See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/270895996

ULTIMATE STRENGTH AND UNCERTAINTY

ANALYSIS OF STIFFENED PANELS
CONFERENCE PAPER OCTOBER 2014

DOWNLOADS

VIEWS

65

47

2 AUTHORS:
Murat Ozdemir

Ahmet Ergin

Istanbul Technical University

Istanbul Technical University

1 PUBLICATION 0 CITATIONS

14 PUBLICATIONS 146 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Murat Ozdemir

Retrieved on: 05 September 2015

TEAM 2014, Oct. 13 - 16, 2014, Istanbul, Turkey

ULTIMATE
STRENGTH
AND
ANALYSIS OF STIFFENED PANELS

UNCERTAINTY

Murat Ozdemir*, Ahmet Ergin

*Research Assistant, Istanbul Technical University, Faculty of Naval Architecture and Ocean
Engineering, Istanbul, Turkey.
E-mail: mozdemir@itu.edu.tr
Abstract
Correct calculation of ultimate strength is one of the most crucial aspects of ship structural design.
Ultimate strength of stiffened panels depends on geometry, material, loads and environmental aspects.
Due to the uncertainty of these parameters, ultimate strength of stiffened panels has inherent variability.
Under this point of view, nonlinear finite element analyses are performed in order to evaluate ultimate
strength of stiffened panels. Results are compared with those in literature. Artificial Neural Network
Model is developed considering nonlinear finite element results. Then, ANN based Monte Carlo
simulations are conducted to investigate the uncertainty of ultimate strength of panels.
Keywords: Ultimate Strength, Uncertainty Analysis, Stiffened Panel, Artificial Neural Network, Monte
Carlo Simulation.

1. INTRODUCTION
Collapse of structure must be identified from the safety point of view. Ultimate limit state of
structures is considered as failure of structure after reaching maximum load carrying capacity
(ultimate strength).
Ultimate strength of structures is relevant with loading condition, geometrical and material
properties. All of the parameters which affect the ultimate strength and collapse behavior of
panels have inherent uncertainties, due to this fact, ultimate strength shows variability. The
variability of structural capacity and reliability of structural system must be quantified.
Reliability techniques have been in development of years [1]. These methods first appeared in
a mathematical form in 1926 by Mayer [2], further developed by Streletzki [3] and Wierzbicki
[4]. With the structural reliability methods, all the uncertainties involved in the description of
the load and the structural capacity can be quantified through probabilistic models [5].
Probabilistic analysis of plate buckling was rare, Ivanov and Rousev [6] presented one of the
first studies of plate buckling problem from a probabilistic viewpoint. Guedes Soares [7]
modeled the uncertainties in plate buckling strength for the different cases. Special attention is
given to the implications of problem formulation and to the methodology of uncertainty
modeling.
A method was proposed by Kimiecik and Soares[8] to determine the cumulative distribution
function of the strength of compressed plates using a response surface approach, used the results

from the nonlinear finite element code to fit a response surface limit state function. Recently,
uncertainty and reliability analysis of stiffened panels have been carried out by [9,10] using
nonlinear finite element results.
2. ULTIMATE STRENGTH ANALYSIS OF STIFFENED PANELS
Nonlinear finite element analyses are conducted to evaluate the ultimate strength of stiffened
panels. Stiffened panels undergo several collapse modes depending on loading, plate and
stiffener properties. Collapse modes of stiffened panels can be classified into six modes [11].
If the stiffeners are rigid enough, local plate buckling occurs between the stiffeners. In case of
lower stiffener bending rigidity, overall buckling may be observed. Beside this, if the stiffeners
are very slender, tripping of stiffeners may be occurred.

Fig. 1. Overall collapse of stiffened panel.

Fig. 2. Tripping of stiffener.

Overall collapse mode is considered in this study. Firstly, linear FEM eigenvalue analyses are
carried out to determine the buckling mode shapes of stiffened panels. Overall buckling mode
shape is imposed as initial imperfection to FEM models.

Fig. 3. Finite element model of stiffened panel.

Finite element model must capture all the mechanisms that could lead to collapse of the
structure. For inelastic analysis, 1-bay model cannot capture the collapse accurately because of
that inter-frame bay deflect in upward or downward half wave, while the next bay would deflect
in opposite sense. The boundary condition at the frame is intermediate between simply
supported and clamped, and cannot be accurately modeled as a loaded edge (see, for instance,
[12]). Because of these reasons, the finite element model is represented as a symmetric 1 1
as seen Fig. 3.

2.1 Boundary conditions

The long edges are considered as simply supported. The displacements in the zdirection and rotations about y and z-axes, as in Fig. 4, are constrained to impose simply
supported boundary conditions.
The transverse frame is not modeled, but displacements in the z-direction along the
transverse frame are constrained.
The short edge, which is the mid-length of the mid-bay of the full three bay model, has
symmetric boundary condition. The symmetric boundary condition is satisfied by
constraining displacements in the x-direction and rotation about y-axis.
The displacements in the y-direction are constrained at the mid-width node in each of
two short edges to prevent rigid body motion.

2.2 Initial imperfections

In the nonlinear collapse analysis, bifurcation buckling is not a critical phenomenon. The
problem has to be considered as a continuous response problem [12]. Therefore, initial
imperfections are introduced for the stiffeners and plating. The imperfect geometry is assumed
as overall buckling mode shape obtained from the eigenvalue analysis. The selected mode shape
has an upward deflection (plate induced) in full bay and a downward deflection in half bay as
seen in Fig. 4. The scaling factor for initial imperfection of the stiffened panel is w0 0.0025a
where, a is the length of one bay.

Fig. 4. Overall buckling mode shape of stiffened panel.

Material properties are same for all the panels as modulus of elasticity 205800 MPa, Poisson
ratio 0.3 and yield stress 352.8 MPa. Material is assumed as elastic-perfectly plastic.

Fig. 5. Stress-strain diagram for the panel material.

3. UNCERTAINTY ANALYSIS OF STIFFENED PANELS

Several parameters affect the ultimate strength of stiffened panels. These parameters are
considered as deterministic in classical ultimate strength analyses. Whereas, the parameters
have inherent uncertainty; since, ultimate strength value of panels shows variability. Practical
aspects affecting ultimate strength behavior of panels can be classified into three groups as
physical aspects, model uncertainties and ageing effects [13].
3.1 Artificial Neural Network
Artificial Neural Networks (ANNs) can be applied to many areas, in this study ANN is applied
to estimate ultimate strength of stiffened panels based on nonlinear FEM results. General
structure of ANN is given in Fig. 6. First layer is input layer where raw input data is prepared
for training process. Second layer is hidden layer where most of the neurons are located and;
third layer is output layer where output of the ANN is given.
Models are constructed as discreetly for 3 stiffened and 5 stiffened panel cases. 270 Nonlinear
FEM analyses are performed to train network for each model.
ANN is simple and relatively accurate method to conduct uncertainty analyses; since, it gives
ultimate strength estimation in a good manner. Also, we can conduct uncertainty simulation by
ANN model in a short CPU time.

Fig. 6. General Structure of ANN[16].

Fig. 7. Single neuron of the model.

3.1.1 Mathematical Background

From the mathematical point of view, developing an empirical formula from experimental or
numerical data is to find an approximate function, which can best represent the relation between
input and output variables [14].
Assume that input and output variables are given Eq. 1 and Eq. 2, respectively.
XT = ( x1 , x2 , x3 ,...xn )

(1)

Y T = ( y1 , y2 , y3 ,... yL )

(2)

Output of the i-th neuron in the hidden layer is expressed as in Eq. 3.

i X f wihik xk bi
k 1

(i=1,,m)

(3)

wihik is the weight of the k-th input variable in the input layer to the i-th neuron in the hidden
layer, and bi is constant.
Mathematical expression of j-th output variable is given as follow;

y j f whoiji X c j
k 1

(j=1,,L)

(4)

Eq. 3 and 4 are expressed in matrix form as given;

Y f W2
I
f W

1 X

(5)

In Eq. 5; I is one by one unit matrix, and the W1 and W2 are weight matrices.
3.2 Monte Carlo Simulation
In reliability analysis of structures Monte Carlo Simulation (MCS) is particularly applicable
when an analytical solution is not attainable and the failure domain cannot be expressed or
approximated by an analytical form.
In reliability analysis of structures limit state function G(x) must be defined principally; the
limit state function contains vector of random variables which define the capacity and load
characteristics. In classical reliability analysis, probability of failure of structure can be
expresses as in Eq. 6.
pf

f x ( x)dx

(6)

G ( x )0

In eq. 6 fx(x) is joint probability distribution function of random variables. Analytical

integration of Eq. 6 is practically compelling issue; since, probability of failure of structure can
be examined by MCS as given in Eq. 7.

pf

1
N

I (x )
j 1

(7)

It is assumed that all variables have normal distribution and standard deviation is %3 of design
value for the Monte Carlo Simulation.
4. RESULTS
A series of nonlinear FEM analyses are performed to estimate ultimate strength of stiffened
panels. Main purpose of this study was to investigate variability of ultimate strength of panels.

Results of FEM analyses are used for training of ANN model. Developed ANN model showed
good agreement with FEM results; since, ANN based Monte Carlo Simulations are conducted
to investigate uncertainty of ultimate strength estimations. Results are given in Table 1 and
Table 2 for 3 stiffened and 5 stiffened panels, respectively.
Table 1. Results for 3 stiffened panels.
PANEL

FEM

ANN

MEAN

STD

MIN

MAX

PANEL

FEM

ANN

MEAN

STD

MIN

MAX

P1
P2
P3
P4
P5
P6
P7
P8
P9
P10
P11
P12
P13
P14
P15
P16
P17
P18
P19
P20
P21
P22
P23
P24
P25

0.39
0.33
0.31
0.66
0.50
0.61
0.93
0.78
0.76
0.34
0.28
0.35
0.48
0.43
0.39
0.80
0.79
0.70
0.25
0.26
0.22
0.46
0.47
0.45
0.27
0.74
0.68

0.38
0.31
0.36
0.64
0.48
0.59
0.91
0.80
0.76
0.33
0.28
0.34
0.45
0.42
0.39
0.77
0.74
0.69
0.24
0.25
0.24
0.48
0.46
0.45
0.25
0.76
0.67

0.39
0.31
0.36
0.64
0.48
0.59
0.90
0.80
0.76
0.33
0.28
0.34
0.46
0.42
0.39
0.77
0.74
0.68
0.24
0.25
0.23
0.48
0.46
0.45
0.25
0.76
0.67

0.02
0.02
0.02
0.04
0.03
0.04
0.03
0.04
0.04
0.02
0.02
0.02
0.03
0.03
0.03
0.04
0.03
0.03
0.01
0.02
0.02
0.02
0.02
0.03
0.02
0.02
0.04

0.28
0.22
0.25
0.48
0.35
0.41
0.76
0.64
0.61
0.26
0.21
0.22
0.33
0.29
0.27
0.61
0.62
0.55
0.18
0.18
0.15
0.38
0.36
0.33
0.18
0.65
0.49

0.50
0.39
0.46
0.79
0.63
0.79
1.01
0.94
0.90
0.44
0.36
0.43
0.63
0.54
0.52
0.92
0.87
0.83
0.31
0.33
0.32
0.59
0.58
0.61
0.31
0.85
0.81

P52
P53
P54
P55
P56
P57
P58
P59
P60
P61
P62
P63
P64
P65
P66
P67
P68
P69
P70
P71
P72
P73
P74
P75
P76
P77

0.72
0.62
0.48
0.40
0.56
0.36
0.26
0.22
0.74
0.67
0.70
0.62
0.48
0.58
0.52
0.28
0.23
0.73
0.67
0.69
0.62
0.50
0.58
0.79
0.30
0.25

0.73
0.64
0.46
0.40
0.58
0.35
0.26
0.21
0.75
0.69
0.69
0.61
0.46
0.55
0.51
0.26
0.22
0.73
0.66
0.68
0.60
0.48
0.54
0.80
0.30
0.24

0.72
0.64
0.46
0.40
0.58
0.35
0.26
0.21
0.75
0.68
0.69
0.61
0.47
0.55
0.51
0.27
0.22
0.73
0.66
0.68
0.60
0.48
0.54
0.80
0.30
0.24

0.03
0.03
0.03
0.03
0.03
0.02
0.02
0.01
0.03
0.04
0.03
0.02
0.03
0.02
0.03
0.01
0.02
0.03
0.04
0.04
0.02
0.03
0.03
0.04
0.02
0.03

0.59
0.53
0.35
0.30
0.46
0.26
0.19
0.16
0.63
0.51
0.55
0.52
0.35
0.45
0.39
0.20
0.15
0.62
0.49
0.55
0.51
0.35
0.44
0.65
0.22
0.14

0.82
0.74
0.61
0.54
0.69
0.45
0.38
0.29
0.88
0.85
0.84
0.73
0.60
0.66
0.66
0.33
0.30
0.89
0.85
0.86
0.71
0.63
0.65
0.95
0.40
0.36

PANEL

FEM

ANN

MEAN

STD

MIN

MAX

P41
P42
P43
P44
P45
P46
P47
P48
P49
P78
P79
P80
P81
P82
P83

0.93
0.89
0.81
0.35
0.35
0.35
0.72
0.64
0.63
0.92
0.92
0.91
0.90
0.86
0.80

0.92
0.88
0.79
0.36
0.36
0.36
0.72
0.66
0.62
0.91
0.91
0.90
0.90
0.85
0.77

0.92
0.88
0.78
0.36
0.35
0.35
0.71
0.66
0.62
0.91
0.91
0.90
0.90
0.85
0.77

0.02
0.02
0.03
0.02
0.02
0.02
0.03
0.03
0.04
0.04
0.02
0.03
0.03
0.04
0.03

0.81
0.78
0.64
0.28
0.23
0.19
0.56
0.51
0.43
0.75
0.81
0.79
0.78
0.73
0.63

1.05
0.99
0.92
0.47
0.43
0.42
0.84
0.77
0.75
1.03
1.03
1.03
1.02
1.00
0.93

P50
P51

Table 2. Results for 5 stiffened panels.

PANEL

FEM

ANN

MEAN

STD

MIN

MAX

P26
P27
P28
P29
P30
P31
P32
P33
P34
P35
P36
P37
P38
P39
P40

0.49
0.39
0.45
0.79
0.61
0.70
0.96
0.87
0.86
0.44
0.35
0.40
0.61
0.54
0.49

0.49
0.40
0.46
0.75
0.61
0.69
0.96
0.86
0.83
0.44
0.35
0.41
0.60
0.53
0.49

0.50
0.40
0.46
0.74
0.61
0.69
0.97
0.86
0.84
0.45
0.35
0.40
0.60
0.53
0.49

0.02
0.02
0.02
0.03
0.03
0.03
0.03
0.02
0.03
0.02
0.02
0.02
0.03
0.03
0.02

0.40
0.30
0.33
0.63
0.48
0.56
0.88
0.76
0.70
0.36
0.25
0.29
0.47
0.40
0.36

0.61
0.50
0.57
0.87
0.72
0.84
1.11
0.98
0.95
0.54
0.43
0.48
0.73
0.65
0.59

Table 2. Results for 5 stiffened panels.(Continue)

PANEL

P84
P85
P86
P87
P88
P89
P90
P91
P92
P93
P94
P95

FEM

0.84
0.62
0.56
0.42
0.91
0.91
0.90
0.89
0.85
0.78
0.82
0.63

ANN

MEAN

STD

MIN

MAX

0.83
0.60
0.54
0.44
0.88
0.89
0.88
0.89
0.82
0.77
0.80
0.61

0.83
0.61
0.54
0.44
0.89
0.90
0.88
0.89
0.82
0.77
0.80
0.61

0.03
0.02
0.02
0.03
0.02
0.02
0.03
0.02
0.03
0.03
0.02
0.03

0.75
0.52
0.45
0.33
0.79
0.80
0.77
0.79
0.72
0.66
0.72
0.51

0.97
0.74
0.64
0.57
1.02
1.01
1.01
1.01
0.96
0.89
0.94
0.74

PANEL

P96
P97
P98
P99
P100
P101
P102
P103
P104
P105
P106
P107

FEM

0.55
0.43
0.90
0.89
0.89
0.89
0.84
0.77
0.78
0.81
0.63
0.55

ANN

MEAN

STD

MIN

MAX

0.52
0.44
0.89
0.87
0.88
0.88
0.81
0.77
0.78
0.79
0.61
0.53

0.52
0.43
0.89
0.87
0.88
0.88
0.81
0.76
0.77
0.79
0.62
0.52

0.03
0.03
0.01
0.03
0.03
0.03
0.03
0.02
0.02
0.02
0.03
0.03

0.41
0.26
0.80
0.74
0.75
0.76
0.72
0.64
0.64
0.71
0.51
0.39

0.62
0.55
1.01
0.96
0.98
0.99
0.95
0.88
0.89
0.91
0.75
0.63

As can be seen from Table 1 and Table 2, ANN results are in good agreement with FEM results.
Monte Carlo Simulation results indicate that uncertainty of ultimate strength is higher than the
individual uncertainty of each design parameter.

Fig. 8. Ultimate Strength Distribution of P6.

Fig. 9. Ultimate Strength Distribution of P28.

5. CONCLUSIONS
Ultimate strength of stiffened panels are estimated by nonlinear FEM and results are used for
ANN simulation as target data. ANN simulations have shown good agreement with FEM
results. Also, ANN based Monte Carlo Simulations are performed to estimate uncertainty in
ultimate strength of panels. Following findings have been obtained from this study:
-

ANN is a powerful tool for the estimation of ultimate strength and uncertainty
simulation.
Uncertainty of design parameters (C.O.V: %3) caused variability of ultimate strength
which level is up to %11 of mean value.
Uncertainty of ultimate strength and probability of failure of the structure must be
quantified from the safety point of view.

REFERENCES
[1] Sobey A.J, Blake J.I.R, Shenoi R.A., Monto Carlo Reliability Analysis of Tophat Stiffened
Composite Plate Structures Under Out of Plane Loading. Reliability Engineering and
System Safety, 110, (2013), 41-49.
[2] Mayer M., Die Sicherheit Der Bauwerke und hre Berechnung Nach Grenzkraften Statt
Nach Zulassigen Spannung, Springer-Verlag,(1926).
[3] Streletzki N.S., Statistical Basis for the Evaluation of the Structural Safety Factor. State
Publishing House for Buildings, (1947), Moscow, Stroizdat.
[4] Wierzbicki W., Safety of Structures as a Probabilistic Problem, Technical Report. Warsaw,
Poland: Przeglad Techniczny.
[5] Gaspar B, Guedes Soares C., Hull Girder Reliability Using a Monte Carlo based Simulation
Method, Probabilistic Engineering Mechanics, 31, (2013), 65-75.
[6] Ivanov L.D, Rousev S.G., Statistical Estimation of Reduction Coefficient of Ships Hull
Plates with Initial Deflections, Naval Architecture, 4, (1979), 158-160.
[7] Guedes Soares C., Uncertainty Modeling in Plate Buckling, Structural Safety, 5, (1988); 1734.
[8] Kmiecik M, Guedes Soares C., Response Surface Approach to The Probability Distribution
of The Strength of Compressed Plates, Marine Structures,15, (2002), 139-156.
[9] Garbatov Y, Tekgoz M, Guedes Soares C., Uncertainty Assessment of the Ultimate Strength
of A Stiffened Panel. Proceedings of the 4th International conference on Marine
Sutructures, (2013), Espoo, Finland.
[10] Gaspar B, Naess A, Leira B. J, Guedes Soares C., System Reliability Analysis of Stiffened
Panel Under Combined Uniaxial compression and Lateral Pressure Loads. Structural
Safety, 39, (2012), 30-43.
[11] Paik J. K, Thayamballi A. K., Ultimate Limit State Design of Steel-Plated Structures,
Wiley, (2003).
[12] Chen Y. Ultimate Strength Analysis of Stiffened Panels Using A Beam-Column Method.
Doctoral Dissertation, Virginia Polytechnic Institute and State University 2003; Virginia,
USA.
[13] Proceedings of The 18th International Ship and Offshore Structures Congress, Committee
III.1 Ultimate Strength, (2012), Rostock, Germany.
[14] Pu Y, Mesbahi E., Application of Artificial Neural Networks to evaluation of Ultimate
Strength of Steel Panels, Engineering Structures, 28, (2006), 1190-1196.