You are on page 1of 31

October 10, 2000 9:4 WSPC/101-CEJ 00018

Coastal Engineering Journal, Vol. 42, No. 4 (2000) 357–387



c World Scientific Publishing Company and Japan Society of Civil Engineers

A RELIABILITY DESIGN METHOD OF CAISSON


BREAKWATERS WITH OPTIMAL
WAVE HEIGHTS

YOSHIMI GODA
Professor Emeritus, Yokohama National University, Ecoh Corporation,
2-6-4 Kita-uneno, Taito-ku, Tokyo 110-0014, Japan
E-mail : goda@ecoh.co.jp
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

HIROSHI TAKAGI
Ujima Construction Office, Chugoku Division, Penta Ocean Corporation,
Hiroshima 734-0016, Japan

Received 11 May 2000


Revised 23 August 2000

The paper presents an improved version of the deformation-based reliability design method
by means of expected sliding distance proposed by Shimosako and Takahashi (1998 and
1999). The process of economic optimization is added, and the concept of the optimal
return period for selection of design wave heights is introduced with a sample diagram for
the optimal return period for three values of service lifetime. It proposes a new tolerable
limit of expected sliding distance of 0.10 m instead of the value of 0.3 m adopted by
Shimosako and Takahashi. The ratio of water depth to the significant wave height is
demonstrated to govern the reliability design of caisson breakwaters, because of random
wave breaking effects. In shallow water, a breakwater can be designed with the cross
section smaller than that by the conventional design, but a larger cross section is needed
in deeper water. Discussions are made on the comparison between the partial safety factor
systems proposed by European groups and on the future problems to be solved for further
refinements of the deformation-based design method.

Keywords: Reliability design method, caisson breakwater, optimal design wave, optimal
return period, expected sliding distance, expected total cost.

1. Introduction
Breakwaters are daily exposed to waves, which grow and decay depending on storm
conditions. The statistics of storm waves can be derived from wave measurements
and/or hindcasting over many years. The probability of high waves is evaluated from
the extreme wave statistics. However, we cannot predict if a storm wave event of a
given height would occur in the next 50 years from now or not, because storms are
stochastic events. The situations are the same for many structures in the field such
as bridges, dams, dikes, and others. They are all subject to stochastic loadings.

357
October 10, 2000 9:4 WSPC/101-CEJ 00018

358 Y. Goda & H. Takagi

The conventional design method in civil engineering is to set a return period of


loading events, to select the design load with a given return period, and to design a
structure with a certain margin of safety. This is the deterministic design method.
Uncertainties in the magnitudes of loading on and resistance of the structure are
supposed to be covered by the safety margin.
A probabilistic method has been developed for structure designs since the 1970s
(e.g. Cornell, 1970). This is called the reliability design method. Application of the
reliability design method to breakwaters began in the mid-1980s. Toyama (1985)
and Suzuki (1987) discussed the safety of caisson breakwaters against sliding on
the basis of reliability design method. For rubble mound breakwaters, van der Meer
(1988) presented a probabilistic approach, and Burcharth (1991 and 1992) gave the
earliest treatment of reliability design methods. Since then, many studies have been
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

carried out on the applications of the reliability design method to vertical and mound
breakwaters.
Most of previous studies have a common feature that they take the design wave
height at a given return period as a fixed input data. The selection of return period
is an ambiguous and subjective process, however. Although the encounter proba-
bility by Borgman (1963) provides engineers with a certain measure for making a
decision, there is no definite criterion for the level of encounter probability to be
employed in structure designs. The computation of the expected sliding distance
(abbreviated as ESD hereinafter) proposed by Shimosako and Takahashi (1998 and
1999) quantitatively clarifies the risk of failure, but it also uses a predetermined
design wave height for a return period equal to the lifetime.
In contrast to these previous studies, the present paper proposes to design the
breakwater with the safety factor of 1.0 and to apply the reliability design method
by temporarily varying the design significant wave height over a range of return
period. Each design section is analyzed for the possibility of sliding by wave thrusts
with the method by Shimosako et al. (1994), and the sliding distances by individ-
ual events are accumulated over a specified lifetime of the breakwater. Statistical
variabilities of design factors are introduced by means of the Monte Carlo simula-
tions. The damage is evaluated with the accumulated sliding distance, with which
the cost of rehabilitation is estimated. The rehabilitation cost is added to the initial
construction cost to yield the total cost.
The results of the expected total costs of the breakwater sections for various
wave height levels are compared in search of the minimum total cost. The ESD over
the lifetime is also checked if it is below the tolerable limit in order to guarantee the
absolute safety of the breakwater design. The optimal design is defined as having
the minimum value of expected total cost while the expected sliding distance being
kept under the tolerable limit. The procedure presented herein would hopefully
shed a light upon the further development of reliability design method of vertical
breakwaters.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 359

2. Failure Mode of Caisson Breakwater


Vertical caisson breakwaters have several modes of failures as listed below. The
failure modes are listed in the order of frequency observed in Japan, which has a
large inventory of caisson breakwaters over several tens of years.

(1) Sliding of caissons.


(2) Displacement of concrete blocks and large rubble stones armoring a rubble foun-
dation mound, often associated with scouring in the sandy seabed in front of a
breakwater.
(3) Breakage and displacement of armor units in the energy-dissipating mound in
front of a caisson.
(4) Rupture of front walls and other damage on concrete sections of a caisson.
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

(5) Circular slip and other failures in the foundation and subsoil.

Although the safety against the overturning of a caisson around its heel is always
checked in the conventional design of vertical breakwaters, a pure form of overturning
never appears in reality because the failures in the foundation and subsoil occur
before the caisson is overturned. Foundation failures have been found in only a
few cases in Japan. It is probably owing to the Japanese practice that the bearing
pressure at the heel has been kept below 500 kPa or so in the conventional design.
Scouring in the seabed in front of breakwater is often referred to as a cause of
offshore tilting of the upright section such as occurred in the Mustapha jetty in
Alger Port in 1934. However, scouring in the seabed only causes deformation of the
rubble mound foundation as listed in the second item and not the tilting of caissons
as far as Japanese experience is concerned. Even for the case of Mustapha jetty,
Oumeraci (1992) suggested the foundation failure as a possible cause of the offshore
tilting.
Ruptures of caisson walls are sometimes reported as occurred under the ex-
ceptionally severe wave conditions. Generation of impulsive breaking wave forces,
wave concentration at a corner formed by two arms of breakwater and others are
cited as the cause of caisson rupture damage. However, the rupture damage is the
event that should be avoided through the improvement of the design methodology
for concrete members of breakwater caissons. It should be dealt with separately
from the overall design of breakwater sections with the reliability method. Breakage
and displacement of armor units listed as the third item are the problems simi-
lar to those of mound breakwaters, and thus they are not discussed in the present
paper.
Because the sliding of caissons comprises the majority of the cases of breakwater
damage, the present paper takes the caisson sliding as the principal failure mode
of breakwaters and develops a reliability design method of caisson breakwaters.
Inclusion of the foundation and subsoil failures will be the subject for future studies.
October 10, 2000 9:4 WSPC/101-CEJ 00018

360 Y. Goda & H. Takagi

3. Selection of Reliability Analysis Method


3.1. Category of reliability design method
The reliability design method is classified into three categories, depending on the
level of probabilistic concepts being employed. The method with Level 1 is the limit
state method, which employs the partial safety factors. It is also called the load and
resistance factor design (LRFD) method. The values of external loads are increased
by multiplying them with the partial safety factors, while the values of resistance
strength are reduced by dividing them with their partial safety factors. Both the
factors are assigned the values equal to or greater than 1.0, in consideration of
the probabilistic behaviors of loads and resistances. Evaluation of the partial safety
factors is the crucial point in the success of the reliability method with Level 1.
Once the partial safety factors are determined, the design calculation proceeds in a
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

deterministic manner.
In the Level 2 reliability design method, the load and resistance are assumed to
distribute normally. Then the safety index β is defined as follows:

mL − mR
β=q (1)
σL2 + σR
2

where mL , mR , σL , and σR denote the means and standard deviations of the load
and resistance, respectively.
The safety index is directly related with the probability of failure (Pf ) such as
Pf = 0.001 at β = 3.08, while β = 2.0 is equivalent to Pf = 0.023, according
to the probability table of normal distribution. Although the probability of fail-
ure can be evaluated with the safety index, a question remains as how to set the
probability level to be employed in the structure design. Unless some regulatory
body specifies the level of acceptable probability of failures to guarantee public
safety, individual engineers must choose the safety level for the structures under
design.
In the Level 3 reliability design method, all the load and resistance factors are
described with the respective probability density functions. The probability of fail-
ures is calculated without assumptions of normal distributions. The deformation-
based reliability design (abbreviated as DBRD) method proposed by Shimosako and
Takahashi (1998 and 1999) belongs to the category of Level 3. They estimate the ex-
pected sliding distance (ESD) of a caisson over its lifetime through the Monte Carlo
simulations by taking the uncertainties of various design factors into computation.
They have set the tolerable limit of the expected sliding distance at SE = 0.30 m,
though they gave no explanation for the reason of selecting that particular value:
they must have refered to various field reports of breakwater damage before they
made the decision.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 361

3.2. Level 2 method with partial safety factors


The reliability design methods with partial safety factors have been used in many
branches of civil engineering. For the design of mound and vertical breakwaters,
European engineers have been trying to promote the system of partial safety factors.
Burcharth and Sφrensen (1998) have reported the recommended values for vertical
breakwaters, and Burcharth and Sφrensen (1999) have summarized the results of
PIANC working group for establishing the partial safety factors for both rubble
mound and vertical breakwaters.
They have categorized the partial safety factors for vertical breakwaters with the
six modes of failures; i.e. foundation failure in sandy subsoil, foundation failure in
clayey subsoil, sliding failure, overturning failure, scour at circular roundheads, and
rock armor failure at toe berm. The partial safety factors against sliding failure are
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

made up of two factors: one to be applied to the wave height as representative of


the load, and the other to the friction coefficient as of the resistance. These partial
safety factors have been assigned particular values depending on the five levels of
the safety index β, for the two classes of reliable and less-reliable sets of wave data.
There are four tables of the partial safety factors in combination of the water depth
and model test verification; i.e. deep and shallow water, and with and without model
tests performed. In total, Burcharth and Sφrensen (1998 and 1999) give 40 sets of
the partial safety factors for wave height and friction coefficient for the failure mode
of sliding. They regard their system to belong to the reliability design method with
Level 2 because of the inclusion of the safety index.

3.3. Level 3 deformation-based reliability design method by means


of expected sliding distance
The three levels of reliability design method are all characterized with the probabil-
ity of failure. However, the concept of failure probability is not easy to understand
when one tries to interprete its meaning in daily design works. In contrast to it, the
concept of the expected sliding distance (ESD) is more understandable, because the
ESD gives a clear image of possible magnitude of breakwater damage. Therefore,
the present paper adopts the essential part of the DBRD method by Shimosako and
Takahasi (1998 and 1999). However, their methodology lacks the economic analysis
for optimal designs. It also leaves the concept of return period not being clarified.
Under their method, the design of a breakwater is calculated by varying the safety
factor until the expected sliding distance comes just below the tolerable limit. Ac-
tually, employment of a high value of safety factor to maintain the expected sliding
distance at a low value is interpreted as equal to the adoption of a large design wave
height and long return period when using a constant value of safety factor. Never-
theless, Shimosako and Takahashi do not specifically mention about the selection of
design wave heights.
October 10, 2000 9:4 WSPC/101-CEJ 00018

362 Y. Goda & H. Takagi

The present paper introduces the economic optimization and wave height evalu-
ation processes into the expected sliding distance method for further refinements of
the latter. It also uses the service lifetime of a breakwater as the input. Thus, the
present method can also solve a question as to how to set the optimal design wave
height for temporary breakwaters to be scheduled to function for a short time span.
Several new insights on caisson breakwater designs have been obtained as discussed
in the following sections.

4. Design Factors and Computational Procedures


4.1. Extremal distribution of offshore significant wave heights
The design of caisson breakwater is affected by various factors. The most important
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

factor is the design wave height in the offshore. The wave period associated with it is
also important. The offshore wave height is converted to the wave height in front of
the breakwater through the wave transformation analysis for the processes of shoal-
ing, breaking, refraction, diffraction, and so on. These processes are influenced by
the sea bottom topography, water depth, tide level, and other factors. Many of these
factors are subject to statistical variation around their mean values. Introduction
of such variability into design process is the essential part of the reliability design
method.
First, the variability of storm wave heights is discussed. The design process of
caisson breakwaters begins from the analysis of extremal statistics of storm wave
heights in the particular region. The design wave height is usually determined by re-
ferring to the extreme wave height distribution, which is constructed with the data of
extreme waves collected through instrumental wave observations and/or hindcasting
projects. The extreme distribution of wave heights is subject to the statistical un-
certainty because of sample variablity, and the return wave height estimated from
an extreme distribution must be treated as having a certain range of confidence
interval.
The Level 3 DBRD method for breakwaters needs the specification of a distribu-
tion function for the annual maximum heights or storm peak heights of significant
waves. In the present study, the following Weibull distribution function is employed
for the annual maximum wave heights:
(   )
(x − B) k
F (x) = 1 − exp − (2)
A

where x stands for the annual maximum significant wave height, A and B are the
scale and location parameters, respectively, and k is the shape parameter which has
been fixed at the value k = 1.0.
The specific values of distribution parameters depend on the severity of storm
wave conditions in a particular region. As a measure of wave severity, the return
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 363

Table 1. Scale and location parameters of the Weibull distribution (k = 1.0) employed for offshore
significant wave heights.

Nominal design height Scale para. Location para. Wave Period


(H1/3 )0D (m) A (m) B (m) (T1/3 )0D (s)

5.0 0.80 1.87 9.1


6.0 1.00 2.09 10.0
7.0 1.30 1.91 10.8
8.0 1.50 2.13 11.5
9.0 1.70 2.35 12.4

wave height with the return period of 50 years is designated here as the nominal
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

design height with the symbol (H1/3 )0D . The scale and location parameters have
been assigned the values listed in Table 1 for several nominal design heights. These
parameter values were selected by referring to those of extreme distributions of
peak wave heights at various stations around Japan. The significant wave periods
of nominal q
design waves have been set so as to yield the wave steepness of 0.039, or
T1/3 = 4.08 (H1/3 )0D . The value of 0.039 was arbitrary chosen within the range of
wave steepness prevailing among storm waves.
When the extreme distribution of peak wave heights at storm events has been
obtained by the peaks-over-threshold method, the extreme distribution of annual
maximum heights is derived from the former with the following formula (e.g., Goda
2000, p. 412):

F (x) = exp{−λ[1 − F∗ (x)]} (3)

where λ denotes the average number of peak waves per year and F∗ (x) is the dis-
tribution function of peak wave heights. In the present study however, the above
conversion process has been neglected for the sake of simplicity.
When the Weibull distributions with k = 1 and other parameters listed in Ta-
ble 1 are used for the annual maximum distributions, they seem to represent the
distributions that spread over quite a wide range of wave heights. However, the main
objective of present study is to exemplify the usefulness and applicability of a new
approach of the reliability design method for caisson breakwaters. Thus the results
of numerical computations with the distribution of Eq. (2) are described without
modifications in the present paper.

4.2. Computation procedure for optimal breakwater design


The procedure of computation for selecting the optimal breakwater cross section
is illustrated in Fig. 1. First, a preliminary design significant wave height is picked
up by referring to the nominal design wave height. This is a parameter for trial
October 10, 2000 9:4 WSPC/101-CEJ 00018

364 Y. Goda & H. Takagi


by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 1. Block diagram of computation procedure for optimal breakwater design.

designs of breakwaters. The preliminary wave height is varied over a certain range
by choosing the return period in a range of 0.5 to 2.0 times the service lifetime.
If the optimal breakwater design is not obtained within the initial range of wave
height, then the range should be expanded until the optimal solution is reached.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 365

With each preliminary wave height, a cross section of caisson breakwater is


designed by the deterministic, conventional method with the safety factor against
sliding being set at S.F. = 1.0. The wave pressure and uplift are calculated by using
the Goda formula with inclusion of the impulsive pressure coefficient proposed by
Takahashi et al. (1994). In the present study, the following conditions are set for
computation of cross sections except for the model case in Sec. 8:
(1) Caisson crest height above high water level: hc = 0.6H1/3
(2) Thickness of rubble mound foundation: t = max{0.2h, 3.0 m}
(3) Height of foot-protection block: t0 = 1.5 m
(4) Rear berm width of rubble mound: b = (1.0 ∼ 1.5)(H1/3 )0D
(5) Mean density of upright section: ρc = 2100 kg/m3 .
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

In the above, h denotes the water depth below the high tide level and max{a, b}
takes the larger value of a or b.
The design crest height of hc = 0.6H1/3 is based on Japanese practice, which
began in the 1960s after its adoption as a recommendation in the design stan-
dard for port and harbor structures. Specification of the thickness of rubble mound
foundation and the height of foot-protection concrete block is needed for com-
putation of wave pressure and uplift by the Goda formula. The berm width of
rubble mound foundation is set at a predetermined value within the range listed
above. When a caisson is slid many times and its center is displaced beyond the
rear end of the mound berm, the caisson is judged as fallen from the mound.
The upright section of a caisson breakwater consists of two parts of the crown
concrete and the sand-filled, reinforced-concrete caisson. The density of the for-
mer part is about 2300 kg/m3 , while that of the latter is about 2100 kg/m3 .
However, the density difference is neglected in the present study for the sake of
simplicity.
The initial construction cost is estimated for the resultant cross section. In the
present study, only the cost of the upright section is taken into consideration with
the unit cost of Japanese Yen 100,000/m3 (about US$950/m3 ). It will be necessary
to make more detailed cost analysis when the present method is applied for actual
breakwater designs.
The designed cross section of breakwater is then subjected to simulated yearly
storm waves over the duration of its service lifetime L. For each simulated yearly
storm, the total sliding distance is calculated. The process of this calculation is
represented as SUBROUTINE A and explained in the next subsection. During the
L years of cyclic calculation, the values of total sliding distance by yearly storms are
accumulated. A rehabilitation cost is evaluated on the basis of accumulated sliding
distance, as discussed in Sec. 6.
The process of L-year cycles is repeated 5000 times, and the total costs (sum of
initial and rehabilitation costs) and accumulated sliding distances thus obtained are
added together to yield the expected values of total cost and sliding distance. The
October 10, 2000 9:4 WSPC/101-CEJ 00018

366 Y. Goda & H. Takagi

number of 5000 times is employed here by referring to Shimosako and Takahashi


(1998 and 1999), who have demonstrated that the coefficient of variation of ESD
decreases less than 3% by doing so.
Through the procedure described above, the relationship between the prelim-
inary wave height and expected total cost can be derived. When the preliminary
wave height for design works is set low, the initial cost is kept low but the rehabil-
itation cost will be high. When the preliminary height is set high, the initial cost
becomes high but the rehabilitation cost will be low. By plotting the expected total
cost against the wave height, we could find the minimum value of total cost. The
breakwater at that condition is the optimal design and the wave height at that point
is called the optimal design wave height. The return period corresponding to that
wave height is the optimal return period.
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

4.3. Introduction of random variations of design factors


The basis of the reliability design method is a due consideration of the stochastic
behaviors of loads and resistances. As described in Sec. 3, the present study employs
the method with Level 3, which introduces the probability density functions of design
factors. Table 2 lists the design factors employed in the present study and their
distribution functions.
Use of random variation of design factors is explained below in conjunction with
the computational block diagram sketched in Fig. 2. First, an annual maximum sig-
nificant wave height is randomly sampled from the extremal distribution function
in the particular region. This process is carried out without any regard to the pre-
liminary wave height for designing a cross section of breakwater, as the latter is a
nominal parameter with little implication of probability. The randomly sampled off-
shore significant wave height is further given a stochastic variation with the normal
distribution. This variation is a representation of the uncertainty in the estimate
of extremal distribution function owing to the limited sample size of extreme wave
data: e.g. see Fig. 11.9 in Goda (2000, p. 406) for an example of extremal wave

Table 2. Characteristics of variations of design factors.

Design Factors Bias Coef. Var. Distr. Funct. Remarks

(H1/3 )0 0 0.10 Normal Mean by extrm. dist.


Water level — Tide amplit. * Triangular —
Wave deformat. −13% 0.10 Normal —
Friction coef. +6% 0.10 Normal µ = 0.6 as the base
Hindiv. * * Rayleigh 2-hours duration
Wave forces −9% 0.10 Normal —

Note: * indicates “not applicable here.”


October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 367
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 2. Details of block diagram for computation of total sliding distance.

analysis. Statistical variability of measured data (Goda, 1988) and inaccuracy in


wave hindcasting also contribute to the above stochastic variation. In the present
study, the coefficient of variation of 10% has been set for this purpose as represen-
tative of the extreme wave data analysis around Japan.
Offshore waves undergo the transformation processes of shoaling, breaking, re-
fraction, and others during their propagation toward the design site. The present
study assumes the situation of waves normally incident to a straight coast with
parallel depth-contours; no effects of refraction and directional spreading are con-
sidered. Wave attenuation by random breaking is evaluated with the model by Goda
(1975). According to Takayama and Ikeda (1993), the current method for prediction
of random wave transformation has a tendency of overestimating wave heights in
shallow water. The bias of the mean is reported as −6% for the significant height
and −13% for the maximum height, respectively. As the maximum wave height gov-
erns the design of breakwater cross section, the mean bias of −13% is adopted in
the present study. The coefficient of variation is assumed to be 10%, same as that
adopted by Takayama and Ikeda.
October 10, 2000 9:4 WSPC/101-CEJ 00018

368 Y. Goda & H. Takagi

The variation of water level by tides is represented with a triangular distribution,


which extends downward from the design water level to the depth equivalent to the
tidal range. Numerical simulations in the present study are carried out with the
tidal range of 0.5 m and 1.0 m. Strictly speaking, the tide level should be computed
with the major four tidal components as done by Kawai et al. (1996). They have
computed annual fluctuations of tide levels at 96 around Japan and have shown the
probability distributions of tide levels. Most of the results indicate the triangular
distribution being a favorable approximation. The use of triangular distribution in
the present study has been made to save computation time.
The coefficient of friction is known to vary stochastically. Takayama and Ikeda
(1993) have reported the mean bias of +6% based on their analysis of previous pro-
totype test data. This bias is also employed in the present study, but the coefficient
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

of variation of 10% is tentatively assumed here.


The Goda formula for wave pressure prediction has been reported to have a
tendency of overestimation. By referring to Takayama and Ikeda (1993), the bias of
−9% is assigned to the mean value and the coefficient of variation is taken as 10%.
The cross section of breakwater designed with a preliminary design wave height is
checked on the possibility of sliding. The check is made for every individual wave so
that a distribution of individual waves must be introduced. Referring to Shimosako
and Takahashi (1998 and 1999), a peak of storm waves is assumed to continue for
2 hours. This assumption of duration reflects the Japanese practice of instrumental
wave observations by every 2 hours. The heights of individual waves are assumed to
follow the Rayleigh distribution in the offshore condition. The periods of individual
waves are set equal to the significant wave period. Individual waves are subjected to
the transformations of shoaling and breaking, which are represented with the linear
shoaling coefficient and the depth-limited breaker process, respectively, for the sake
of simplicity. The resultant individual wave heights are then given the stochastic
variation for wave transformations listed in Table 2.

4.4. Computation of sliding distance


The wave thrusts and uplifts are computed for individual waves as before. The
results are further given the stochastic variation according to Table 2 to simulate
random variations in the nature. With these wave loads, the possibility of caisson
sliding is examined for individual waves. Sliding occurs when the following condition
is met:

Pmax > µ[g(Mc − Mw ) − U ] (4)

where Pmax denotes the peak value of wave thrust upon the caisson, µ the coefficient
of friction, g the acceleration of gravity, Mc the mass of caisson, Mw the mass of
water displaced by the caisson, and U the uplift exerted upon the bottom of caisson.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 369

The distance of caisson sliding is estimated with the model presented by Shi-
mosako et al. (1994). The model assumes a triangular time-history of wave thrust
variation with a short duration, which simplifies the pattern of breaking wave pres-
sures. Shimosako and Takahashi (1998 and 1999) later improved the model by adding
a sinusoidal thrust variation to represent standing wave pressures. However, the com-
posite model demands much longer computer time to estimate the sliding distance
than the triangular model does. Thus, the earlier model with a triangular variation
is employed in the present study. According to this model, the time variation of
wave thrust is expressed as follows:

 2t τ0

 Pmax :0≤t≤

 τ 2
 0 
P (t) = 2 1 − t P τ0
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

< t ≤ τ0 (5)

 max :

 τ0 2


0 : τ0 < t

The time variation of uplift is described with the form same as in Eq. (5).
The term τ0 in Eq. (5) represents the duration of triangular wave thrust and is
related to the wave period as below.

τ0 = k0 τ0F (6)

where the constant k0 and the time τ0F are given by

k0 = [(α∗ )0.3 + 1]−2 (7)


   
H
τ0F = max 0.4T, 0.5 − T (8)
8h

in which α∗ is the wave pressure coefficient which takes the larger value of the
parameter α2 by the Goda formula (1974) or the impulsive pressure coefficient αI
by Takahashi et al. (1994).
The sliding distance when Eq. (4) is satisfied is derived after some manipulation
under the assumption of a rigid body motion as below.

τ02 (FS − µWe )3 (FS + µWe )


S= (9)
8µMc We FS2

where FS = Pmax + Umax and We = g(Mc − Mw ).


For individual waves during 2-hours at the peak of annual maximum storm, a
check is made if sliding of the caisson takes place. If sliding occurs, the distance of
sliding is calculated and the result is accumulated. By this process, the total distance
of caisson sliding by an annual maximum storm is obtained.
October 10, 2000 9:4 WSPC/101-CEJ 00018

370 Y. Goda & H. Takagi

5. Evaluation of Sliding Failure Probability under Conventional


Design Method
5.1. Probability of sliding failure and expected sliding distance
The present method of analysis enables to examine the probability of sliding failure
of the breakwaters designed by the conventional method. Instead of introducing a
preliminary design significant wave height, the nominal design wave is used to cal-
culate a cross section of breakwater. The safety factor is set as S.F. = 1.2 according
to the conventional method under the design condition given in Sec. 4.2. The evalu-
ation procedure is obtained by modifying the block diagram of Fig. 1. The top two
boxes of the extremal distribution function and preliminary design wave are taken
out, and the bottom box of the optimal breakwater section is omitted together with
the outer repetition cycle. The middle cycle for 5000 repetitions and the inner cycle
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

for L-years are maintained.


The event of sliding failure is defined as a single or multiple occurrences of
sliding during an annual maximum storm. A single check with Eq. (4) against the
highest wave Hmax in front of the breakwater suffices for the purpose of examining a
sliding failure. If a designed section of breakwater indicates sliding by any maximum
storm during a cycle of 50 years lifetime, that particular cycle is termed as having a
sliding failure. By counting the number of L-year cycles having sliding failures Nf ,
the probability of sliding failure is defined as Pf = Nf /5000 in the present study.
The expected sliding distance (ESD) is also computed by the procedure described
in Sec. 4.
Examples of the sliding failure probability are shown in Fig. 3. Five levels of the
offshore design significant wave heights ranging from 5.0 m to 9.0 m listed in Table 1
are used to calculate the sliding failure probability. The water depth below the high
tide level varies from 10 m to 26 m. The calculation indicates the sliding failure
probability ranging from 2 × 10−3 to 10−1 . Kawai et al. (1997) examined various
reports of breakwater damage in Japan and concluded that the mean rate of sliding
failure per year is 7 × 10−4 and the probability of sliding failure over 50 years can be
estimated as 3 × 10−2 . Thus, the results shown in Fig. 3 are in agreement with the
performance of prototype breakwaters. It can be taken as an evidence supporting
the validity of the present method of analysis.
The ESD of the breakwaters designed by the conventional method are shown in
Fig. 4. The design conditions are the same as those employed in Fig. 3.
Both the sliding failure probability and ESD increase as the water depth in-
creases. In shallow water, the wave thrust has an upper limit because the maximum
wave height is restricted by the water depth. In deeper water, the limitation is re-
laxed and a possibility of larger wave thrusts exerted upon a breakwater increases.
The variations in the sliding failure probability and ESD reflect the effect of ran-
dom wave breaking process on breakwater stability. The maximum sliding distance
appears at the condition of h ' 3.5(H1/3 )0D . In the water deeper than that, the
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 371
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 3. Estimated probability of sliding failure of breakwaters designed by the conventional method
over 50 years.

Fig. 4. Expected sliding distance of breakwaters designed by the conventional method over 50 years.
October 10, 2000 9:4 WSPC/101-CEJ 00018

372 Y. Goda & H. Takagi

probability and sliding distance decreases. In such deep water, the bearing capac-
ity of the foundation would become the critical design factor, and the danger of
breakwater damage would increase in reality.
The maximum value of ESD increases as the offshore design wave height in-
creases. The results of Figs. 3 and 4 give a warning to engineers that more caution
should be paid when designing breakwaters in deeper water than in shallower water
area.

5.2. Influence of design parameters on sliding failure


The degree of sliding failure is influenced by various design factors other than the
design wave condition and water depth. Computations have been carried out by
varying the values of several desing factors. First, two additional bottom slopes of
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

1/20 and 1/100 have been tested for the nominal design wave height of (H1/3 )0D =
6.0 m. As the sea bottom slope increases, the probability of sliding failure increases
in shallow water area, because the breaker height increases with the sea bottom
slope. The ESD is found little affected, however.
In addition to the tidal range of 1.0 m in Figs. 3 and 4, two tidal ranges of
0.5 m and 2.0 m have been tested. Increase in the tidal range tends to decrease the
probability and distance of sliding. As the present study takes the water level below
the high tide level as the reference, a large tidal range increases the possibility of
smaller water depth and thus lesser wave thrusts.
For the examination of the effect of wave periods, the period for waves with
(H1/3 )0D = 6.0 m has been varied from 9.8 s to 11.0 s and 12.2 s. With the increase
in the wave period, the probability and distance of sliding increase; the latter increase
is more pronounced than the former. The wave period effect on breakwater damage
by sliding is caused by the fact that both the wave pressure intensity and the wave
thrust duration (τ0 ) increase with the wave period getting longer.

6. Economic Optimization of Breakwater Design


Economic optimization is the objective to be achieved in any structure design. Voort-
man et al. (1998) have discussed an economic optimization of vertical breakwaters
in details, by taking various modes of failure into analysis. The subsoil failure was
the dominant factor in their design case, but such a failure is rare in Japanese ex-
perience as discussed in Sec. 2. In the present study, the failure mode of sliding is
considered in economic optimization.
The objective function is the total cost inclusive of the probability of failure.
A sliding of breakwater caissons over a few meters itself does not cause immediate
damage on port operation in a short term. Thus, the cost of sliding damage can
be evaluated by the cost of rehabilitation of breakwaters only. Because there is no
guideline for the estimation of rehabilitation cost, three simple models for cost esti-
mation are employed here as shown in Fig. 5. The cost of rehabilitation is normalized
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 373
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 5. Three models for estimating rehabilitation cost as a function of total sliding distance.

with the initial construction cost and plotted in the ordinate. The abscissa is the
accumulated distance of sliding. The distance Smax is the threshold sliding distance
beyond which a caisson is judged as fallen from the mound and defined by
Smax = b + B/2 (10)
where b is the rear berm width of rubble mound foundation and B is the caisson
width. When the accumulated sliding distance reaches Smax , it is assumed that the
damaged caisson is destroyed and the breakwater is rebuilt with the full amount
equal to the initial cost.
Case 1 assumes a linear increase of rehabilitation cost beyond the threshold
distance of 0.30 m, below which no rehabilitation work is made. This threshold
value is based on the authors’ subjective judgment, and it should not be taken as
the same as the tolerable limit of SE = 0.30 m by Shimosako and Takahashi (1998
and 1999). Case 2 employs a parabolic increase of rehabilitation cost with the sliding
distance. In Case 3, the rehabilitation cost increases rapidly with the sliding distance
and approaches toward the initial construction cost gradually. Actual damage on
breakwaters by storm waves takes various forms and the degree of rehabilitation
depends largely on the judgment of engineers in charge of inspection. The three
models depicted in Fig. 5 do not represent individual cases of rehabilitation works.
They are rather indicators of dependency of the rehabilitation cost on the sliding
distance; such dependency could be quantitated by the analysis of many field records
of rehabilitation works. The long term change of the monetary values by inflation and
others are not considered in the present analysis. Its effect as well as the introduction
of interest cost would be the subject of future studies.
October 10, 2000 9:4 WSPC/101-CEJ 00018

374 Y. Goda & H. Takagi


by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 6. Example of expected total cost curves versus safety coefficient against sliding.

When a cross section of breakwater designed against a preliminary wave height is


found to have slid several times beyond Smax in one simulation of a service lifetime,
the rehabilitation cost is the multiple of initial cost plus the rehabilitation cost
corresponding to the remainder of sliding distance. The rehabilitation costs through
5000 repetitions of a given lifetime are averaged together to yield the expected value,
which is then added to the initial construction cost for estimation of the expected
total cost. Figure 6 is an example of the calculation of expected total cost. The
ordinate is the total cost normalized with the initial cost of breakwater section
which is designed with the safety factor of S.F. = 1.0 for the nominal design wave
height. In this calculation, the breakwater section has been designed by varying the
safety factor instead of varying preliminary design wave heights in Fig. 1. Use of
various values of the design safety factor enables easy judgment on the condition of
safety of the optimal breakwater design.
Figure 6 is the case of a breakwater at the water depth h = 16 m for the nominal
design wave height (H1/3 )0D = 6.0 m at the return period of 50 years. The service
lifetime is set at 50 years, same as the return period. The bottom slope is 1/50 and
the tidal range is 1.0 m. The optimal design, or the cross section with the minimum
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 375

value of expected total cost, depends on the model for rehabilitation cost. With the
model Case 3, the optimal design is obtained at S.F. = 0.95, but with the models
Cases 1 and 2 the condition of S.F. ' 0.75 yields the optimal design with a narrower
caisson width. Because the model Case 3 assumes more rapid rise of rehabilitation
cost at a small sliding distance than Cases 1 and 2, the expected rehabilitation cost
becomes larger and the optimal design point shifts to a higher S.F. value.
The optimal value of safety factor for designing a breakwater depends on the
water depth. Under the same wave height, tidal level, and bottom slope, the optimal
design with the rehabilitation cost model Case 3 has been found to appear at S.F. =
0.75 in the water depth h = 12 m, while the optimal safety factor becomes S.F. ' 1.1
in h = 20 m. Figure 6 also shows the ESD with the scale on the upper portion.
The optimal breakwater section with the rehabilitation cost model Case 3 has the
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

ESD of about 0.65 m. The optimal breakwater sections at depths 12 m and 20 m


also indicate almost the same value of expected sliding distance. In Fig. 4, the
breakwaters designed with the constant safety factor of S.F. = 1.2 are shown to
have the ESD increasing with the water depth. That is to say, a severer damage on
breakwaters is expected in deeper water than in shallow water if a breakwater is
designed with a constant value of safety factor. With the optimal design method,
the extent of expected damage and its probability remain independent of the water
depth. A similar conclusion has been obtained by Shimosako and Takahashi (1998
and 1999) with their DBRD method by means of ESD.

7. Frequency Distribution of Total Sliding Distance Per Storm


The optimal designs of breakwaters discussed above have the expected sliding dis-
tance of about 0.65 m. Engineers would ask if this amount of sliding is acceptable
in common sense. The expected value in statistics is the ensemble average of a
population, or the arithmetic average over a large number of samples. A range of
distribution also needs to be known for the examination of the characteristics of vari-
ates. For this purpose, the frequency distribution of the total sliding distance over
the lifetime of breakwater has been constructed from the data of 5000 simulations.
Figure 7 is an example of such examinations.
The breakwater section has been designed for the nominal wave height
(H1/3 )0D = 6.0 m with the safety factor S.F. = 1.0 in water of the depth h = 15 m.
The sea bottom slope is 1/50 and the tidal range is 1.0 m. The frequency distribu-
tion is much skewed positively with a long tail on the right. Thus, the mean sliding
distance of upper 5% becomes 1.89 m and the mean of upper 10% is 1.26 m, while
the overall mean is only 0.19 m: this mean value is about one half of the case shown
in Fig. 6 at h = 16 m (SEXP = 0.41 m) because of a smaller water depth. The
characteristics of large skewness of the sliding distance distribution have been con-
firmed in other design conditions as well as in the example given by Shimosako and
Takahashi (1998).
October 10, 2000 9:4 WSPC/101-CEJ 00018

376 Y. Goda & H. Takagi


by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 7. Example of frequency distribution of accumulated sliding distance during one cycle of life-
time [(H1/3 )0D = 6.0 m, h = 15 m, bottom slope of 1/50, and tidal range of 1.0 m].

Fig. 8. Relationship between the mean of upper 10% sliding distance and the expected sliding
distance.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 377

The large skewness of the sliding distance distribution is exemplified with the
relationship between the mean of upper 10% and the overall mean as in Fig. 8. The
data have been obtained for the nominal design wave heights ranging from 5.0 m to
8.0 m, the bottom slopes of 1/20 and 1/50, and the tidal ranges of 0.5 m and 1.0 m.
The breakwater sections have been designed with the safety factor S.F. = 1.0. A
clear straight line of regressions can be drawn with the mean of upper 10% being
7.5 times the expected sliding distance.
The ESD value by definition can be taken as the average over many caissons
along a long stretch of breakwater extending in water of the same water depth. If
the ESD is computed as SEXP = 0.3 m at a particular design condition, 3 or 4
among 100 caissons would be displaced more than 2.0 m when the design storm hits
the breakwater, according to the regression line shown in Fig. 8. The event would
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

be reported to Port Authority as a major damage of breakwater. In consideration of


such situations, the tolerable limit of ESD is better set at SE = 0.10 m rather than
0.30 m proposed by Shimosako and Takahashi (1998 and 1999). Since they do not
show the relationship between the ESD and the mean of upper 10%, it is unable to
know what is their mean value of upper 10% corresponding to SE = 0.3 m. There
is a possibility that the ratio of the former to the latter is smaller than that shown
in Fig. 8, because of the difference between the extremal distribution functions of
design wave heights employed by them and us. Examination of the effect of the
extremal distribution function on the optimal breakwater design would be a major
task for future studies.

8. Rational Breakwater Design with Economic Optimization and


Safety Consideration
8.1. Model calculation of breakwater section
To illustrate the newly proposed method of reliability-based design of caisson break-
water, a model calculation is made for the following conditions:

(1) Elevation of seabed: −19 m


(2) Rubble mound thickness: t = 2.5 m
(3) Foot-protection block height: t0 = 1.5 m
(4) Berm width of rubble mound: b = 10.0 m (front) and 7.0 m (rear)
(5) Extremal wave distribution: Weibull with
q A = 1.7 m, B = 2.59 m, and k = 1.0
(6) Significant wave period: T1/3 = 4.6 (H1/3 )0D
(7) Tide level: H.W.L. = + 0.5 m, L.W.L. = ±0.0 m
(8) Sea bottom slope: 1/100
(9) Incident wave angle: 0◦ .

A cross section of breakwater is designed for the wave height (H1/3 )0D = 9.24 m
and T1/3 = 14.0 s corresponding to the return period R = 50 years. The significant
October 10, 2000 9:4 WSPC/101-CEJ 00018

378 Y. Goda & H. Takagi


by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 9. Curves of expected total cost and sliding distance for test case of breakwater design
[(H1/3 )0D = 9.24 m, h = 19.5 m, bottom slope of 1/100, and tidal range of 0.5 m].

wave height in front of the breakwater is calculated as 8.50 m at the high tide level.
Under the conventional method, the required caisson width is B = 24.05 m with
the safety factor S.F. = 1.2. The probability of sliding failure and the ESD over
the lifetime L = 50 years are computed as Pf = 0.0076 and SEXP = 0.0054 m,
respectively. It would be said that this is a rather conservative design.
Under the new method of reliability-based design, the breakwater section is com-
puted by varying the return period of design wave from 20 to 100 years. Figure 9
shows the results of computations; the solid line is the expected sliding distance, the
dashed lines are the expected total costs with the rehabilitation cost models Cases 1
and 3, and the two horizontal dash-dot lines indicate the tolerable limit of sliding
distance at SE = 0.30 m and 0.10 m. The expected total cost is normalized with
the construction cost of the breakwater under the conventional design.
If the rehabilitation cost model of Case 1 is employed, the economic optimiza-
tion dictates the return period be set at around 25 years with a cost reduction of
24% compared with the conventional design. However, the expected sliding distance
becomes SEXP = 1.2 m and the probability of sliding failure more than 0.2. There
would be too many cases of breakwater damage if the breakwaters are designed
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 379

Table 3. Comparison of breakwater sections under the conventional and new reliability-based design
methods.

Item Conventional SE = 0.30 m SE = 0.10 m

Safety factor S.F. 1.2 1.0 1.0


Return period R (year) 50 35 80
Offshore wave height (H1/3 )0D (m) 9.24 8.63 10.04
Local wave height H1/3 (m) 8.50 7.95 9.23
Caisson width B (m) 24.05 19.68 20.45
Height of upright section (m) 21.60 21.27 22.04
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.

by the economic optimization principle alone. Thus, we propose to satisfy both the
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

economic optimization and the condition to keep the ESD below a tolerable limit. If
the design at the minimum point of the expected total cost curve indicates the ESD
exceeding a tolerable limit, the design with the distance just below the tolerable
limit should be adopted as the optimal design section; i.e. it yields the minimum
cost within the allowable sliding distance. For the rehabilitation cost model Case 3,
the minimum point of the expected total cost curve coincides with the condition
of SEXP = SE = 0.30 m at the return period of 35 years. However, the optimal
design is obtained at the return period of 80 years when the tolerable limit is set at
SE = 0.10 m.
The characteristics of design comparison are listed in Table 3. The cross sections
by the conventional and new methods are sketched in Fig. 10. Under the new design
method, the breakwater with SE = 0.10 m has a narrower but higher cross section
than the conventional one. A narrower width is owing to the acceptance of an in-
crease in the expected sliding distance. The increase of caisson height is owing to
the use of a higher wave height, because the crest height is set as hc = 0.6H1/3 .

8.2. Rational selection of return period


The example above is the case with the service lifetime of L = 50 years. When a
breakwater is planned for temporary protection of some structures in a short time
span, a question arises as how to select the return period for design purpose. The
conventional procedure is to consider the encounter probability by Borgman (1963).
If the encounter probability is taken at 0.2, for example, the return period needs
to be taken as about 4.5 times the lifetime. The newly proposed method can yield
the optimal return period Ropt as demonstrated above. Thus a series of simulations
have been made to examine the optimal return period for various design conditions.
Simulations have coveredq a wide range of design factors, but the wave periods have
been set as T1/3 = 4.08 (H1/3 )0D , which is shorter than the model breakwater
design in Sec. 8.1. The rehabilitation cost has been estimated with the models of
October 10, 2000 9:4 WSPC/101-CEJ 00018

380 Y. Goda & H. Takagi


by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 10. Comparison of cross sections designed by the conventional and new methods.

Cases 1 and 3 shown in Fig. 5. However, within the range of present investigation,
little difference was observed between the results by the two cost models, because
the optimal designs were determined not by the minimum point of expected total
cost curve but by the limit of the ESD.
The optimal return period is defined here as the return period which yields
the optimal breakwater section satisfying both the condition of the minimum value
of expected total cost and that of the ESD not exceeding the tolerable limit of
SE = 0.10 m. The results of simulations are shown in Fig. 11. Three levels of the
service lifetime, L = 10, 50, and 100 years, are tested. The ordinate of Fig. 11 is the
optimal return period and the abscissa is the water depth relative to the nominal
wave height, i.e. h/(H1/3 )0D . Three exponential curves have been fitted by the least
squares method to the data and are shown with the dash-dot, solid, and dash lines
for the three lifetimes of L = 10, 50, 100 years, respectively. The optimal return
period increase rapidly as the relative water depth increases, and the influence of
service lifetime is rather small. The optimal return period for L = 100 years is about
twice that for L = 10 years.
The results of Fig. 11 suggest that a simple application of encounter probability
does not lead to a rational design of breakwaters. When a breakwater is built in
shallow water, there is no need to take a long return period. For the relative water
depth of h/(H1/3 )0D = 2.0, for example, the optimal return period is about 30 and
40 years for L = 50 and 100 years, respectively. When a breakwater is built in deeper
water, however, the optimal return period becomes longer than the service lifetime.
For h/(H1/3 )0D = 3.0, for example, the optimal return period is about 120 and 150
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 381
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Fig. 11. Examples of optimal return period curves for various levels of service lifetime (Weibull
distribution with k = 1.0).

years for L = 50 and 100 years, respectively. Such effects of the relative water depth
are all due to the breaker height limitation by water depth.
Another noteworthy point in Fig. 11 is the necessity to take a relatively long
return period for a short service lifetime. If a temporary breakwater with L = 10
years is to be built in water with the depth h = 3(H1/3 )0D , for example, the return
period of Ropt = 80 years must be employed in design works. This is another effect
of wave breaking process, because waves of the height with a short return period
have only a small probability of breaking and the breakwater designed against such
waves tends to have a slender cross section. However, there remains a probability
October 10, 2000 9:4 WSPC/101-CEJ 00018

382 Y. Goda & H. Takagi

of large waves attacking the breakwater and the slender section experiences large
sliding once such large waves come.
In the range of h > 3(H1/3 )0D , the optimal return period indicates a tendency of
decrease with some scatter for the case of L = 10 years. It is owing to the limitation
of the present study that includes the sliding failure mode only without considering
the foundation failure mode. It is similar with general tendency observed in Figs. 3
and 4.
The diagram of Fig. 11 cannot be applied for general cases of breakwater design.
When the extremal wave height distribution difference from Eq. (2) is introduced,
different results will be obtained. However, the diagram serves for clarifying the
effects of the relative water depth and the length of service lifetime upon the rational
selection of return period for breakwater design. It is recommended to employ the
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

extremal distribution established at the design site and to derive a diagram similar
to Fig. 11 through the Monte Carlo simulations.

9. Discussions
9.1. Comments on the design method with partial safety factors
In comparison with the DBRD method by Shimosako and Takahashi (1998 and 1999)
as well as the present method of optimal reliability design method, the Level 2 reli-
ability design method with the partial safety factor systems of vertical breakwaters
by Burcharth and Sørensen (1998 and 1999) have the following drawbacks:

(1) The systems are very complicated with various categorizations. Engineers in
charge of breakwater design would find it difficult to choose an appropriate set
of the partial safety factors.
(2) Some guideline must be deviced for the admissible probability of failure. With-
out an appropriate guideline, confusion and inconsistency in the safety level of
breakwater construction would occur when viewed in regional basis.
(3) The categorization of water depth into the two classes of deep and shallow water
is too simple, and no definition is provided for the water depth to differentiate
deep water from shallow one. The classification disregards the process of wave
deformation by random breaking, which progresses gradually from deep to shal-
low water.
(4) Because of the above, the great danger of failure in deep water due to nonbreak-
ing waves such as exhibited in Figs. 3, 4, and 11 is not addressed appropriately.
The DBRD method, on the other hand, has the following advantage over
the partial safety factor systems:
(5) The degree of damage is quantitatively evaluated at the design stage. This
provides engineers with a measure to judge the safety of breakwaters under
design.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 383

(6) The degree of safety against sliding can be set at the same level regardless of
the design conditions. This feature cannot be realized in the conventional design
method or by the partial safety factor systems.
(7) The Monte Carlo simulations can be carried out quite easily with the present
capacity of personal computers. The CPU time for computing one case of opti-
mizing a breakwater design such as the case shown in Fig. 9 is in the range of
2 to 4 minutes for Pentium II 266 MHz prossessors. Thus, engineers can utilize
the method as a tool for daily design works.
(8) The DBRD method can bring a cost reduction in breakwater construction in
shallow to relatively shallow water, as demonstrated in Fig. 10 and by Shimosako
and Takahashi (1998 and 1999).
(9) Though the deformation analysis for the modes of failure other than sliding has
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

not been established, the sliding failure constitutes the majority of the damage
cases of caisson breakwaters. Thus, the introduction of the deformation-based
method at the present stage greatly contributes to rationalization of breakwater
designs.

9.2. Comparison of the authors’ method with that by Shimosako


and Takahashi
As mentioned in Sec. 3, the present study is an extension of the DBRD method by
Shimosako and Takahashi (1998 and 1999). A main difference between them is the
description of design wave height. Shimosako and Takahashi do not vary the design
height explicitly but make trial designs by varying the value of safety factor. They
take the design section that has the ESD just below SE = 0.3 m as the optimal
one. If the safety factor for the optimal design is greater than 1.2, i.e. the value
under the conventional method, it implies the actual design wave height being larger
than the nominal design height and the corresponding return period being longer
than 50 years. Thus, it can be said that they are varying the design wave height
implicitly.
The present method varies the design wave height explicitly while keeping the
safety factor at S.F. = 1.0. Even if the same tolerance limit for ESD is adopted,
a small difference appears however in the crest heights of breakwaters determined
by the two methods, because of Japanese practice of setting the crest height at 0.6
times the design significant wave height.
Although the analysis of expected total cost did not exercise the governing power
for determination of optimal designs within the range of present study, it can become
the critical factor by a change in the rehabilitation cost model or by a use of another
extremal distribution function for storm wave heights. Inclusion of the cost analysis
for economic optimization is the important factor for further development of DBRD
methods for caisson breakwaters.
October 10, 2000 9:4 WSPC/101-CEJ 00018

384 Y. Goda & H. Takagi

9.3. Problems to be solved by future studies


Apparently the DBRD method is the direction to be followed by engineers from
now. However, the following problems need to be solved before the method becomes
the daily design tool of engineers in charge of breakwater designs:

9.3.1. Examination of partial-duration series data


The present study examined the annual maximum storm waves that continue for
2 hours only, following the approach taken by Shimosako and Takahashi. Actually
many events of storm waves occur every year. Extreme data collected from such
events are called the partial-duration series data against the annual maximum series
data of the former. It has been known that the estimate of extreme events for a long
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

return period is not affected by the choice between the two series. However, it is
necessary to investigate whether the use of annual maximum series data can correctly
predict the probability of sliding failures and the expected sliding distance. The
effect of storm duration on the possibility of sliding failure should also be studied;
some assumption for time-variation of storm wave heights will be necessary for such
investigations. A limited number of case study will suffice for clarification of the
effect of partial-duration series data.

9.3.2. Extremal distribution functions


As mentioned in Sec. 4.1, the present study employed the Weibull distribution with
the shape parameter k = 1.0 for the extremal function for the annual maximum wave
heights. In fact, the extreme wave heights have mostly been studied by researchers
in the world with the peaks-over-threshold method, i.e. for the partial-duration se-
ries data. Shimosako and Takahashi (1998 and 1999) say that they used the Weibull
distribution but did not report the parameter values. It is expected that the choice
of extremal distribution function will significantly affect the prediction of ESD in
deep to relatively deep water where the wave breaking process plays a minor role.
The frequency distribution of total sliding distance such as shown in Figs. 7 and 8,
the optimal return period in terms of the service lifetime such as shown in Fig. 11
and others will behave differently from those of the present study, when other ex-
tremal distributions are employed. It is recommended to study the effect of extremal
distribution functions through the Monte Carlo simulations.

9.3.3. Refinement of computational algorithm with examination of design data


uncertainty
The biases and coefficients of variations of the design factors employed in the present
study are tentative ones. Further studies are needed to establish the variability of
design factors before the methodology is applied for actual design works. Sensitivity
analysis will help to determine the degree of accuracy required to be incorporated in
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 385

the Monte Carlo simulations. As for the computational algorithm, individual wave
heights in front of breakwater are better estimated with the original random wave
breaking model by Goda (1975), which can predict the probability density function
of wave heights at any location from the offshore to the shoreline.

9.3.4. Survey of field reports of sliding failures


The technology in civil engineering depends largely on the experience in the past.
The DBRD method should also refer to the performance of existing structures.
Fortunately in Japan, there are 9000 caissons for vertical breakwaters and 7000
caissons for breakwaters protected by moundes of energy-dissipating concrete units,
according to Kawai et al. (1997). They have examined the sliding damage in the
period from 1989 to 1993. It is recommended to extend the survey over a much
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

longer period so that the probability of sliding failures in the prototype breakwaters
could be established more definitely. The survey could also clarify the frequency
distribution of sliding distance as well as the threshold for the mean sliding distance
beyond which the caisson displacement is judged as the breakwater damage. Such
results would provide a quantitative data for the establishment of the tolerable limit
of ESD.

9.3.5. Design diagram of optimal return period


The results shown in Fig. 11 are tentative ones, because the extremal distribution
function of wave heights may not represent the prototype condition. However, a
diagram of optimal return periods for various service lifetimes would be of much
help for engineers in making judgment for the determination of design wave heights.
As mentioned in Sec. 8.2, it is encouraged to prepare such diagrams at respective
sites with the extremal distribution functions established there.

9.3.6. Inclusion of other failure modes in reliability design method


Among various failure modes other than sliding, the circular slip through the founda-
tion and subsoil seems to deserve the most caution. When it does occur, the upright
section is displaced over a large distance and the whole section must be replaced.
It is an event of sudden and total collapse. Definition of progressive deformation is
difficult in contrary to the case of sliding failure. Nevertheless, some measure for
judging the degree of safety against this type of failures must be established within
the scope of reliability design method. Knowledge and experience of geotechnical
engineers are sought for.

10. Concluding Remarks


Construction of a breakwater is an expensive undertaking. It must be safe enough
against the attacks of future severe storms, but the design should not be too
October 10, 2000 9:4 WSPC/101-CEJ 00018

386 Y. Goda & H. Takagi

conservative to economize the project cost. The Level 3 reliability design method
provides engineers with a tool to meet the above contradictory objectives.
The present paper has extended the methodology of ESD analysis by Shimosako
and Takahashi, by adding the concept of economic optimization and by showing
a technique to determine the optimal design wave height and return period. The
present paper can be said to have strengthened the basis for application of the re-
liability design method for caisson breakwaters through clarification of the ideas
involved and by presentation of more simulation data in the present paper. Of
course, there are many problems to be solved for full utilization of the reliability
design method. We wish many engineers will join in the task of solving the prob-
lems, establishing the full model of DBRD method by means of ESD with economic
optimization, and making available it for daily design works.
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

Acknowledgment
The authors wish to express their thanks to the reviewers for their critical comments
and advices given to the original manuscript, which much assisted the authors in
improving the text of the present paper.

References
Borgman, L. E. (1963). Risk criteria, J. Waterway Harbor Div., Proc. ASCE 89, WW3, pp. 1–35.
Burcharth, H. F. (1991). Introduction of partial coefficient in the design of rubble mound
breakwaters, Proc. Coastal Structures Breakwaters, Institute of Civil Engineers., London,
pp. 543–565.
Burcharth, H. F. (1992). Reliability-based design of coastal structures: Structural integrity, Proc.
Short Course, 23rd Int. Conf. Coastal Eng., Venice, ASCE, pp. 511–545
Burcharth, H. F. and J. D. Sørensen (1998). Design of vertical wall caisson breakwater using partial
safety factors, Proc. 26th Int. Conf. Coastal Eng., Copenhagen, ASCE, pp. 2138–2151.
Burcharth, H. F. and J. D. Sørensen (1999). The PIANC safety factor system for breakwaters (by
ed. I. J. Losada), Spain, Balkemare, Coastal Structures ’99, pp. 1125–1144.
Cornell, C. A. (1970). A first order reliability theory of structural designs, Structural Reliability and
Codified Design, SM Study No. 3, University of Waterloo, Ontario, Canada.
Goda, Y. (1974). New wave pressure formulae for composite breakwaters, Prof. 14th Int. Conf.
Coastal Eng., Copenhagen, ASCE, pp. 1702–1720.
Goda, Y. (1975). Irregular wave deformation in the surf zone, Coastal Eng. Japan, JSCE 18: 13–26.
Goda, Y. (1988). Statistical variability of sea state parameters as a function of wave spectrum,
Coastal Eng. Japan, JSCE 31, 1, pp. 39–52.
Goda, Y. (2000). Random Seas and Design of Maritime Structures, 2nd edn., World Scientific,
Singapore, p. 443.
Kawai, H., H. Fujisaki and Y. Suzuki (1996). Occurrence probability of tidal level and its effect of
breakwaters, Proc. Civil Eng. Ocean, JSCE 12: 261–266 (in Japanese).
Kawai, H., T. Takayama, Y. Suzuki and T. Hiraishi (1997). Failure probability of breakwater caisson
for tidal level variation, Report of the Port and Harbour Research Institute 36, 4, pp. 3–41
(in Japanese).
Oumeraci, H. (1992). Review and analysis of vertical breakwater failures — Lessons learned, Coastal
Eng. 22, 3, pp. 3–30.
October 10, 2000 9:4 WSPC/101-CEJ 00018

A Reliability Design Method of Caisson Breakwaters with Optimal Wave Heights 387

Shimosako, K., S. Takahashi and K. Tanimoto (1994). Estimating the sliding distance of composite
breakwaters due to wave forces inclusive of impulsive forces, Proc. 24th Int. Conf. Coastal
Eng., Kobe, ASCE, pp. 1580–1594.
Shimosako, K. and S. Takahashi (1998). Reliability design method of composite breakwater using
expected sliding distance, Report of the Port and Harbour Research Institute 37, 3, pp. 3–30
(in Japanese).
Shimosako, K. and S. Takahashi (1999). Application of deformation-based reliability design for
coastal structures — Expected sliding distance method of composite breakwaters —, Coastal
Structures ’99, ed. I. J. Losada, Spain, Balkemare, pp. 363–371.
Suzuki, S. (1987). Application of the reliability design method to the safety of caisson breakwaters
against sliding (2nd study), Technical Note of Port and Harbour Research Institute 582: 37
(in Japanese).
Takahashi, S., K. Tanimoto and K. Shimosako (1994). A proposal of impulsive pressure coefficient
for design of composite breakwaters, Proc. Int. Conf. Hydro-Tech. Eng. Port Harbor Constr.
(Hydro-Port ’94), Yokosuka, Port and Harbour Research Institute, pp. 489–504.
by FLINDERS UNIVERSITY LIBRARY on 02/02/15. For personal use only.

Takayama, T. and N. Ikeda (1993). Estimation of sliding failure probability of present breakwaters
Coast. Eng. J. 2000.42:357-387. Downloaded from www.worldscientific.com

for probabilistic design, Report of the Port and Harbour Research Institute 31, 5, pp. 3–32.
Toyama, S. (1985). Application of the reliability design method to the safety of caisson breakwaters
against sliding, Technical Note of Port and Harbour Research Institute 540: 49 (in Japanese).
van der Meer, J. W. (1988). Deterministic and probabilistic design of breakwater armour layers,
J. Waterway Port Coastal Ocean Eng., ASCE, 114, 1, pp. 66–80.
Voortman, H. G., H. K. T. Kuijper and J. K. Vrijling (1998). Economic optimal design of vertical
breakwaters, Proc. 26th Int. Conf. Coastal Eng., Copenhagen, ASCE, pp. 2124–2137.

You might also like