Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
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.
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 23
structural demand (for a given maximum wave height). The idea by Visser [34]. Also, PoD curves for MPI and Eddy current tech-
behind inspecting an offshore structural system is to detect the niques can be obtained from the DNV guidelines [35]. PoD curves
presence and growth of size cracks, in order to perform the neces- can be represented by a logistics, exponential and lognormal mod-
sary repairs and maintenance to the structure later on. Several els. In the present study a PoD curve which is represented by the
authors have developed inspections and maintenance plans for off- following exponential model is used [36,37]:
shore structures: (a) based on risk and reliability of welded con- x a
min
nections subject to fatigue [25–27]; (b) using methodologies that PoDðxÞ ¼ 1 exp ; a > amin ð2Þ
k
take into account fatigue sensitivity analyses in steel joints [28];
(c) implementing simplified approaches and using practical design The inspection quality is given by the parameter k which is the
parameters such as fatigue design factors [29] and/or reserve mean size above amin (equal to 2 mm, [37]) of the smallest detect-
strength ratios [30]; (d) considering the damage caused by fatigue, able size. In the optimization analysis of the present study the fol-
buckling and dents on structural elements [31]; and (e) using lowing auxiliary measure of inspection quality is introduced:
Bayesian techniques [32]. q ¼ 1=k ð3Þ
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 q = 0 refers to no inspection, q = 1 is related to a perfect quality
structural capacity on the optimal time interval (based on a cost– inspection, while q 0.2–0.3 can be related to visual inspection.
benefit analysis) for inspection and maintenance of offshore struc-
tures. Three scenarios are considered: (1) structural demand (for a 2.2. Present value of the expected cost of inspection
given intensity) varies in time, while structural capacity remains
constant, (2) structural capacity deteriorates over a time interval, If a crack size is detected (given by Eq. (3)), the present value of
while structural demand remains constant, (3) both structural the expected cost of inspection during the time interval [0, Dt) can
capacity and structural demand vary simultaneously over time. be obtained as:
The optimal interval for inspection and maintenance corre- Z Dt
sponds here to the lowest total cost associated with inspection, re- C I ð0; DtÞ ¼ C I ðsjqÞ pS ½dðsÞ ecs ds ð4Þ
0
pair and failure, considering risk and reliability of the structure.
The difference between the present study and those mentioned where CI(s|q) is the cost of inspecting the structure in the instant s,
above is that in this paper reliability is expressed in terms of the for a given inspection quality q, it is considered that the structure
expected number of failures over a time interval by means of sim- with cumulative damage d has survived up to the instant s with a
plified closed-form mathematical expressions that take into ac- probability of pS[d(s)]; and ecs is a factor that converts the cost
count the structural deterioration over time. None of the to its present value, given a discount rate c.
methods above has solved the problem by means of this simplified Considering that the inspection of the structure will be done if a
approach. The mathematical expressions are then incorporated crack is detected at the end of the interval [0, Dt), and assuming
into the cost optimization formulation. Besides, it is demonstrated that the structure survives up to the end of Dt, Eq. (4) is simplified
that for estimating the optimal interval of inspection and mainte- as:
nance of a deteriorating structure, the variation in time of both C I ðDtÞ ¼ C Ijq;Dt ecðDtÞ pS ½dðDtÞ ð5Þ
structural capacity and structural demand (for a given maximum
wave height) must be taken into account (case 3), otherwise the where CI|q,Dt is the inspection cost at the end of [0, Dt) for a given
optimal time interval could be overestimated. 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)].
2. Expected total present value of the cost function over a time
On the other hand, if it is assumed that the occurrence of the
interval
structural failure corresponds to a non-homogeneous Poisson pro-
cess, then:
The expected total cost function is defined as the summation of
R Dt
the total expected cost, C, of inspection (I), repair (R), and failure (F) mF ðsÞds
pS ½dðDtÞ ¼ e 0 ¼ egF ð0;DtÞ ð6Þ
at the end of a time interval [0, Dt) as follows:
where mF ðsÞ represents the expected annual structural failure rate at
C Total ð0; DtÞ ¼ C I ð0; DtÞ þ C R ð0; DtÞ þ C F ð0; DtÞ 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
where C I ð0; DtÞ, C R ð0; DtÞ and C F ð0; DtÞ are the present value of the
section. Finally, the present value of the expected inspection cost,
expected cost of inspection, repair and failure, respectively. The
for a given inspection quality, at the end of the interval [0, Dt) is:
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 I ð0; DtÞ ¼ C Ijq;Dt ecðDtÞgF ð0;DtÞ ð7Þ
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.
2.3. Present value of the expected cost of repair
2.1. Quality of inspection (q)
Following an inspection, a decision must be made regarding if a
crack is detected. The repair decision will depend on the inspection
The inspection quality is related to the probability of detecting a
quality q [38]. If a crack is not detected, repair actions will not be
crack of a given size. That probability depends on the crack size, the
performed, in the case that a cracked is detected at the end of a
inspection method and the inspection team. The reliability with re-
time interval of interest, the present value of the expected cost of
spect to the probability of detecting a crack is defined by the prob-
repair within the time interval [0, Dt) is given by the following
ability of detection (PoD) curve which is identical to the
equation:
distribution function of the smallest detectable crack size [33].
Z Dt X
n
PoD curves for different inspections techniques (i.e. magnetic par-
ticle inspection [MPI], Eddy current, alternate current methods C R ð0; DtÞ ¼ C rj ½dðsÞjq pRj ½dðsÞjS pS ½dðsÞ ecs ds ð8Þ
0 j¼1
[ACFM and ACPD] and ultrasonic techniques) has been compiled
24 D. Tolentino, S.E. Ruiz / Engineering Structures 61 (2014) 22–30
where n is the number of elements to be repaired; C rj ½dðsÞjq is the 3. Evaluation of structural reliability
repair cost of element j having a cumulative damage equal to or
greater than d, at instant, s, given an inspection quality q; and In this paper, structural reliability is expressed by means of the
pRj[d(s)|S]ds is the probability of repairing the jth element with expected number of structural failures at the end of the time interval,
cumulative damage d, given that the structure has survived up to gF ð0; DtÞ, which is an extension of the expected annual structural
instant s. Eq. (8) considers that the structure with cumulative dam- failure rate concept (m(t), [39]). Based on the simplified approach
age d has survived up to the instant s with a probability of pS[d(s)]. proposed by Cornell and collaborators [40,41] to evaluate the reli-
When the damage level of the jth structural element is equal or ability of structures subject to seismic loads, and later on, extended
exceeds a certain threshold, the probability of repairing the by Torres and Ruiz [42], it is possible to obtain gF ð0; DtÞ as shown
element, is as follows: in Eq. (16). It is noticed that in the present study, the simplified ap-
proach by Cornell and collaborators [41] constitutes the basis to
PRj ½dðDtÞjS ¼ PðDj ðDtÞ dÞ ð9Þ evaluate the reliability of marine offshore structures subject to
wave loads, instead of the reliability of structures subjected to seis-
where P(Dj(Dt) P d) is the conditional probability that element j
mic loads. The use of the simplified approach is possible because
with certain level of cumulative damage D at time Dt, is higher or
the concepts involved in both types of structures (i.e., statistical
equal to an established damage level d.
description of the structural capacity and of the structural demand,
Considering that the damage has been accumulated in element j
simplified representation of environmental hazard curves, repre-
during the time interval [0, Dt), Eq. (8) is simplified as:
sentation of the structural deterioration by means of a simple
X
n mathematical expression, etc.), are similar.
C R ð0; DtÞ ¼ C rj jq;Dt PðDj ðDtÞ dÞ ecðDtÞ P S ½dðDtÞ ð10Þ Z " # r
j¼1
Dt b sÞ bðsÞ
Cð
gF ð0; DtÞ ¼ k
0 aðsÞ
where C rj jq;Dt represents the expected cost of repair of element j, at !
r2
the end of interval [0, Dt), given an inspection quality q, and it is 2 2 2
exp 2
r ln Djhmax ð sÞ þ r
ln C ð sÞ þ r sÞ
UT ð ds
considered that the structure with cumulative damage d has sur- 2b ðsÞ
vived up to the end of Dt with a probability of PS[d(Dt)].
ð16Þ
Assuming that the occurrence of repairing the elements corre-
sponds to a non-homogeneous process, and that the structure In Eq. (16) r2ln Djhmax ðsÞ and r2ln C ðsÞ represent the variances of the nat-
has survived up to the end of Dt (similar to Eq. (6)), then the ex- ural logarithms of the structural demand given a maximum wave
pected repair cost at the end of [0, Dt) can be obtained as follows: height hmax, and of the structural capacity, at instant s, respectively.
It is considered that these uncertainties present a lognormal type
X
n
distribution [5]. The expression r2UT ¼ r2UD þ r2UC represents the epi-
C R ð0; DtÞ ¼ C rj jq;Dt PðDj ðDtÞ dÞ ecðDtÞgF ð0;DtÞ ð11Þ
j¼1
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
2.4. Present value of the expected cost of failure 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
The present value of the expected cost of failure that can occur instant s, and presents a lognormal type distribution; a(s) and
within the interval [0, Dt) can be calculated as: b(s) are parameters defining the relation of the median structural
demand with the maximum wave height hmax, at instant s.
Z Dt In this study Eq. (16) is solved for three specific cases:
C F ð0; DtÞ ¼ C f ½dðsÞ pF ðsÞ ecs ds ð12Þ
0
1. Structural demand (for a maximum wave height) varies in
where Cf[d(s)] is the cost of the structural failure, given the level of time, while structural capacity remains time invariant.
cumulative damage d at instant s; pF(t) is the probability density 2. Structural capacity varies in time, while structural
function of waiting time to failure, and is equal to: demand remains time invariant.
3. Both structural demand and structural capacity vary in
Rt
mF ðsÞ time, simultaneously.
pF ðtÞ ¼ mF ðtÞe 0 ¼ mF ðtÞegF ð0;tÞ ð13Þ
Considering that the expected cost of failure and the present va- 3.1. Considering that structural demand for a given maximum wave
lue factor are constant values during the time interval of interest, height varies in time, while structural capacity is time independent
expression (12) becomes as follows: (case 1)
Z Dt
C F ð0; DtÞ ¼ C f jDt ecðDtÞ mF ðsÞegF ð0;sÞ ds ð14Þ Considering only that the structural demand varies in time, it is
0 necessary to assume that the median structural capacity, C b ðsÞ, and
its standard deviation, r2ln C ðsÞ, have constant values during the
Solving Eq. (14) for small intervals of time (e.g., Dt = 1 year), the b ðsÞ ¼ a and
time interval of interest, which means that C
present cost value of the possible structural failure within the
interval [0, Dt) is equal to:
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 .
b
X
N
C F ð0; DtÞ ¼ C f jDt gF ðtk tk1 ÞecðDtÞgF ð0;tk tk1 Þ ð15Þ It is important to notice that Cornell et al. [41] assumed the
k¼1 median structural demand (for a given intensity) given by
b ¼ a yb when structural demand is invariant in time (without
D
where N b is the number of time intervals of interest; gF ðt k t k1 Þ is considering structural deterioration).
the expected number of failures for the time intervals of interest, Taking into account the assumptions above, the expected num-
b years.
and t 1 ; t2 . . . tN ¼ 1; 2 . . . N 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
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 a0 is the initial crack value, and af is the crack size after
ing in the tubular joints. The critical nodes affected by fatigue were
N = m0 t cycles.
selected on the basis of non-linear static analyses (‘‘pushover’’),
Considering that the joint’s cracking affects the resistance and
which were performed by decreasing the capacity of the selected
stiffness of the structural elements, some authors [51,52] recom-
joint being analyzed and assessing its contribution to the global
mend to modify the capacity of the intact joint, Pk, by a linear cor-
capacity. The critical joints correspond to nodes 1, 2, 3, 4 and 6
rection factor, as shown in Eq. (25), where Pc is the capacity of the
(see Fig. 1). Two points were considered for each structural ele-
cracked joint and Pk can be calculated by means of the API recom-
ment connected to the critical joints. The points correspond to
mendations [53]:
the zone of maximum and minimum stresses in the cross-section
of the element. Two points were analyzed for nodes 1 and 3, Pc ¼ Pk ð1 ðAcrack =Ajoint ÞÞ ð25Þ
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 Ajoint and Acrack are the areas of the cross-section and of the crack,
minimum and maximum stresses of the selected points were ob- respectively, corresponding to all the elements connected to the
tained by subjecting the structure to several dynamic time history joint.
‘‘step by step’’ analyses, using a set of simulated wave histories As an example, percentages of simulated cracked areas at the
associated with different return intervals, TR. The wave frequency end of different time intervals, corresponding to node 2, are shown
content is based on the Pierson–Moskowitz spectrum [46]. The in Fig. 3.
sea surface is represented by means of a homogeneous Gaussian
stationary process, which is expressed as a linear superposition 4.2. Evaluation of the structural capacity over a time interval
of regular waves with random generation in their phase angles
(with uniform distribution between 0 and 2p). 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
Fig. 2. Points considered for the fatigue analysis. 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.
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)
capacity increases from the first subinterval (0.076) to the fourth 4.5. Expected cost of inspection, repair and failure
subinterval (0.218). The median of capacity, C b T , and the standard
deviation of its natural logarithm, rlnC,T correspond to the twenty The expected costs of inspection, repair and failure for cases 1, 2
capacity curves mentioned above. It is noticed that as the time and 3 (see Sections 3.1–3.3) are shown in Figs. 7–9, respectively. It
interval gets longer, there is a higher probability that the structure is assumed that the design life of the structure is about 30 years,
be loaded with more operational and storm waves, so the accumu- and the annual discount rate is c = 6%.
lation of damage at critical joints increases and the response be- The offshore platform handles an approximate production, Pmax
comes highly nonlinear, giving place to more uncertainty in the about 184,000 barrels of oil per day [55]. The cost of the drilling
structural response. 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-
4.3. Evaluation of the structural demand (for a given maximum height ment replacement, etc. In this study it is considered that the repair
wave) over an interval 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
The structural demand (given a maximum wave height) over is 20,000 USD for each critical point [57].
time, was evaluated by subjecting the structure to dynamic time The cost of failure includes costs due to equipment damage, pol-
history ‘‘step by step’’ analyses, using a set of twenty simulated lution, deferred production and indirect losses. The costs are
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
Median of demand, D̂ (m)
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
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
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 29
6.00E+06 Total cost (section 3.1) [6] Ditlevsen O. Structural reliability analysis and the invariance problem. Report
Total cost (section 3.2) no. 22. University of Waterloo, Solid Mechanics Division, Waterloo, Canada;
5.00E+06 1973.
Total cost (section 3.3) [7] Ang AH-S, Cornell CA. Reliability bases of structural safety and design. J Struct
4.00E+06
Cost USD
[38] Frangopol DM, Lin K, Estes A. Life-cycle cost design of deteriorating structures. [50] Newman JR, Raju IS. An empirical stress intensity factor equation for the
J Struct Eng 1997;123(10):1390–401. surface crack. Eng Fract Mech 1981;15(1–2):185–92.
[39] Jalayer F, Cornell CA. A technical framework for probability-based demand and [51] Stacey A, Sharp JV, Nichols NW. Static strength assessment of cracked tubular
capacity factor design (DCFD) seismic formats. PEER report 2003/ joints, 15th international conference on offshore mechanics and arctic
08. Berkeley: Pacific Earthquake Engineering Research Center, University of engineering OMAE, vol. 3; 1996. p. 211–24.
California; 2003. [52] Burdekin FM. The static strength of cracked joints in tubular members. Offshore
[40] Cornell CA. Calculating building seismic performance reliability: a basis for technology report 2001/080. London: Health and Safety, Executive; 2002.
multi-level design norms. In: 11th World conference on earthquake [53] American Petroleum Institute. Recommended practice for planning, designing
engineering. Paper 2122; 1996. and construction fixed offshore platforms – load and resistance factor design.
[41] Cornell CA, Jalayer F, Hamburger RO, Foutch DA. The probabilistic basis for the API RP 2a – LRFD. Washington, DC; 1993.
2000 SAC/FEMA steel moment frame guidelines. J Struct Eng ASCE [54] Soreide TH, Amdahl J, Eberg E, Holmas T, Hellan O. USFOS – a computer
2002;128(4):526–33. program for progressive collapse analyses of steel offshore structures. Report
[42] Torres MA, Ruiz SE. Structural reliability evaluation considering capacity STF71 F88038. Trondheim, Norway; 1993.
degradation over time. Eng Struct 2007;29:2183–92. [55] Instituto Mexicano del Petróleo. Resumen ejecutivo de costos promedio de
[43] Tolentino D, Ruiz SE, Torres MA. Simplified closed-form expressions for the estructuras típicas ubicadas en la sonda de Campeche. Departamento de costos
mean failure rate of structures considering structural deterioration. Struct del IMP. México, DF; 1998 [in Spanish].
Infrastruct Eng 2012;8(5):483–96. [56] Raine A. The development of Alternating Current Field Measurement (ACFM)
[44] Rainville ED. Intermediate course in differential equations. USA: John Wiley & technology as a technique for the detection of surface breaking defects in
Sons Inc.; 1961. conducting material and its use in commercial and industrial applications. In:
[45] Petróleos Mexicanos. Diseño y Evaluación de Plataformas Marinas Fijas en la 15th World conference on non-destructive testing. Rome; 2000.
Sonda de Campeche. NRF-003-PEMEX-2000. Mexico; 2000 [in Spanish]. [57] The Marine Technology Directorate Limited. Review of repairs to offshore
[46] Pierson WJ, Moskowitz L. A proposed spectral form for fully-developed wind structures and pipelines. Publication 94/102; 1994.
sea based on the similarity law of S.A. Kitaigorodoskii. J Geophys Res [58] Campos D, Rodriguez M, Martinez M, Ramos R. Assessment of consequences of
1964;69(24):5181–90. failure in jacket structures. In: 18th International conference on offshore
[47] Silva FL, Heredia E. Effect of uncertainties on the reliability of fatigue damaged mechanics and arctic engineering OMAE. OMAE99/S&R-6015. St. Johns,
systems. In: 23th International conference on offshore mechanics and arctic Newfoundland, Canada; 1999.
engineering OMAE, vol. 2. Vancouver, Canada; 2004. p. 427–34. [59] Lehr WJ. Oil spill modeling and processes. In: Brebbia CA, editor. Review of
[48] Paris P, Erdogan FA. Critical analysis of crack propagation laws. J Basic Eng modeling procedures for oil spill weathering behaviour. WIT Press; 2001.
Trans ASME 1963;85:258–534. [60] Stahl B. Reliability engineering and risk analysis. In: McClelland B, Reifel MD,
[49] Sobczyk K, Spencer FB. Random fatigue: from data theory. Academic Press Inc.; editors. Planning and design of fixed offshore platforms. New York: Van
1992. Nostrand Reinhold Co.; 1986.