You are on page 1of 26

Sesam Users Conference 03.Dec.

2012
Optimization of a Floating
Platform Mooring System Based
on a Genetic Algorithm
Aidin Rezvani Sarabi
Nelson Szilard Galgoul
NSG Engenharia, Projetos e
Representacao Comercial Ltda.
1
Sesam Users Conference 03.Dec.2012
2
Objective
Optimization of the platform heading

Optimization of the mooring pattern

Searching for the tension or length of the
mooring lines

Choosing the optimum line material and size
Sesam Users Conference 03.Dec.2012
3
Choosing an Optimization
Method
Many optimization problems in practical
engineering are quite hard to be solved by
conventional optimization techniques.

So there has been an increasing interest in
solving such hard optimization problems by
imitating the behavior of living beings.
Sesam Users Conference 03.Dec.2012
4
Choosing Optimization Method
Simulating the natural evolutionary process of
living beings results in stochastic optimization
techniques called evolutionary algorithms.

The most widely developed type of
evolutionary algorithms are known today as
Genetic Algorithms (GAs).
Sesam Users Conference 03.Dec.2012
5
Genetic Algorithm Fundamentals
GAs work with a coding of the solution set, not
the solutions themselves
GAs search for a population of solutions, not a
single solution
Genetic Algorithms use payoff information
(Fitness Functions), not derivatives or other
auxiliary knowledge
GAs use probabilistic transition rules, not
deterministic rules

Sesam Users Conference 03.Dec.2012
6
Analysis of Mooring System
using Mimosa
First, Mimosa determines an equilibrium
position by applying a numerical procedure
that solves the equation below:


The solution to this equation is the
equilibrium position that defines the
platform coordinates and heading under
static loads
( ) ( ) ( ) ( ) 0
6 2 1 6
,
2
,
1
= + + + x
wa
f x
wi
F x
cu
F x x x
mo
F
Sesam Users Conference 03.Dec.2012
7
Analysis of the Mooring System
using Mimosa
The actual platform motions are computed
by performing a dynamic analysis, where the
corresponding responses are categorized as
high frequency (HF) and low frequency (LF)
motions

The HF responses are calculated using a
linear spectral analysis.

Sesam Users Conference 03.Dec.2012
8
Analysis of Mooring System
using Mimosa
The LF responses are horizontal motions
(Surge, Sway and Yaw) which result from the
solution of equation below:

LF LF
LF LF
F Kx x C x M = + +

Sesam Users Conference 03.Dec.2012
9
Analysis of the Mooring System
using Mimosa
In order to calculate the extreme values for
the combinations of HF and LF motions,
Mimosa uses a heuristic equation which is
based on model tests and simulation studies
as given in the equation below for one
variable

+
+
=
HF
ext
LF
Sign
LF
ext
HF
Sign
tot
ext
x x
x x
x max
Sesam Users Conference 03.Dec.2012
10
Objective Function Formulation
Each floating unit has six degrees of freedom
(DOF) which include surge, sway, yaw, roll,
pitch and heave. The mooring system is only
capable of controlling the surge, sway and yaw
responses i.e. horizontal responses.

To reduce roll, pitch and heave, i.e. vertical
responses, the vessel shape and dimensions
may be optimized.
Sesam Users Conference 03.Dec.2012
11
Objective Function Formulation
Here optimization of the mooring design,
means to minimize the surge and sway
responses. Surge and sway are platform
longitudinal and transverse displacements
Sesam Users Conference 03.Dec.2012
12
Objective Function Formulation
Minimizing the horizontal translational
response (Platform Offset) is our optimization
problem objective function.
The objective function of the mooring design
optimization problem, the optimization
parameter boundaries and the problem
constraints could be defined as:
( ) ( ) ( ) ( ) | |
=
+ =
=

m
i
i
y
i
x
i
a
m
i
i
i
a Minimize
1
2 2
1
2
. :
( )

=
=
p k
k
g
n j
j j j
to Subjected
,... 1 , 67 . 1
,... 1 ,
max min
:


Sesam Users Conference 03.Dec.2012
13
Penalty Function Formulation
The sequence presented in the previous slide
has led to a constrained optimization problem
which now must be solved
The penalizing strategy is chosen to handle the
constraints. So a constrained problem is
transformed into an unconstrained problem by
penalizing unfeasible solutions. The penalty
function is described as below:
( )

= = <
= = =
p k m i
k i
g if
i
P
p k m i
k i
g if
i
P
,..., 1 , ... 1 , 67 . 1
,
9 . 0 5 . 0
,..., 1 , ... 1 , 67 . 1
,
1
Sesam Users Conference 03.Dec.2012
14
Fitness Function Formulation
Fitness is a quality value that is a measure of
the reproducing efficiency of individuals in a
population.
A potential solution with a higher fitness value
will have greater probability of being selected
as a parent in the reproduction process.
Therefore, the minimization problem must be
transformed into a maximization problem of a
fitness function, using the following
expressions:

2
2
avg
i
i

=
i
i
i
P F . 1
max
|
|
.
|

\
|
=

Sesam Users Conference 03.Dec.2012


15
Genetic Algorithms
As already mentioned above GAs differ from
conventional optimization methods and
search procedures in several fundamental
ways. A GAs basic execution cycle can be
described by the following steps:

Step 1: Reproduction
Step 2: Recombination
Step 3: Replacement
If some convergence criteria is satisfied,
Stop
Otherwise, go to step 1

Sesam Users Conference 03.Dec.2012
16
Implementation Details
1- Coding Design Variable
Design variables were coded using a fixed-length
binary-digit {0,1} string
2- Decoding
To obtain the real values of the design variables in the
domain region, each chromosome must be decoded
3- Offset Computation
Dynamic analyses are carried out in the frequency
domain using Mimosa
4- Fitness Function Calculation
The fitness value of each chromosome is computed by
considering offset values obtained from Mimosa




Sesam Users Conference 03.Dec.2012
17
Implementation Details
5- Selection
Chromosomes are selected as parents to produce
children and this selection depends on fitness values
6- Crossover Operator
The two-point crossover operator (2X), has been
adopted herein, for example 000000 and 111111
make 001100 and 110011
7- Mutation Operator
This operator changes the bit from 1 to 0 or vice versa
8- Generation Gap
It is a parameter that controls percentage of the
population that will be replaced in each generation


Sesam Users Conference 03.Dec.2012
18
Computational Procedure
1- Start
Initialize parameters: population size, crossover and
mutation probabilities
2- Seeding
Initial population is generated randomly
Initial population is decoded
Fitness value of each individual is computed by using
the Mimosa software applying fitness equation
3- Reproduction
Chromosomes are selected as parents
Application of the crossover operator
Application of the mutation operator





Sesam Users Conference 03.Dec.2012
19
Computational Procedure
4- Updating
The new offspring chromosomes substitute the worst
chromosomes of the current population
5- Evaluation
The new chromosomes are decoded
Fitness of the new chromosomes is computed
6- Stopping Criterion Satisfied
If so, then go to step 7; else, go back to step 3 (Here
maximum number of 8000 iterations is considered
7- Repeat
8- End



Sesam Users Conference 03.Dec.2012
20
Case Study
The procedure described in the previous
sections, has been implemented in a
computer program (based on Matlab) which
has been written to solve a mooring design
optimization problem using Mimosa.

As a case study, a
floating unit anchored
by 10 mooring lines,
was considered.




Sesam Users Conference 03.Dec.2012
21
Case Study
The 10 lines were divided into 4 groups with
side constraints, as given in the Table below:



The floating unit is subjected to a set of
environmental conditions that are combined
according to a collinear approach, i.e. with
currents, winds and waves acting
simultaneously in the same direction.
Sesam Users Conference 03.Dec.2012
22
Case Study
In this case study, eight combinations have
been considered. The JONSWAP spectrum for
the Caspian Sea conditions was used to
calculate wave HF responses, while the API
spectrum was used for determining the time
varying part of the wind forces.



Sesam Users Conference 03.Dec.2012
23
Case Study
The total number of iterations considered
here was 8000 and the minimum value of the
objective function was reached at the 43rd
generation in offspring 4269.
The next table presents the final results of the
mooring design optimization problem
including azimuths of each line, anchor
position, line length, line size and line
material. Also the platform heading is 180 deg.
relative to true North.




Sesam Users Conference 03.Dec.2012



Case Study



Sesam Users Conference 03.Dec.2012
25
Case Study
The optimized mooring pattern is illustrated
below. Line size and material in the ordinary
case is 3.5 chain and 4 chain in the optimized
design.
Sesam Users Conference 03.Dec.2012
26
Case Study
The responses are reduced by the optimized
mooring design to up to 3.5 times less than in an
ordinary design. This matter has a great effect on
platform workability because of the reduction of
down-time.

You might also like