Engineering Structures 61 (2014) 22–30

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Influence of structural deterioration over time on the optimal time

interval for inspection and maintenance of structures
Dante Tolentino ⇑, Sonia E. Ruiz
Instituto de Ingeniería, Mecánica Aplicada, Universidad Nacional Autónoma de México, Coyoacán 04510, México, DF, Mexico

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

Article history: The influence of the time variation of the structural demand and/or of the structural capacity on the opti-
Received 13 January 2012 mal time interval (based on a cost–benefit analysis) for inspection and maintenance of offshore structures
Revised 18 December 2013 is analyzed. Reliability is expressed in terms of the expected number of failures over a time interval by
Accepted 9 January 2014
means of closed-form mathematical expressions which consider the structural degradation. The mathe-
Available online 7 February 2014
matical expressions are incorporated into the cost optimization formulation. Three scenarios are consid-
ered: (1) structural demand (for a given intensity) varies in time, while structural capacity remains
Keywords:
constant, (2) structural capacity deteriorates over a time interval, while structural demand remains con-
Cost–benefit analysis
Structural deterioration
stant, (3) both structural capacity and structural demand vary simultaneously over time. The optimal
Optimal time interval time interval for inspection and maintenance corresponds to the lowest cost of inspection, repair and fail-
Closed-form expressions ure. The cost optimization is applied to an offshore jacket platform. The damage condition is given by the
Maintenance fatigue crack size at critical joints. It is shown that in order to estimate the optimal time interval for
inspection and maintenance of the structure, it is necessary to take into account the variation in time
of both its structural capacity and its structural demand (case 3); if one of them were ignored, the optimal
time interval could be overestimated.
Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction structural maintenance [14–16], as well as with structural mainte-

Structures are subjected to loads that can lead to degradation of
the structural mechanical properties as time goes by, which leads
to the decreasing of the structural capacity while the structural de-
mand increases. As a consequence, the structural reliability is mod-
ified; therefore, it is convenient to develop tools oriented to
evaluate the structural reliability considering that the structural
capacity and/or the structural demand change(s) over time. After
the structural reliability is known, an optimization analysis can
be applied in order to find the optimal time interval for inspection
and maintenance of the structure. The results of the optimization
analysis will depend on the hypothesis made about the variation
in time of the structural capacity and of the structural demand.
To establish an optimization model, Forsell [1] formulated the
problem by minimizing the total cost. The problem has been stud-
ied by using probability-based methods [2]. Later, reliability for-
mulations were introduced into probability-based design codes
[3–8] and optimal design criteria were developed on the basis of
probability concepts [9–13]. During the last fifteen years several
authors have applied the concept of time-dependent reliability in-
dex to optimize the life cycle of deteriorating structures without
⇑ Corresponding author. Tel.: +52 (55) 56233600x8480.
E-mail address: DTolentinoL@iingen.unam.mx (D. Tolentino).
nance [16]. Later, maintenance programs for existing structures
using the concept of multiobjective optimization were developed
considering different optimization objectives: i.e., (a) minimizing
the maintenance cost and maximizing the load-carrying capacity
and durability [17], (b) minimizing the cost and maximizing the
structural performance [18], (c) minimizing the maximum condi-
tion index, maximizing the minimum safety index, and minimizing
the present value of cumulative maintenance cost [19–21], and (d)
minimizing the maximum probability of failure, maximizing the
minimum redundancy index, and minimizing the life-cycle cost
[22]. A criterion to find optimal time intervals for maintenance
based on multiobjective optimization which considers confidence
factors obtained by means of closed-form expressions, damage in-
dex and the expected cumulative total cost of structures with
degrading structural properties, has been developed [23]. Recently
a generalized probabilistic framework for optimum inspection and
maintenance planning that maximizes the expected service life
and minimizes the total life-cycle cost has been formulated [24].
For the particular case of offshore structures, the degradation of
structural properties is mainly due to fatigue phenomena caused
by waves continuously acting on the structures. The mechanical
deterioration caused by fatigue is reflected by cracking of the tubu-
lar joints, giving place to a decrease of the resistance capacity of
the structural system, and as a consequence, an increase in its

0141-0296/$ - see front matter Ó 2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2014.01.012
 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30
structural demand (for a given maximum wave height). The idea
behind inspecting an offshore structural system is to detect the
presence and growth of size cracks, in order to perform the neces-
sary repairs and maintenance to the structure later on. Several
authors have developed inspections and maintenance plans for off-
shore structures: (a) based on risk and reliability of welded con-
nections subject to fatigue [25–27]; (b) using methodologies that
take into account fatigue sensitivity analyses in steel joints [28];
(c) implementing simplified approaches and using practical design
parameters such as fatigue design factors [29] and/or reserve
strength ratios [30]; (d) considering the damage caused by fatigue,
buckling and dents on structural elements [31]; and (e) using
Bayesian techniques [32].
The main objective of the present study is to analyze the influ-
ence of the time variation of the structural demand and/or of the
structural capacity on the optimal time interval (based on a cost–
benefit analysis) for inspection and maintenance of offshore struc-
tures. Three scenarios are considered: (1) structural demand (for a
given intensity) varies in time, while structural capacity remains
constant, (2) structural capacity deteriorates over a time interval,
while structural demand remains constant, (3) both structural
capacity and structural demand vary simultaneously over time.
The optimal interval for inspection and maintenance corre-
sponds here to the lowest total cost associated with inspection, re-
pair and failure, considering risk and reliability of the structure.
The difference between the present study and those mentioned
above is that in this paper reliability is expressed in terms of the
expected number of failures over a time interval by means of sim-
plified closed-form mathematical expressions that take into ac-
count the structural deterioration over time. None of the
methods above has solved the problem by means of this simplified
approach. The mathematical expressions are then incorporated
into the cost optimization formulation. Besides, it is demonstrated
that for estimating the optimal interval of inspection and mainte-
nance of a deteriorating structure, the variation in time of both
structural capacity and structural demand (for a given maximum
wave height) must be taken into account (case 3), otherwise the
optimal time interval could be overestimated.
2. Expected total present value of the cost function over a time
interval
The expected total cost function is defined as the summation of
the total expected cost, C, of inspection (I), repair (R), and failure (F)
at the end of a time interval [0, Dt) as follows:
C Total ð0; DtÞ ¼ C I ð0; DtÞ þ C R ð0; DtÞ þ C F ð0; DtÞ
where C I ð0; DtÞ, C R ð0; DtÞ and C F ð0; DtÞ are the present value of the
expected cost of inspection, repair and failure, respectively. The
optimal interval corresponds to the lowest cost over the design life
of the structure. The present value of the expected cost C I ð0; DtÞ,
C R ð0; DtÞ and C F ð0; DtÞ considering the structural deterioration over
a time interval, is defined in Sections 2.1–2.4, respectively.
2.1. Quality of inspection (q)
The inspection quality is related to the probability of detecting a
crack of a given size. That probability depends on the crack size, the
inspection method and the inspection team. The reliability with re-
spect to the probability of detecting a crack is defined by the prob-
ability of detection (PoD) curve which is identical to the
distribution function of the smallest detectable crack size [33].
PoD curves for different inspections techniques (i.e. magnetic par-
ticle inspection [MPI], Eddy current, alternate current methods
[ACFM and ACPD] and ultrasonic techniques) has been compiled
23
by Visser [34]. Also, PoD curves for MPI and Eddy current tech-
niques can be obtained from the DNV guidelines [35]. PoD curves
can be represented by a logistics, exponential and lognormal mod-
els. In the present study a PoD curve which is represented by the
following exponential model is used [36,37]:
x  a 
min
PoDðxÞ ¼ 1  exp ; a > amin ð2Þ
k
The inspection quality is given by the parameter k which is the
mean size above amin (equal to 2 mm, [37]) of the smallest detect-
able size. In the optimization analysis of the present study the fol-
lowing auxiliary measure of inspection quality is introduced:
q ¼ 1=k ð3Þ
q = 0 refers to no inspection, q = 1 is related to a perfect quality
inspection, while q  0.2–0.3 can be related to visual inspection.
2.2. Present value of the expected cost of inspection
If a crack size is detected (given by Eq. (3)), the present value of
the expected cost of inspection during the time interval [0, Dt) can
be obtained as:
Z Dt
C I ð0; DtÞ ¼ C I ðsjqÞ  pS ½dðsÞ  ecs ds ð4Þ
0
where CI(s|q) is the cost of inspecting the structure in the instant s,
for a given inspection quality q, it is considered that the structure
with cumulative damage d has survived up to the instant s with a
probability of pS[d(s)]; and ecs is a factor that converts the cost
to its present value, given a discount rate c.
Considering that the inspection of the structure will be done if a
crack is detected at the end of the interval [0, Dt), and assuming
that the structure survives up to the end of Dt, Eq. (4) is simplified
as:
C I ðDtÞ ¼ C Ijq;Dt  ecðDtÞ  pS ½dðDtÞ ð5Þ
where CI|q,Dt is the inspection cost at the end of [0, Dt) for a given
inspection quality q, ec(Dt) is the present value cost factor. Eq. (5)
considers that the structure with cumulative damage d has survived
up to the end of Dt with a probability of pS[d(Ds)].
On the other hand, if it is assumed that the occurrence of the
structural failure corresponds to a non-homogeneous Poisson pro-
cess, then:
R Dt
 mF ðsÞds
pS ½dðDtÞ ¼ e 0 ¼ egF ð0;DtÞ ð6Þ
where mF ðsÞ represents the expected annual structural failure rate at
ð1Þ instant, s; gF ð0; DtÞ is the expected number of failures at the end of a
time interval Dt, gF ð0; DtÞ is defined with more detail in the next
section. Finally, the present value of the expected inspection cost,
for a given inspection quality, at the end of the interval [0, Dt) is:
C I ð0; DtÞ ¼ C Ijq;Dt  ecðDtÞgF ð0;DtÞ ð7Þ
2.3. Present value of the expected cost of repair
Following an inspection, a decision must be made regarding if a
crack is detected. The repair decision will depend on the inspection
quality q [38]. If a crack is not detected, repair actions will not be
performed, in the case that a cracked is detected at the end of a
time interval of interest, the present value of the expected cost of
repair within the time interval [0, Dt) is given by the following
equation:
Z Dt X
n
C R ð0; DtÞ ¼ C rj ½dðsÞjq  pRj ½dðsÞjS  pS ½dðsÞ  ecs ds ð8Þ
0 j¼1
24
where n is the number of elements to be repaired; C rj ½dðsÞjq is the
repair cost of element j having a cumulative damage equal to or
greater than d, at instant, s, given an inspection quality q; and
pRj[d(s)|S]ds is the probability of repairing the jth element with
cumulative damage d, given that the structure has survived up to
instant s. Eq. (8) considers that the structure with cumulative dam-
age d has survived up to the instant s with a probability of pS[d(s)].
When the damage level of the jth structural element is equal or
exceeds a certain threshold, the probability of repairing the
element, is as follows:
PRj ½dðDtÞjS ¼ PðDj ðDtÞ  dÞ
where P(Dj(Dt) P d) is the conditional probability that element j
with certain level of cumulative damage D at time Dt, is higher or
equal to an established damage level d.
Considering that the damage has been accumulated in element j
during the time interval [0, Dt), Eq. (8) is simplified as:
X
n mathematical expression, etc.), are similar.
C R ð0; DtÞ ¼ C rj jq;Dt  PðDj ðDtÞ  dÞ  ecðDtÞ  P S ½dðDtÞ
j¼1
where C rj jq;Dt represents the expected cost of repair of element j, at
the end of interval [0, Dt), given an inspection quality q, and it is
considered that the structure with cumulative damage d has sur-
vived up to the end of Dt with a probability of PS[d(Dt)].
Assuming that the occurrence of repairing the elements corre-
sponds to a non-homogeneous process, and that the structure
has survived up to the end of Dt (similar to Eq. (6)), then the ex-
pected repair cost at the end of [0, Dt) can be obtained as follows:
X
n
C R ð0; DtÞ ¼ C rj jq;Dt  PðDj ðDtÞ  dÞ  ecðDtÞgF ð0;DtÞ
j¼1
2.4. Present value of the expected cost of failure
The present value of the expected cost of failure that can occur
within the interval [0, Dt) can be calculated as:
Z Dt
C F ð0; DtÞ ¼ C f ½dðsÞ  pF ðsÞ  ecs ds
0
where Cf[d(s)] is the cost of the structural failure, given the level of
cumulative damage d at instant s; pF(t) is the probability density
function of waiting time to failure, and is equal to:
Rt
 mF ðsÞ
pF ðtÞ ¼ mF ðtÞe 0 ¼ mF ðtÞegF ð0;tÞ
Considering that the expected cost of failure and the present va-
lue factor are constant values during the time interval of interest,
expression (12) becomes as follows:
Z Dt
C F ð0; DtÞ ¼ C f jDt ecðDtÞ mF ðsÞegF ð0;sÞ ds
0 necessary to assume that the median structural capacity, C
Solving Eq. (14) for small intervals of time (e.g., Dt = 1 year), the
present cost value of the possible structural failure within the
interval [0, Dt) is equal to:
b
X
N
C F ð0; DtÞ ¼ C f jDt gF ðtk  tk1 ÞecðDtÞgF ð0;tk tk1 Þ
k¼1
where N b is the number of time intervals of interest; gF ðt k  t k1 Þ is
the expected number of failures for the time intervals of interest,
b years.
and t 1 ; t2 . . . tN ¼ 1; 2 . . . N
D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30
3. Evaluation of structural reliability
In this paper, structural reliability is expressed by means of the
expected number of structural failures at the end of the time interval,
gF ð0; DtÞ, which is an extension of the expected annual structural
failure rate concept (m(t), [39]). Based on the simplified approach
proposed by Cornell and collaborators [40,41] to evaluate the reli-
ability of structures subject to seismic loads, and later on, extended
by Torres and Ruiz [42], it is possible to obtain gF ð0; DtÞ as shown
in Eq. (16). It is noticed that in the present study, the simplified ap-
proach by Cornell and collaborators [41] constitutes the basis to
ð9Þ evaluate the reliability of marine offshore structures subject to
wave loads, instead of the reliability of structures subjected to seis-
mic loads. The use of the simplified approach is possible because
the concepts involved in both types of structures (i.e., statistical
description of the structural capacity and of the structural demand,
simplified representation of environmental hazard curves, repre-
sentation of the structural deterioration by means of a simple
ð10Þ Z " # r
Dt b sÞ bðsÞ
Cð
gF ð0; DtÞ ¼ k
0 aðsÞ
!
r2  
2 2 2
 exp 2
r ln Djhmax ð sÞ þ r
ln C ð sÞ þ r sÞ
UT ð ds
2b ðsÞ
ð16Þ
In Eq. (16) r2ln Djhmax ðsÞ and r2ln C ðsÞ represent the variances of the nat-
ural logarithms of the structural demand given a maximum wave
height hmax, and of the structural capacity, at instant s, respectively.
It is considered that these uncertainties present a lognormal type
distribution [5]. The expression r2UT ¼ r2UD þ r2UC represents the epi-
ð11Þ
stemic uncertainty related to the demand (UD) and to the limit state
capacity (UC), respectively; k and r are parameters that define the
characteristics of the wave hazard curve around the maximum
r
wave height of interest; m(hmax) is defined by mðhmax Þ ¼ k  hmax
b
[40,41]. C ðsÞ represents the median of the limit state capacity at
instant s, and presents a lognormal type distribution; a(s) and
b(s) are parameters defining the relation of the median structural
demand with the maximum wave height hmax, at instant s.
In this study Eq. (16) is solved for three specific cases:
ð12Þ
1. Structural demand (for a maximum wave height) varies in
time, while structural capacity remains time invariant.
2. Structural capacity varies in time, while structural
demand remains time invariant.
3. Both structural demand and structural capacity vary in
time, simultaneously.
ð13Þ
3.1. Considering that structural demand for a given maximum wave
height varies in time, while structural capacity is time independent
(case 1)
ð14Þ Considering only that the structural demand varies in time, it is
b ðsÞ, and
its standard deviation, r2ln C ðsÞ, have constant values during the
b ðsÞ ¼ a and
time interval of interest, which means that C
r2ln C ðsÞ ¼ r2ln C , respectively.
It is assumed that the median of the structural demand is given
by Dð b sÞ ¼ ða þ f  sÞ  yb .
ð15Þ It is important to notice that Cornell et al. [41] assumed the
median structural demand (for a given intensity) given by
b ¼ a  yb when structural demand is invariant in time (without
D
considering structural deterioration).
Taking into account the assumptions above, the expected num-
ber of failures at the end of a time interval results as follows:
 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30 25

abr  
r2  where XCD,T(t, Dt) is a correction function for the expected number
gFD ð0; DtÞ ¼ k exp 2
r2ln Djhmax þ r2ln C þ r2UT XD ð0; DtÞ
of failures which considers the variation in time of both the struc-
a 2b
ð17aÞ tural capacity and the structural demand (for a given maximum
wave height), simultaneously. In this paper the hypergeometric
where function F(A, B; C; x), which appears in Eq. (19b), is solved by means
" bþr  r #
ba a1  a b a b
of the following hypergeometric series [44]:
XD ð0; DtÞ ¼  þ ð17bÞ AB AðA þ 1ÞBðB þ 1Þ 2
f ðb þ rÞ a a þ f ðDtÞ a FðA; B; C; xÞ ¼ 1 þ xþ x þ 
1!C 2!CðC þ 1Þ
where XD(0, Dt) is a correction function for the expected number of AðA þ 1Þ    ðA þ n  1ÞBðB þ 1Þ    ðB þ n  1Þ n
failures over the time interval [0, Dt). þ x
CðC þ 1Þ    ðC þ n  1Þn!
3.2. Considering that structural capacity varies in time, while ð20Þ
structural demand is time independent (case 2) r r r f ðbT þ aT Þ
where A ¼ 1  ; B ¼  ; C ¼ 2  ; x ¼ .
b b b f aT  abT
In this section, only the variation in time of the structural capac-
b ¼ a  yb and 4. Illustrative example
ity is considered. For this case, the median demand D
the standard deviation of the natural logarithm of the demand
In this section the optimal interval for inspection and mainte-
r2ln Djy are constant values during the interval of interest. It is also
nance a jacket offshore platform during its design life (DL) is for-
assumed that the median capacity varies linearly in each time sub-
b ðsÞ ¼ aT þ b s, T = 1, 2 . . . N. The total time interval mulated, assuming the following hypotheses:
interval (T): C T
contains N subintervals. A linear assumption was also used by Tor-
(a) The structure survives up to the end of the time interval of
res and Ruiz [42] who considered that the structural capacity var-
interest.
ies linearly over the entire range of the interval under study;
(b) Intervals between inspections are equidistant.
however, in the present study this assumption is generalized to
(c) The damage is accumulated on the critical nodes.
N subintervals.
(d) The elements recover their fully capacity after repair.
According to the hypotheses above, Eq. (16) is solved as follows:
a br  
T r2  The expected total function cost given by Eq. (1), is expressed as
gFC;T ð0; DtÞ ¼ k exp 2
r2lnDjhmax þ r2lnC;T þ r2UT XC;T ð0; DtÞ
follows (using Eqs. (7), (11), and (15)):
a 2b
ð18aÞ C Total;DL ð0; DtÞ ¼ C I;DL ð0; DtÞ þ C R;DL ð0; DtÞ þ C F;DL ð0; DtÞ ð21Þ
where
where
" 1r #
aT b b Dt b
X
NI
XC;T ð0; DtÞ ¼ 1þ T 1 ð18bÞ
bT ðb  rÞ aT C I;DL ð0; DtÞ ¼ C Im jq;Dt  ecm ðDtÞgF;T m ð0;DtÞ ð21aÞ
m¼1
where XC,T(0, Dt) is a correction function for the expected number
of failures which considers the deterioration of the structural capac- X
NI X
n

ity in the subinterval T. C R;DL ð0; DtÞ ¼ C rj ;mjq;Dt  PðDj;m ðDtÞ

m¼1 j¼1

3.3. Considering that both structural capacity and structural demand  dÞ  ecm ðDtÞgF;T m ð0;DtÞ ð21bÞ
(for a given maximum wave height) vary in time (case 3)
X
NI b
X
N
For this case, it is considered that both structural capacity and C F;DL ð0; DtÞ ¼ C fm jDt gF;T m ðtk  tk1 Þecm ðDtÞgF;T m ð0;tk tk1 Þ ð21cÞ
structural demand vary in time. It is assumed that the median m¼1 k¼1
structural capacity varies linearly in each subinterval: C b ðsÞ ¼
where NI in the number of inspections which will be done over the
aT þ bT s and, simultaneously, the median structural demand, ex-
design life, DL, of the structure, at a number of constant time inter-
pressed in terms of the maximum wave height hmax, varies in time
b sÞ ¼ ða þ f  sÞ  yb . Recently, Tolentino et al. [43] proposed a vals NI = DL/Dt.
as: Dð
The fixed offshore steel jacket platform is located in Campeche
similar expression, but in this paper, it is generalized for different
Bay, Mexico. One of the platform internal frames was represented
subintervals, T; besides, the hypergeometric function is solved in a
by a simplified 2D model. An illustration of the steel frame ana-
simpler way, resulting the following equations:
 lyzed is shown in Fig. 1. The cross-section dimensions of the ele-
abr r2  
ments indicated in Fig. 1 are shown in Table 1. The expected
gFCD;T ð0; DtÞ ¼ k exp 2
r2lnDjhmax þ r2lnC;T þ r2UT XCD;T ð0; DtÞ
a 2b weight of the deck was assumed equal to 500 ton. The water depth
ð19aÞ at the site is 45.11 m. The structure is 48 m high. The structural
model takes into account the contribution of the piles inside the
where jacket legs to the lateral shear strength. In this study, extreme con-
  r ditions of wave, wind and sea currents (0%, 50%, and 95% of the to-
ba abT b
XCD;T ð0; DtÞ ¼ ½FðA; B; C; xÞ tal depth) were obtained from the Design and Evaluation Code of
bT ðb  rÞ af þ bT a
Marine Platforms located at Campeche Bay [45].
  r !
f bT Dt b
þFðA; B; C; xðDtÞÞ    1 þ 4.1. Fatigue analysis
bT a
   r   r #
b Dt a þ bT Dt b a b An offshore structure is continuously subjected to wave loads
1þ T ð19bÞ (due to operational and/or storm waves) along its entire life. The
a a þ f Dt a
waves may deteriorate the structure giving place to fatigue crack-
26 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30

+3.658 m The average crack size simulation process of the selected points
±0 m subjected to random load was performed using the Monte Carlo
technique. There were considered 25,000 samples. It was verified
-6.906 m that a larger number of samples did not change the statistical
results. Operational and storm waves were considered. The arrival
times between storms were assumed with exponential distribu-
-17.678 m tion. The storm wave heights followed a Gumbel distribution.

48.00 m
45.11 m
The statistical values applied for crack simulation are shown in
Table 2. The values are based on those used by Silva and
-31.394 m Heredia [47] for marine platforms located in Campeche Bay.
The mean crack growth under stochastic loads was calculated
using the following modified differential equation [48,49]:
0
da
-45.110 m ¼ CðDK mr Þm m0 ð22Þ
dt
Fig. 1. Simplified 2D model of the jacket platform. pffiffiffiffiffiffiffi
DK mr ¼ YSmr pa0 ð23Þ

where C and m are parameters related to the material properties,

Table 1
Cross sections of the elements.
DKmr is the mean stress intensity range, m0 is the rate of crossings
through zero with positive slope in a given time, Smr is the mean
Element Diameter (m) Thick (m) stress range of the random response of the elements [49], Y is a fi-
C, F, I, L, O, R 1.334 0.0159 nite geometrical correction factor [50], and a0 is the crack size. In Eq.
A, B, D, E 0.660 0.0159 (22), the random load was replaced by an equivalent cyclic load
G, H, J, K 0.559 0.0127
with amplitude and frequency expressed as functions of the mean
M, N 0.457 0.0159
P, Q 0.457 0.0127 properties of the original stochastic process.
S, T 0.508 0.0095 Substituting Eq. (23) in (22), the following expression is
U, X 1.334 0.0318 obtained:
V, W 0.457 0.0238
Z af pffiffiffiffiffiffiffi m
Y 0.406 0.0127
da=Y m ð pa0 Þ ¼ CSm 0
mr m t ð24Þ
a0

ing in the tubular joints. The critical nodes affected by fatigue were
selected on the basis of non-linear static analyses (‘‘pushover’’),
which were performed by decreasing the capacity of the selected
joint being analyzed and assessing its contribution to the global
capacity. The critical joints correspond to nodes 1, 2, 3, 4 and 6
(see Fig. 1). Two points were considered for each structural ele-
ment connected to the critical joints. The points correspond to
the zone of maximum and minimum stresses in the cross-section
of the element. Two points were analyzed for nodes 1 and 3,
whereas four points were considered for nodes 2, 4 and 6 because
these nodes are connected to two elements, as shown in Fig. 2. The
minimum and maximum stresses of the selected points were ob-
tained by subjecting the structure to several dynamic time history
‘‘step by step’’ analyses, using a set of simulated wave histories
associated with different return intervals, TR. The wave frequency
content is based on the Pierson–Moskowitz spectrum [46]. The
sea surface is represented by means of a homogeneous Gaussian
stationary process, which is expressed as a linear superposition
of regular waves with random generation in their phase angles
(with uniform distribution between 0 and 2p).
Fig. 2. Points considered for the fatigue analysis.
where a0 is the initial crack value, and af is the crack size after
N = m0 t cycles.
Considering that the joint’s cracking affects the resistance and
stiffness of the structural elements, some authors [51,52] recom-
mend to modify the capacity of the intact joint, Pk, by a linear cor-
rection factor, as shown in Eq. (25), where Pc is the capacity of the
cracked joint and Pk can be calculated by means of the API recom-
mendations [53]:
Pc ¼ Pk ð1  ðAcrack =Ajoint ÞÞ ð25Þ
Ajoint and Acrack are the areas of the cross-section and of the crack,
respectively, corresponding to all the elements connected to the
joint.
As an example, percentages of simulated cracked areas at the
end of different time intervals, corresponding to node 2, are shown
in Fig. 3.
4.2. Evaluation of the structural capacity over a time interval
The lateral structural capacity in terms of the maximum global
displacement of the jacket (corresponding to node 14 in Fig. 1) was
obtained by means of nonlinear ‘‘push-over’’ static analyses using
twenty different lateral load simulated profiles. These are related
to the acting forces corresponding to the simulated waves that pro-
duce the maximum base shear response. The capacity reduction at
the critical joints obtained from the fatigue analysis is considered
as damage condition. USFOS program [54] was used for the
non-linear analysis. Fig. 4 shows the median of the capacity, C bT ,
represented in terms of global displacement of the platform
considering cumulative damage, for different subintervals, T. Each
median structural capacity is fitted to the following expression:
b ðDtÞ ¼ a þ b Dt, T = 1, 2, 3 and 4. Fig. 4 shows that the median
C T
of the structural capacity does not follow a linear behavior in the
time interval (0, 15 years) analyzed. Table 3 shows that the
standard deviation of the natural logarithm of the structural
 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30 27

Table 2
Statistical parameters used for crack simulation.

Parameters m0 Smr a0 /c m* ln C* a0
Mean value As function of joint and time As function of joint and time 0.25 3.0 40.39 0.00011
Standard deviation As function of joint and time As function of joint and time – 0.3 0.69067 –
Distribution Lognormal Rayleigh – Normal Normal –
*
Correlation coefficient qlnC,m = 0.9.

100 function of the platform global displacement, for different

Ajoint/Acrackx100

80 intervals.
Fig. 5 shows that the median values of demand D b present varia-
60
tions for maximum wave heights larger than 16.7 m (wave heights
40
smaller than 16.7 m do not produce significant damage to the struc-
20
ture). Fig. 5 also shows that for waves greater than 16.7 m, higher
0 values of the median demand occur as the time interval becomes
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time (years) larger. This obeys to the nonlinear behavior of the structure. The va-
lue of the median structural demand, given a maximum wave
Fig. 3. Percentages of simulated cracked areas. Joint 2. height, is given by Dðb DtÞ ¼ ð3:75E  04 þ 5:0E  08  DtÞ  h2:0 (see
max
Section 3.3), and the standard deviation of the natural logarithm
of the demand, for a maximum wave height, is given by
(m)

0.260 rln Djhmax ¼ ð1:65E  03 þ 1:5E  05  DtÞ  h1:5

max .
0.250
Median of capacity

0.240 4.4. Expected number of failures over time

0.230
0.220 The expected numbers of failures at the end of different inter-
vals calculated with Eqs. (17a), (18a), (19a) and for the case with-
0.210
out structural damage are presented in Fig. 6. The epistemic
0.200
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
uncertainties associated with the structural demand, rUD, and with
the structural capacity, rUC, were assumed equal to 0.15. The
Time (years)
parameters k and r fitted to the wave hazard curve are equal to
b T , corresponding to four time subintervals.
Fig. 4. Medians of capacity, C k = 5.0E03 and r = 5, which correspond to a maximum wave height
equal to hmax = 23 m and to a return period of 1485 years.

capacity increases from the first subinterval (0.076) to the fourth
subinterval (0.218). The median of capacity, C b T , and the standard
deviation of its natural logarithm, rlnC,T correspond to the twenty
capacity curves mentioned above. It is noticed that as the time
interval gets longer, there is a higher probability that the structure
be loaded with more operational and storm waves, so the accumu-
lation of damage at critical joints increases and the response be-
comes highly nonlinear, giving place to more uncertainty in the
structural response.
4.3. Evaluation of the structural demand (for a given maximum height
wave) over an interval
The structural demand (given a maximum wave height) over
time, was evaluated by subjecting the structure to dynamic time
history ‘‘step by step’’ analyses, using a set of twenty simulated
time histories of waves associated with different maximum wave
heights (corresponding to different return periods, TR). The crack
growth simulations and the capacity reductions at selected joints
4.5. Expected cost of inspection, repair and failure
The expected costs of inspection, repair and failure for cases 1, 2
and 3 (see Sections 3.1–3.3) are shown in Figs. 7–9, respectively. It
is assumed that the design life of the structure is about 30 years,
and the annual discount rate is c = 6%.
The offshore platform handles an approximate production, Pmax
about 184,000 barrels of oil per day [55]. The cost of the drilling
equipment is estimated as 130,000,000 USD [55]. The inspection
cost using visual inspection technique is 3518 USD [56].
There are different repair strategies: welding, grinding and ele-
ment replacement, etc. In this study it is considered that the repair
will take place when the crack depth is greater or equal to a third of
the thickness of the plate (5 mm approx.), and the cost of welding
is 20,000 USD for each critical point [57].
The cost of failure includes costs due to equipment damage, pol-
lution, deferred production and indirect losses. The costs are
Median of demand, D̂ (m)

0.25 Without damage

considered to evaluate the capacity C b over time were also used Δt=3 years
in this section. Fig. 3 shows the median structural demand as a 0.20 Δt=6 years
Δt=9 years
0.15 Δt=12 years
Δt=15 years
Table 3 0.10
Standard deviations of the natural logarithm rlnC,T.
0.05
Time sub-interval (years) rlnC,T
0 6 Dt 6 4 7.60E02
0.00
0 5 10 15 20 25
4 6 Dt 6 8 1.65E01
8 6 Dt 6 11 1.99E01
h max (m)
11 6 Dt 6 15 2.18E01
b for different intervals.
Fig. 5. Median of demand, D,
28 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30

0.018 Equation 14 (without damage) The cost due to equipment damage is obtained by a linear ratio
Equation 15a (case 1)
Expected number of

0.016 between the global damage index and the drilling equipment. The
0.014 Equation 16a (case2)
cost is given by the following equation:
failures ηF,T

0.012 Equation 17a (case 3)

0.01 C ED ¼ D  C DE ð27Þ
0.008
0.006 The pollution cost is obtained considering the volume of oil spill
0.004 (related with a certain global damage level) by means of the fol-
0.002
lowing expression:
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
PC ¼ C OR  ðSA =10002 ÞE ð28Þ
Time (years)
where COR is the oil recovery cost (541.57 USD/h) with efficiency, E,
Fig. 6. Expected number of failures over a time interval.
of 0.81 h/km2 [58]; SA is the area of stain [59], which is obtained as
follows:
3.5E+06 Inspection cost 2 1
Repair cost SA ¼ 2:27ðDw  V s Þ3 t 2 þ 0:04ðDw  V s  U 4 Þ ð29Þ
3.0E+06
Failure cost
2.5E+06 Total cost where Dw is the density difference between crude oil and water; t is
Cost USD

2.0E+06 the oil spill time; U is the wind speed at height equal to 10 m, with
values Dw = 0.06341, U = 30 m/s and t = 2400 s. The volume of oil
1.5E+06
spill VS is obtained as a function of the global damage index as
1.0E+06
follows:
5.0E+05
0.0E+00 V s ¼ Pmax  D4 ð30Þ
0 2 4 6 8 10
Inspection interval (years) The deferred production cost CDP is given by the following
expressions [60]:
Fig. 7. Expected costs for case 1 (Section 3.1). Z Z
DL DLþRp
C DP ¼ RðsÞecðstÞ ds  RðsÞecðstÞ ds ð31Þ
t t
3.5E+06 Inspection cost
Repair cost RðsÞ ¼ C C  Pmax  U C  365 ð32Þ
3.0E+06 Failure cost
2.5E+06 Total cost
where Rp is the recovery time of product, equal to 4 years; CC is the
Cost USD

2.0E+06 crude oil cost, equal to 17 USD/barrel; UC is the profit of marketing

1.5E+06 the product, equal to 12% [31].
1.0E+06 The indirect losses cost, which takes into account agriculture,
5.0E+05
coal, gas, oil, minerals, petrochemical, basic chemical, tires, elec-
tricity and textile, is obtained as follows:
0.0E+00
0 2 4 6 8 10
C PI ¼ ðC F1 þ C F2 ÞðP max =Ptotal ÞD4 ð33Þ
Inspection interval (years)
where CF1 and CF2 are the first and second phase loss cost due to
Fig. 8. Expected costs for case 2 (Section 3.2).
structural damage, equal to CF1 + CF2 = 1.086  1010 USD; Ptotal is
the oil total production in Campeche bay equal to 2,100 000 barrels
6.0E+06 Inspection cost of oil per day [55].
Repair cost
5.0E+06 Failure cost
Total cost 4.6. Results
4.0E+06
Cost USD

3.0E+06 Figs. 7–9 show results of costs of inspection, repair, failure and
total costs for cases 1, 2 and 3, respectively. Figs. 7 and 8 show that
2.0E+06
the expected costs of inspection and repair have a significant influ-
1.0E+06 ence when intervals of time smaller than 4 years are considered.
0.0E+00 This is due mainly to the low probabilities of failure and to the high
0 2 4 6 8 10 number of inspections that must be made during the design life of
Inspection interval (years) the structure; however, as the interval between inspections in-
creases, the influence of the cost of failure becomes more signifi-
Fig. 9. Expected costs for case 3 (Section 3.3).
cant because the probabilities of failure become higher. Figs. 7
and 8 indicate that the optimal intervals for maintenance actions
obtained based on a global damage index, expressed in terms of are equal to 6 and 5 years, respectively. For the latter case (see
global displacement of the offshore structure as follows: Fig. 9), the optimal time interval results equal to 4 years.
Fig. 10 compares the expected total cost during the design life of
the structure for cases 1, 2 and 3 (which correspond to Sections
D ¼ ðd0  du Þ=ðd0  dd Þ ð26Þ
3.1–3.3, respectively). Fig. 10 shows that the optimal time intervals
of inspection and maintenance actions are equal to 6, 5 and 4 years
where d0 is the maximum global displacement of the structure for cases 1, 2 and 3, respectively. These intervals indicate that
without damage, du represents the maximum global displacements ignoring the contribution of the deterioration phenomenon on
that consider structural deterioration, and dd is the ultimate global the structural capacity, or on the demand, or in both, overestimate
displacement demanded by the structure. the intervals for inspection and maintenance actions.
 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30
6.00E+06 Total cost (section 3.1)
Total cost (section 3.2)
5.00E+06
Total cost (section 3.3)
4.00E+06
Cost USD
3.00E+06
2.00E+06
1.00E+06
0.00E+00
0 2 4 6
Inspection interval (years)
Fig. 10. Total costs for cases 1, 2 and 3 (Sections 3.1–3.3).
5. Conclusions
The influence of the structural deterioration over time on the
optimal time interval for inspection and maintenance of structures
was analyzed. The structural reliability is expressed by means of
simple closed-form mathematical expressions which are incorpo-
rated into the cost–benefit analysis. The criterion was applied to
a fixed steel jacket platform located in Campeche Bay, Mexico.
The results indicate that the variation in time of the structural
capacity (case 2) has a greater influence than the time variation
of the structural demand (case 1); e.g., it was obtained that the ex-
pected number of failures for an interval was higher when it was
considered only deterioration in structural capacity (case 2) than
when it was considered only variation in time of the structural de-
mand (case 1). As a consequence, the optimal interval for inspec-
tion and maintenance actions for case 2 resulted about 5 years.
Meanwhile, when only the deterioration in the structural demand
(for a given maximum wave height) was considered, then the opti-
mal time interval was about 6 years. It is concluded that for evalu-
ating the structural reliability of deteriorating structures, and for
making plans for inspection and maintenance actions, it is recom-
mended to take into account the variation in time of both the
structural capacity and the structural demand (for a given maxi-
mum wave height), otherwise the optimal time interval could be
overestimated.
The cost/benefit criterion presented, together with the closed-
form mathematical expression used, constitutes an efficient tool
for making inspection and maintenance plans for offshore struc-
tures; moreover, it can be improved to take into account different
type of inspection and maintenance actions formulated-within a
multiobjective optimization framework, in order to make decisions
for the most appropriate alternative.
Acknowledgements
Thanks are given to D. De León (formerly at IMP) for his valu-
able comments related to the reliability analysis of the marine
platform. This research project had the support of DGAPA-UNAM
(PAPIIT IN107011 and IN102114). The first author thanks CONA-
CYT for the economical support to develop his PhD research.
References
[1] Forsell C. Ekonomi och Byggnadsvasen (Economy and construction). Sunt
Fornoft 1924;4:74–7 [in Swedish].
[2] Freudenthal AM. Safety and the probability of structural failure. Trans ASCE
1956;121(1337):1375.
[3] Cornell CA. A probability-based structural code. J Am Concr I 1969;66(12):
974–85.
[4] Lind NC. Consistent partial safety factors. J Struct Div ASCE 1971;97(6):
1651–70.
[5] Rosenblueth E, Esteva L. Reliability basis for some Mexican codes. Probabilistic
design of reinforced concrete buildings. ACI Publication SP-31; 1972. p. 1–41.
29
[6] Ditlevsen O. Structural reliability analysis and the invariance problem. Report
no. 22. University of Waterloo, Solid Mechanics Division, Waterloo, Canada;
1973.
[7] Ang AH-S, Cornell CA. Reliability bases of structural safety and design. J Struct
Div ASCE 1974;100(9):1755–69.
[8] Hasofer AM, Lind NC. Exact and invariant second moment code format. J Eng
Mech Div ASCE 1974;100(1):111–21.
[9] Rosenblueth E, Mendoza E. Reliability optimization in isostatic structures. J
Eng Mech Div ASCE 1971;97(6):1625–42.
[10] Mau ST. Optimum design of structures with a minimum expected cost
criterion. Report no. 340. Cornell University, Department of Structural
Engineering, Ithaca, NY; 1971.
8 10
[11] Moses F. Design for reliability-concepts and applications. In: Gallagher RH,
Zienkiewicz OC, editors. Optimum structural design. New York, NY: John
Wiley; 1973. p. 241–61.
[12] Frangopol DM. A reliability-based optimization technique for automatic
plastic design. Comput Method Appl Mech 1984;44(1):105–17.
[13] Esteva L, Díaz OJ, García J, Sierra G, Ismael E. Life-cycle optimization in the
establishment of performance-acceptance parameters for seismic design.
Struct Saf 2002;24(2–4):187–204.
[14] Thoft-Christensen P. Assessment of the reliability profiles for concrete bridges.
Eng Struct 1998;20(11):1004–9.
[15] Frangopol DM, Kong JS, Gharaibeh ES. Reliability-based life-cycle management
of highway bridges. J Comput Civil Eng 2001;15(1):27–34.
[16] Kong JS, Frangopol DM. Life-cycle reliability-based maintenance cost
optimization of deteriorating structures with emphasis on bridges. J Struct
Eng ASCE 2003;129(6):818–28.
[17] Miyamoto A, Kawamura K, Nakamura H. Bridge management system and
maintenance optimization for existing bridges. Comput-Aided Civ Infrastruct
Eng 2000;15:45–55.
[18] Furuta H, Kameda T, Nakahara K, Takahashi Y, Frangopol D. Optimal bridge
maintenance planning using improved multi-objective genetic algorithm.
Struct Infrastruct Eng 2006;2(1):33–41.
[19] Liu M, Frangopol DM. Optimal bridge maintenance planning based on
probabilistic performance prediction. Eng Struct 2004;26:991–1002.
[20] Neves LC, Frangopol DM, Cruz PJS. Probabilistic lifetime-oriented
multiobjective optimization of bridge maintenance: single maintenance
type. J Struct Eng ASCE 2006;132(6):991–1005.
[21] Neves LC, Frangopol DM, Petcherdchoo A. Probabilistic lifetime-oriented
multiobjective optimization of bridge maintenance: combination of
maintenance types. J Struct Eng ASCE 2006;132(11):1821–34.
[22] Okasha NM, Frangopol DM. Lifetime-oriented multi-objective optimization of
structural maintenance considering system reliability, redundancy and life-
cycle cost using GA. Struct Saf 2009;31(6):483–9.
[23] Tolentino D, Ruiz SE. Time intervals for maintenance of offshore structures
based on multiobjective optimization. Math Probl Eng 2013. Article ID
125856: 15 pages.
[24] Kim S, Frangopol DM, Soliman M. Generalized probabilistic framework for
optimum inspection and maintenance planning. J Struct Eng ASCE
2013;139(3):436–47.
[25] Skjong R. Reliability based optimization of inspection strategies, vol. 3. Kobe,
Japan: ICOSSAR; 1985. p. 614–8.
[26] Madsen HO, Sørensen JD, Olsen R. Optimal inspection planning for fatigue
damage of offshore structures. ICOSSAR. San Francisco, USA; 1989. p. 2099–
106.
[27] Sørensen JD, Faber MH, Rackwitz R, Thoft-Christensen P. Modeling in optimal
inspection and repair. In: 10th International conference on offshore mechanics
and arctic engineering OMAE. Stavenger, Noruega; 1991. p. 281–8.
[28] Thoft-Christensen P, Sørensen JD. Optimal strategies for inspection and repair
of structural systems. Civil Eng Syst 1987;4:94–100.
[29] Straub D, Sørensen JD, Goyet J, Faber MH. Benefits of risk based inspection
planning for offshore structures. In: 25th International conference on
offshore mechanics and arctic engineering OMAE. Hamburg, Germany;
2006. p. 1–10.
[30] Faber MH. Reliability based inspection planning of fatigue damaged offshore
platforms. In: 1st International symposium on analysis of structural risk and
reliability for offshore facilities. IMP, Mexico; 2001.
[31] Ortega C, De León E. Development of a cost–benefit model for inspection of
offshore jacket structures in Mexico. In: 22th International conference on
offshore mechanics and arctic engineering OMAE. Cancun, Mexico; 2003. p.
87–98.
[32] Sørensen JD, Ersdal G. Risk based inspection planning of ageing structures. In:
27th International conference on offshore mechanics and arctic engineering
OMAE, vol. 3. Estoril, Portugal; 2008. p. 399–408.
[33] Madsen HO, Torhaug R, Cramer EH. Probability-based cost benefit analysis of
fatigue design, inspection and maintenance. Marine structural inspection,
maintenance and monitoring symposium. IIE1-IIE11, Arlington, Virginia;
1991.
[34] Visser W. POD/POS curves for non-destructive examination. Offshore
technology report 2000/018. UK: Health & Safety Executive; 2002.
[35] DNV. Guideline for offshore structural reliability analysis. Technical report no.
95–2018. Høvik, Norway; 1995.
[36] Rangel-Ramírez JD, Sørensen JD. Optimal risk-based inspection planning for
offshore wind turbines. Int J Steel Strut 2008;4:295–303.
[37] Moan T. Reliability-based management of inspection, maintenance and repair
of offshore structures. Struct Infrastruct Eng 2005;1(1):33–62.
30 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30
[38] Frangopol DM, Lin K, Estes A. Life-cycle cost design of deteriorating structures.
J Struct Eng 1997;123(10):1390–401.
[39] Jalayer F, Cornell CA. A technical framework for probability-based demand and
capacity factor design (DCFD) seismic formats. PEER report 2003/
08. Berkeley: Pacific Earthquake Engineering Research Center, University of
California; 2003.
[40] Cornell CA. Calculating building seismic performance reliability: a basis for
multi-level design norms. In: 11th World conference on earthquake
engineering. Paper 2122; 1996.
[41] Cornell CA, Jalayer F, Hamburger RO, Foutch DA. The probabilistic basis for the
2000 SAC/FEMA steel moment frame guidelines. J Struct Eng ASCE
2002;128(4):526–33.
[42] Torres MA, Ruiz SE. Structural reliability evaluation considering capacity
degradation over time. Eng Struct 2007;29:2183–92.
[43] Tolentino D, Ruiz SE, Torres MA. Simplified closed-form expressions for the
mean failure rate of structures considering structural deterioration. Struct
Infrastruct Eng 2012;8(5):483–96.
[44] Rainville ED. Intermediate course in differential equations. USA: John Wiley &
Sons Inc.; 1961.
[45] Petróleos Mexicanos. Diseño y Evaluación de Plataformas Marinas Fijas en la
Sonda de Campeche. NRF-003-PEMEX-2000. Mexico; 2000 [in Spanish].
[46] Pierson WJ, Moskowitz L. A proposed spectral form for fully-developed wind
sea based on the similarity law of S.A. Kitaigorodoskii. J Geophys Res
1964;69(24):5181–90.
[47] Silva FL, Heredia E. Effect of uncertainties on the reliability of fatigue damaged
systems. In: 23th International conference on offshore mechanics and arctic
engineering OMAE, vol. 2. Vancouver, Canada; 2004. p. 427–34.
[48] Paris P, Erdogan FA. Critical analysis of crack propagation laws. J Basic Eng
Trans ASME 1963;85:258–534.
[49] Sobczyk K, Spencer FB. Random fatigue: from data theory. Academic Press Inc.;
1992.
[50] Newman JR, Raju IS. An empirical stress intensity factor equation for the
surface crack. Eng Fract Mech 1981;15(1–2):185–92.
[51] Stacey A, Sharp JV, Nichols NW. Static strength assessment of cracked tubular
joints, 15th international conference on offshore mechanics and arctic
engineering OMAE, vol. 3; 1996. p. 211–24.
[52] Burdekin FM. The static strength of cracked joints in tubular members. Offshore
technology report 2001/080. London: Health and Safety, Executive; 2002.
[53] American Petroleum Institute. Recommended practice for planning, designing
and construction fixed offshore platforms – load and resistance factor design.
API RP 2a – LRFD. Washington, DC; 1993.
[54] Soreide TH, Amdahl J, Eberg E, Holmas T, Hellan O. USFOS – a computer
program for progressive collapse analyses of steel offshore structures. Report
STF71 F88038. Trondheim, Norway; 1993.
[55] Instituto Mexicano del Petróleo. Resumen ejecutivo de costos promedio de
estructuras típicas ubicadas en la sonda de Campeche. Departamento de costos
del IMP. México, DF; 1998 [in Spanish].
[56] Raine A. The development of Alternating Current Field Measurement (ACFM)
technology as a technique for the detection of surface breaking defects in
conducting material and its use in commercial and industrial applications. In:
15th World conference on non-destructive testing. Rome; 2000.
[57] The Marine Technology Directorate Limited. Review of repairs to offshore
structures and pipelines. Publication 94/102; 1994.
[58] Campos D, Rodriguez M, Martinez M, Ramos R. Assessment of consequences of
failure in jacket structures. In: 18th International conference on offshore
mechanics and arctic engineering OMAE. OMAE99/S&R-6015. St. Johns,
Newfoundland, Canada; 1999.
[59] Lehr WJ. Oil spill modeling and processes. In: Brebbia CA, editor. Review of
modeling procedures for oil spill weathering behaviour. WIT Press; 2001.
[60] Stahl B. Reliability engineering and risk analysis. In: McClelland B, Reifel MD,
editors. Planning and design of fixed offshore platforms. New York: Van
Nostrand Reinhold Co.; 1986.