull girder bending stresses. B. The distribution laws of both the design total hull girder bending moment and SMdeck refer to the whole time period. Again, the composition of the laws of distribution of the constituent variables (in this case the total design hull girder bending moment, Mt, and the deck section modulus, SMdeck) is applied to obtain the cumulative distribution function (c.d.f.) and the probability density function (p.d.f.) of the total hull girder bending stresses in the deck: Fs where:
Fs
t ,deck

t,deck

(s t,deck ) = FM t (s t, deck .SM deck ).f SM

0

deck

(SM deck ) d (SM deck )

(15)

(s t ,deck )

= c.d.f. of st,deck stress in the deck, st,deck, and the deck section modulus, SMdeck. It is calculated either by integrating Eq. (10) or directly by Eq. (11)

FM t (s t , deck .SM deck ) = c.d.f. of Mt,. which is expressed through the product of the bending

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

21

ABS TECHNICAL PAPERS 2004

f SM

deck

(SM deck )

= p.d.f. of SMdeck. At the beginning of each time period it is Gaussian (see Fig. 10) and is determined with the equation 1
1 SM deck SM deck exp SM deck 2 2
2

f SM

deck

(SM deck ) =

SM deck

(16)

SM deck = mean value of SMdeck

SM

deck

= standard deviation of SMdeck (SM deck ) is determined either by Eq.

For the whole time periods of 10 and 20 years, f SM (13) or Eq. (14)
SM d e ck , ma x

deck

fs

t , deck

(s t , deck ) =

SM d e ck , min

fM

( M t = s t deck SMdeck ) f SM deck (SM deck )SMdeck d (SMdeck )

,

(17)

where:
fs
t , deck

(s t , deck ) = p.d.f. of st,deck

The results of the calculations of st,deck for the two cases with 10 and 20 years operational life are shown in Fig. 12 together with the permissible total bending stresses of 1784 kg/cm2 given in the Classification Rules [1]. One can clearly observe in Fig. 12 two trends: a) the probability density functions (p.d.f.) of st,deck is shifted towards the permissible total bending stresses when the distribution laws refer to the whole time period under consideration, b) within one set of calculations (A or B), the difference between the p.d.f. is not big, especially relative to the permissible stresses located in the asymptotic tails (numerical results of the calculations for set A and B are given in the next Section). In the authors view, the use of distribution laws of the hull girder geometric properties corresponding to the whole time interval is preferable because it is more consistent with the approach to calculating the loads acting on the hull structure.
VIII. A MEASURE OF RELIABILITY

In mathematical terms, the calculation of the probabilities of exceedence of the permissible stresses requires very high accuracy because the permissible stresses are in the asymptotic tails of the p.d.f. (see Fig. 12). Even very small change of the ordinates of the distribution laws in the asymptotic tails leads to substantial change of the probability of exceedence of given limit. Therefore, to make the calculations more stable, a measure of the probability of exceedence proposed in [2] is used. It is based on the proposal of Prof. V. V. Bolotin to measure the reliability of structures with very high reliability: R = log10 ( p.o.e.) [ Bells] (18)

Using Bells as a measure of the reliability (defined as the probability P(st,deck < [st])), the following results for the p.o.e. of the permissible total bending stresses for each of the four probability density functions in Fig. 12 were obtained (for comparison, the results are also shown in the traditional format of probability of exceedence):

22

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

ABS TECHNICAL PAPERS 2004

14 permissible total bending stress [st,deck] = 1784 kg/cm2

12

10

8 p.d.f. 6 4 2 st [kg/cm ] 0 200 400 600 800 1000 1200 1400 1600 1800 p.d.f. of s(total) for period To = 15-25 years (set B) p.d.f. of s(total) for period To = 0-20 years (set B) p.d.f. of s(total) for period To = 0-20 years (set A) p.d.f. of s(total) for period To = 15-25 years (set A) permissible total stress for new design
2

Fig. 12 Probability density functions (p.d.f.) of the total bending stresses in the deck for the two sets of calculations for operational life of 10 and 20 years A. The distribution law of the design total hull girder bending moment refers to the whole time period while the distribution law of SMdeck at the beginning of the time period is used For 20 years operational life: P(st < [st]) = 4.27 Bells P(st > [st]) = 5.34.10-5 For 10 years operational life: P(st < [st]) = 4.17 Bells P(st > [st]) = 6.72.10-5 B. The distribution laws of both the design total hull girder bending moment and SMdeck refer to the whole time period For 20 years operational life: P(st < [st]) = 2.18 Bells P(st > [st]) = 0.0067 For 10 years operational life: P(st < [st]) = 2.37 Bells P(st > [st]) = 0.0043

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

23

ABS TECHNICAL PAPERS 2004

One can observe in the first set of calculations, A, (i.e., when the distribution law of Mt refers to the whole time period of 10 or 20 years, but that of SMdeck is determined for the beginning of each time period) that the difference between the probability of exceedence (p.o.e.) of the permissible total bending stresses for 10 and 20 years service life (measured in Bells) is about 2.4%. In the second set of calculations, B, (i.e., when the distribution laws of Mt and SMdeck refer to the whole time period of 10 or 20 years) this difference (measured in Bells) is about 8.7%. The numbers might be slightly influenced by the unavoidable inaccuracy (although the accuracy of the curve-fit is extremely high) when the ordinates of the probability density functions or cumulative distribution functions were substituted by analytical equations, as aforementioned. However, even with this in mind, one can conclude that the differences between the p.o.e. of the permissible total bending stresses in the Classification Rules for new design within one set of calculations are quite small, especially when compared with the uncertainties in the loads, corrosion wear, construction tolerances, etc. However, the difference between the probabilities of exceedence (measured in Bells) of the same permissible total bending stresses in the two sets of calculations, A and B, is much greater (between 176 - 196%). It does not mean that the reliability of the hull structure has changed by the same numbers. The probability of exceedence (i.e., the probability of failure) is method-dependent. Direct comparison between results obtained by two different methods is not meaningful. Comparisons can be made only between results obtained by the same method. In this sense, any set of calculations (A or B) is acceptable, provided all parameters are presented in probabilistic terms, as described in this paper. In other words, whatever set of calculations is used (A or B), it is reasonable to use the permissible bending stresses given in the Classification Rules for new design in assessing the strength of old ships, provided the calculations are performed in probabilistic terms, as described in the paper. In plain language, there is no need to develop a new set of permissible stresses for aging ships, provided all parameters determining the ships strength are calculated in probabilistic terms in the same way as for new design. It is interesting to mention that the probability of exceedence of the wave-induced bending moment given in the Rules is 1.10-8. However, when all other uncertainties are taken into consideration (still water bending moment, geometric properties of structural components and hull girder), the probability of exceedence of the permissible total bending stresses increases up to 4.3 6.7x10-3. The accuracy of the calculations with the analytical method presented in this paper was double checked by repeating all calculations with Monte Carlo simulation method. The maximum difference between the results obtained by the two methods is 2%. The analytical method provides accurate solution in closed form within very short time, including in areas of the asymptotic tails. Thus, the traditional application of the Monte Carlo method in such problems can be avoided, especially when high accuracy requires long computational time.
CONCLUSIONS

This paper illustrates that time-consuming numerical analyses need not necessarily be the only workable tool for solving probabilistic formulations. An analytical method has been developed for calculation of the probability of exceedence of given permissible total hull girder stresses. It explicitly accounts for the uncertainties associated with age related hull structure degradation due to corrosion, probabilistic nature of still water and wave-induced bending moments. This probability of exceedence can be used as a measure at the new design stage or at any point in the service life of the ship to account for the desired level of structural safety. The method does not require

24

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

ABS TECHNICAL PAPERS 2004

application of specially developed computer programs for calculating the reliability of the hull structure. It is believed that such an approach will facilitate the application of time-dependent Reliability theory in ship structural design, maintenance and repair.
ACKNOWLEDGEMENT

The authors wish to thank ABS management for supporting this publication. The opinions expressed are those of the authors and not those of the company.
REFERENCES

1. 2. 3. 4.

5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15.

American Bureau of Shipping Rules for Building and Classing Steel Vessels, part 3, Hull Construction and Equipment, 2002 Bolotin V V Mechanics of the rigid body and reliability theory (in Russian), Proceedings of the Second Congress on Theoretical and Applied Mechanics, Publishing House Nauka (Science), Moscow, 1966 Huang R - Corrosion Protection of Ballast and Cargo Tanks of Crude Oil Tankers, Proceedings of the International Workshop on Corrosion Control for Marine Structures and Pipelines, Galveston, Texas, February 9-11, 1999 Ivanov L D Probabilistic Presentation of the Geometric Properties of Shipbuilding Structural Profiles when Assessing Elastic Bending Strength, vol. 1 Theory, 143 pages, vol. 2 Numerical examples, pp. 144-340, American Bureau of Shipping, Technical Report RD-2001-19, December 2001 Ivanov L D The statistical estimation of still water bending moments for cargo ships, Fifth International Ship Structures Congress, Hamburg, September 1973, Proceedings, vol. 2, pp. 18-19 Ivanov L D A Probabilistic Assessment of all Hull Girder Geometric Properties at any Ships Age, American Bureau of Shipping, Technical Report R&R-2002-1, December 2002, 166 pages Ivanov L D, Madjarov H The statistical estimation of still water bending moments for cargo ships, Shipping World and Shipbuilder, 1975, No. 3908, pp. 759-762 Ivanov L D, Kamenov L Probabilistic estimation of the resource of the main ships hull structure, Fourth International Symposium Ships Reliability and Producibility91 16-17 May, 1991, Varna, Bulgaria, Proceedings, pp. 145-155 Ivanov L.D. Level of Confidence: Definition, Calculation and Use for New Designs and Aging Vessels, American Bureau of Shipping, Technical Report R&R-2003-1, November 2003 Lewis E V, Hoffman D, MacLean W M, van Hooff R, Zubaly R B Load Criteria for Ship Structural Design, Ship Structure Committee Report SSC-240, Washington, 1973 Livshits N A, Pugachev V N Probabilistic Analysis of Automatic Control Systems (in Russian), part I, Publishing House Soviet Radio, Moscow, 1963 Mano H, Kawabe H, Iwakawa K, Mitsumune N Statistical Character of the Demand on Longitudinal Strength (Second Report) Long Term Distribution of Still Water Bending Moment, Journal of the Society of Naval Architects of Japan, 1977, pp. 255 263 Maximadji A I Estimation of the Still Water Bending Moments for Tankers (in Russian), Proceedings of the Central Scientific Research Institute of the Merchant Fleet (CNIIMF), Leningrad, 1973, vol. 169, pp. 16-33 Melchers R E - Probabilistic modelling of immersion marine corrosion, Proceedings of ICOSSAR97 The 7th Int. Conf. On Structural Safety and Reliability, Kyoto, 24-28 Nov. 1997, edited by Shiraishi, Shinozuka & Wen, 1998, Balkema, Roterdam, pp. 81-88 Melchers R E - Probabilistic Modelling of Marine Corrosion of Steel Specimens, Proceedings

Evaluating Corrosion Wastage and Structural Safety of Aging Ships

25

ABS TECHNICAL PAPERS 2004

16. 17. 18. 19. 20. 21. 22. 23. 24.

of the 5th (1995) International Offshore and Polar Engineering Conference, The Hague, The Netherlands, June 11-16, 1995, pp. 204-210 Ochi M K Applied Probability and Stochastic Processes in Engineering and Physical Sciences, John Wiley & Sons, 1989 Paik J K, Kim S K, Lee S K - Probabilistic Corrosion Rate Estimation Model for Longitudinal Strength Members of Bulk carriers, Ocean Engineering, 1998, vol. 25, No. 10, pp. 837-860 Soares C G, Moan T - Statistical Analysis of Still Water Bending Moments and Shear Forces in Tankers, Ore and Bulk Carriers, Norwegian Maritime Research, No.3 1982, pp. 33-48 Soares C. Guedes Influence of Human Control on the Probability Distribution of Maximum Still-water Load Effects in Ships, Marine Structures, vol. 3, No. 4, 1990, pp. 319-339 Suhir E Applied Probability for Engineers and Scientists, McGraw-Hill, 1997 TSCF (Tanker Structure Co-operative Forum) Condition evaluation and maintenance of tanker structures, Witherby & Co. Ltd., London, 1992 TSCF (Tanker Structure Co-operative Forum) Guidance manual for tanker structures, Witherby & Co. Ltd., London, 1997 Wang G, Spencer J, Elsayed T Estimation of corrosion rates of oil tankers, 22nd International Conference on Offshore Mechanics and Arctic Engineering, 8 13 June 2003, Cancun, Mexico Wang G, Spencer J, Sun H H Assessment of corrosion risks to aging ships using an experience database, 22nd International Conference on Offshore Mechanics and Arctic Engineering, 8 - 13 June 2003, Cancun, Mexico

26

Evaluating Corrosion Wastage and Structural Safety of Aging Ships