Sie sind auf Seite 1von 17

NAME 6121

Marine Structural Vibration


Comparative Analysis of Ship Vibration FEA Code developed
in FORTRAN and MAESTRO

Submitted To: Brandon M. Taravella, PhD,P.E.

Submitted By: Ahmed Islam


Date of Submission: 11/25/2014

Department of Naval Architecture and Marine


Engineering

Considering any ship as a beam model, this project compared different mode shapes and the
vibratory displacement using different methods developed in a FORTRAN and excel program.
These values were then compared against the well-known Finite Element Analysis program
Maestro.
General Characteristics of the Vessel
General Characteristics of the Vessel
Length(L)
Breadth(B)

173
32.2

m
m

T(Draft)
Cms

12.1
0.99

Vertical Moment of Inertia

238.4

m4

Structural Characteristics
Modulus of Elasticity

200

Gpa

Structural damping tan

0.001

Using the full load condition


1. Continuous beam analytical solution
The analytical solution contains the following assumptions:
a.
b.
c.
d.
e.
f.
g.
h.
i.

Constant cross section (Inertia of Mid-ship, no shear deflection)


Constant Mass Distribution
Rigid body heave and pitch induced mode shapes
Exciting force from the propeller is being used(~6kN)
Discarded the damping terms
The mass includes added mass and physical mass.
The mode shapes are independent and superimposed
Ship (beam) is considered as Euler beam, ignoring rotational inertia and shear
stiffness.
Ratio of rigid frequency to flexural frequency is considered to be roughly ~ 1.

The equation of motion defining the vibration of ship is given as:

4
5
2

+
+ = (, )
4
4
4

where,
4

is the beam bending

4 is the structural damping

2
4

is the mass(physical + added mass) inertia


is the hydrodynamic damping
is the hydrostatic stiffness

The beam bending and the structural damping terms come from Voigt Kelvin model for
viscoelasticity. Important to note that the Voigt Kelvin method represents damping with a
spring and dashpot configured parallel to each other. The beam considered for this analytical
solution is the Euler beam which ignores the rotational inertia and shear stiffness. In the Euler
beam the cross section is perpendicular to the axis of bending. In a Timoshenko beam rotation
is taken into account between the cross section and the bending line. The rotation comes from

a shear deformation, which is not included in an Euler beam. The excitation force (, ) can be
represented in Fourier series as = () = cos().
Using the boundary conditions on the Eulers beam the equation of motion of the beam
becomes,

=0
=0
=0
=0

i.e moment = 0 @x=0 and x=L


i.e shear = 0 @x=0 and x=L

Solving the differential equation,

(1 + )

(2 ) = ( 3 )
3

Considering F(x) concentrated at x=1 and discarding the damping:


Equation of motion results in W- 4 = 0 and the boundary conditions are,
= = 0
= 0, =

@ x=0
@ x=L

Where
2

= ( ) 2 2 ( ) ( )

3
=
, characteristics flexural frequency

= ,

= 2 ,

characteristics rigid body frequency


hydro-dynamic damping
structural damping

When a concentrated force, such as the thrust or vertically applied force on a propeller acts on
the ship, we can deduce the resonating frequency by using the beam deflection equation,
() =

[(cos() cosh() (sinh() + sin())


4
2
1 cosh() cos()
+ (sinh() sin())(cosh() + cos())]

1 = cosh() cos() is where the resonant frequency will occur.


Therefore the kappa values can be used to calculate the natural frequency,
2

+( )

Further modal expansion technique yields,


() = 1 () + 2 cos() + 3 sinh() + 4 cosh()
Applying the end conditions, the mode shapes can be represented as
() = 4 [( ) + cosh( ) +

sin( ) + sinh( )
sin( ) + sinh( )
cos( ) cosh( )

the modal stiffness as


= 4 () dx
the modal exciting force as
= () () dx
and the vibratory displacement is,

2
2
=1 1 (
)
+
(2
(

) ]
[

()cos( )

Resulting in the following,

W(x1)

W(x2)

W(x3)

W(x4)

W(x5)

W(x6)

W(xf)

4.56459E-10

-4.5918E-10

6.97462E-10

-9.85293E-10

1.30954E-10

-4.45419E-11

-2.0414E-10

0.05

3.50452E-10

-2.7923E-10

3.16238E-10

-3.04065E-10

2.03258E-11

-6.70738E-13

1.03051E-10

0.1

2.45196E-10

-1.0457E-10

-3.65003E-11

2.80262E-10

-6.32891E-11

2.7064E-11

3.48165E-10

0.15

1.42417E-10

5.37791E-11 -3.08215E-10

6.13644E-10

-8.51336E-11

2.29096E-11

4.39401E-10

0.2

4.46121E-11

1.82235E-10 -4.48475E-10

5.99263E-10

-3.99521E-11 -5.29989E-12

3.32383E-10

0.25

-4.52742E-11 2.68384E-10 -4.33041E-10

2.73668E-10

3.63334E-11

-2.92325E-11

7.0837E-11

0.3

-1.24154E-10 3.03967E-10 -2.76336E-10

-1.95211E-10

8.84443E-11

-2.55313E-11

-2.28821E-10

0.35

-1.89076E-10 2.86897E-10 -3.02471E-11

-5.79494E-10

7.90876E-11

2.44977E-12

-4.30382E-10

0.4

-2.37468E-10 2.22089E-10

2.29419E-10

-6.94768E-10

1.45075E-11

2.80548E-11

-4.38165E-10

0.45

-2.67342E-10

1.2108E-10

4.24295E-10

-4.84946E-10

-6.01448E-11

2.68553E-11

-2.40203E-10

0.5

-2.77444E-10 5.34133E-13

4.96015E-10

-4.76873E-11

-9.25752E-11

5.54744E-15

7.8848E-11

0.55

-2.67345E-10

-1.2009E-10

4.23155E-10

4.12056E-10

-6.00575E-11 -2.68495E-11

3.60871E-10

0.6

-2.37473E-10

-2.2132E-10

2.27472E-10

6.78784E-10

1.46208E-11

-2.80599E-11

4.34029E-10

0.65

-1.89083E-10

-2.8645E-10

-3.2438E-11

6.28175E-10

7.91474E-11

-2.46083E-12

1.96893E-10

0.7

-1.24163E-10

-3.0389E-10

-2.78144E-10

2.86127E-10

8.84084E-11

2.55248E-11

-3.06141E-10

0.75

-4.5284E-11

-2.6868E-10

-4.33966E-10

-1.82351E-10

3.62262E-11

2.92367E-11

-8.64813E-10

0.8

4.46023E-11

-1.8281E-10

-4.48303E-10

-5.48633E-10

-4.0057E-11

5.31098E-12

-1.16989E-09

0.85

1.42408E-10

-5.4512E-11

-3.071E-10

-6.23771E-10

-8.51664E-11 -2.29021E-11

-9.51044E-10

0.9

2.45189E-10

1.03863E-10 -3.49498E-11

-3.3885E-10

-6.3236E-11

-2.70664E-11

-1.1505E-10

0.95

3.50448E-10

2.78764E-10

3.17457E-10

2.39743E-10

2.04061E-11

6.62776E-13

1.20748E-09

4.56459E-10

4.59178E-10

6.97455E-10

9.76038E-10

1.30954E-10

4.45419E-11

2.76463E-09

Vibratory Displacement for


k = 4.73,7.85,11, 14.14,17.28,20.42 and
Superimposed
3E-09

Displacement W(x)

2.5E-09

2E-09
1.5E-09
1E-09

5E-10
0
-5E-10 0

0.2

0.4

0.6

0.8

-1E-09
-1.5E-09

Mode 1

Mode 2

Mode 3

Mode 4

ModeLocation
5

Mode 6

Superimposed Mode

1.2

2. 20 beam FEA program with bending stiffness only (FORTRAN code)


Validation of the FORTRAN program
Using a simple translational system with 2-DOF

For the given mass-spring system,


we can write that,
1 = 21 + 2
22 = 22 + 1
Assuming that the displacements are:
1 = 1
2 = 2
We can write the following system of equation,
2
[2

1
0

2 ] [ ] = [0]
2 2
2

for which the characteristic equation results in,


(2 2 )(2 22 ) 2 = 0
Solving the roots for the natural frequency, we get
1 = 0.634

2 = 2.366

For verification purpose the mass and stiffness was: = [

10 0
100 50
] = [
]
0 20
50 100

Using the following values in the FORTRAN CODE, the resulting values were:
(see appendix for complete FORTRAN routine)

m1,1
m2,2

10

k1,1
k1,2
k2,1
K2,2

100
-50
-50
100

20

omega1 1.780449
omega2 3.439477
The omega values were same when used the equation for 1 and 2

The 20 beam FORTRAN uses the following assumptions:


a.
b.
c.
d.
e.

Different Inertia and Mass distribution over the length of the Beam
No exciting force
Discarded the damping terms
The mass includes added mass and physical mass.
Ship (beam) is considered as Euler beam, ignoring rotational inertia and shear
stiffness (not a Timoshenko beam).
f. The method also ignored the high fluctuating inertia resulting from the bulkheads
inside the hull.

Inside the Fortran Code:


Considering the beam element for the Fortran program, the ends of the elements are rigidly
connected to the adjoining structure instead of being pinned, the element will therefore
simulate just like a beam with moments, and lateral forces acting on them. The Fortran
program reads the masses of the beam elements and places them in a mass matrix form:
1
=[ .
.

.
2
.

.
. ]

Then the Fortran reads the Inertia values for each beam element. E, modulus of elasticity for
the beam has been recorded earlier as E=204GPa and the length was considered as the length
of the given vessel.
In order to assemble the stiffness matrix, the program has a segment where it computes the
stiffnesses based on the E, I and the L values and then places them in a matrix form.

Figure: Beam Model

Figure: Stiffness Matrix

Having the mass and the stiffness matrix, the program then computes the dynamic matrix or
the eigen vectors which represents the mode shapes. For a beam the vibration model signifies
an equation of the form,
[ + ] = 0
where M and K are mass and stiffness matrices, is the eigenvalue related to the natural
frequency by = 2 . [1]
2 [][] + [][] = [0]
[][] + [][] = [0]
[][] = [][]
therefore the final accomplishment of the program is to compute the mode shape by utilizing
the following:
[]1 [][] = []
3. 20 beam FEA program with bending stiffness and Shear Stiffness (FORTRAN code)
In addition to the bending stiffness, the program was made to compute the mode shapes and
displacement taking shear stiffness into considerations. The shear stiffness is used when the
program was made to compute the shear deflection of the beam. Assumptions remain same,
and the beam is still a Eulers beam since it does not take torsional moments into account and
therefore cannot be a Timoshenko beam.
1

12

For shear stiffness the matrix was multiplied by 1+ where = 2 , and the matrix was

manipulated by :
2,2 =

(4 + )

2,4 =

(2 )

4,2 =

(2 )

4,4 =

(4 + )

4. 11,000 element Maestro FEA model


a. Different Inertia and Mass distribution over the length of the Beam
b. No exciting force
c. The mass includes added mass and physical mass (wet mode).
d. Ship (beam) is considered as Timoshenko beam, considering rotational inertia and
shear stiffness.
e. The method also took into account of the high fluctuating inertia resulting from the
bulkheads inside the hull.
f. Maestro had pin-pin constraints.
Maestro used beam element which consists of only the beam itself. That is, both the area A
(and the parameter = AE/L) and the moment of inertia I (and the parameter = EI/L3) refer
to the beam alone, and I is calculated about the beam's own centroid, the centroidal axes being
denoted as x,y. Where maestro used plate elements, it used the following equation to
compute the displacement due to vibration:

Comparison of the frequencies: all in rad/s

1st Flexural
Mode

Analytical Method
Fortran-1 (using only beam stiffness)
Fortran-2(using beam and shear stiffness)
Maestro

% difference
Analytical Method
Fortran-1
Fortran-2
Maestro

Analytical
Method
0.00
0.38
0.34
0.98

4.342
6.952
6.536
2.190

Fortran-1

Fortran-2

Maestro

0.60
0.00
0.06
2.17

0.51
0.06
0.00
1.98

0.50
0.68
0.66
0.00

-------------------------------------------------------------------------------------------------------------------------------

2nd Flexural
Mode

Analytical Method
Fortran-1 (using only beam stiffness)
Fortran-2(using beam and shear stiffness)
Maestro

% difference
Analytical Method
Fortran-1
Fortran-2
Maestro

Analytical
Method
0.00
0.33
0.41
0.88

Fortran-1
0.48
0.00
0.12
1.79

11.979
17.762
20.242
6.3634

FortranMaestro
2
0.69
0.47
0.14
0.64
0.00
0.69
2.18
0.00

------------------------------------------------------------------------------------------------------------------------------

3rd Flexural Mode

Analytical Method
Fortran-1 (using only beam stiffness)
Fortran-2(using beam and shear stiffness)
Maestro

% difference
Analytical Method
Fortran-1
Fortran-2
Maestro

Analytical
Method
0.00
0.35
0.33
1.55

Fortran-1
0.26
0.00
0.50
0.88

23.525
31.9567
35.024
9.23207

FortranMaestro
2
0.49
0.61
1.02
0.47
0.00
0.74
2.79
0.00

Comparison Mode Shape 1:


Analytical Solution:
Vibratory Displacement k=4.73

Displacement W(x)

2E-09
1E-09
0
-1E-09

0.2

0.4

-2E-09

0.6

0.8

1.2

30

40

50

30

40

50

Location X

Fortran -1

Displacement

Mode Shape 1

1.20E+00
1.00E+00
8.00E-01
6.00E-01
4.00E-01
2.00E-01
0.00E+00
-2.00E-01 0
-4.00E-01
-6.00E-01

10

20

Location

Fortran-2
Mode shape 1

Displacement

2.00E+00
1.00E+00
0.00E+00
-1.00E+00

10

20

-2.00E+00
-3.00E+00

Location

Maestro:

Comparing Mode Shape -2


Analytical Solution
Vibratory Displacement k=7.85

Displacement W(x)

2E-09
1E-09
0
0

0.2

0.4

0.6

0.8

1.2

40

50

-1E-09
-2E-09

Location X

Fortran -1
Mode Shape 2

1.50E+00

Displacement

1.00E+00
5.00E-01
0.00E+00
-5.00E-01 0

10

20

30

-1.00E+00
-1.50E+00

Location

Fortran-2

Mode Shape 2
1.50E+00

Displacement

1.00E+00

5.00E-01
0.00E+00
-5.00E-01 0

10

20

30

40

50

-1.00E+00
-1.50E+00

Location

Maestro

Comparing Mode Shape -3


Analytical Solution:
Vibratory Displacement k=11.00

Displacement W(x)

1E-09
5E-10
-1E-24
-5E-10 0

0.2

0.4

0.6

-1E-09
-1.5E-09
-2E-09

Location X

0.8

1.2

Fortran-1
Mode Shape 3

Displacement

5.00E+00
0.00E+00
0

10

20

30

40

50

30

40

50

-5.00E+00
-1.00E+01

Location

Fortran-2

Mode Shape 3

Displacement

1.50E+00
1.00E+00
5.00E-01
0.00E+00
0
-5.00E-01

Maestro

10

20
Location

Summary and Discussion of Result:


Among all the methods that have been compared, Maestro model seems to be more accurate.
The analytical solution takes into account a number of assumptions which reduces the accuracy
of the examined data and result. Also the finite element method has increasingly improved
result with high number of nodes or elements. As it was said earlier, the Maestro model
considered rotational inertia, and the results in the Maestro considered torsional vibration
which the Fortran nor the analytical solution considered. Also in regards to the mass
distribution, the Maestro model has more accurate mass distribution (~distributed along 42
station) than the Fortran(~distributed along 20 stations) or the analytical solution ( single mass).
In the Fortran program the structural damping was ignored, whereas the structural damping
was included in the analytical solution and the Maestro model. Maestro used different shear
areas to calculate the shear deflections in more degrees of freedom than the other methods.
There are certain discrepancies in the assumption for the Maestro model took place. Since we
usually consider the ship to be a free-free beam, the Maestro was forced to have pin-pin
connection for the functionality of the constraints in its FEM method. There are certain number
of other features that Maestro have taken into account that was different than the analytical
and Fortran code. The added mass calculation in Maestro did not use the same method as
Volker Bertrams hmasse subroutine used to calculate the added mass in Fortran.
In future, the addition of torsional and rotational stiffness would improve the Fortran codes
ability to predict of the mode shapes of simple beams. Better and efficient way to re-distribute
the mass over the length of the beam/ship may increase the outcome of the results from the
Fortran code.

Das könnte Ihnen auch gefallen