TECHNISCHE UNIVERS1TT Laboratorium voor ScheepshydromechaflCa
Archief
Mekelweg 2, 2628 CD Deft
Tel: O15786873 - Fax 015° 781833
INSTITUT FÜR SCHIFFBAU DER UNIVERSITÄT HAMBURG Schrift Nr. 2404
Fatigue Analysis and Test
of 3-D Specimens of Ship Structures
Nie Wu
FATIGUE ANALYSIS AND TEST OF 3-D SPECIMENS OF SHIP STRUCTtJRES
*
NIE WTJ
1. Introduction
In a joint research project on fatigue strength analysis of
weld ship structures between Institute of Naval architecture in Hamburg University (IFS) and Harbin Shipbuilding Engineering Ins-titute (HSEI) a fatigue analysis and test of 3-d specimens
modeling a typical ship structure have been performed since
October. 1992. It will be continued after May. 1993. This project
is supported financially by the Federal German Ministry of
Re-search and Development (BMFT).
Design which takes fatigue into account is necessary for all
moving structures, especially for welded ship structures. They
are subjected to various cyclic loads during their lifetime.
Structural fatigue failures of large ships, specially of tankers,
or at hatch coamings of 'open deck' ships were reported in the
past decades. The cracks were found mainly at the rounded edges
of cutouts in transverse members connected to longitudinal
stif-feners. A number of investigations were carred out [1]. [2]. [3]. and useful methods for estimation of fatigue life were proposed. Moreover, a section on fatigue design of welded ship structures
is adopted in the newest issue of ¡ Rules for Classification and Construction, I-ship Technology ' [4]. [5] published by
GermarliS-cher Lloyd in 1992. It is the first Rule on ship structural
design considering fatigue analysis in the world.
Some influence parameters, however, can not be considered by
the detail classification, and large scatter of fatigue strength
exists in previous analyses. The fatigue strength analysis may be carried out by special fatigue tests for lately adopted structu
*
Assoc. Professor, Harbin Shipbuilding Engineering Institute
rai details and unconventional hull shapes, etc. For structural components welded probably by Roboter the fatigue analysis and
tests are necessary to follow the development of welding Roboter in manufacturing ship structures.
This joint research project is divided into four steps:
selection of details to be investigate according to testing
facilities available in Harbin or in Hamburg and definition of
concepts to be followed.
manufacture specimens for tests and select numerical
analy-sis method for stress distribution in the experimental model.
experiment with the first two specimens and compare the
re-suits of measurement and calculation.
a series of testing and investigations of fatigue strength in Harbin Shipbuilding Engineering Institute.
2. Modelling longitudinal/transverse web intersection and numerical calculation
2.1 specimen type
The specimen type would be representive, and be available to
the testing facilities in HSEI and IFS. The intersection between
a longitudinal stiffener and transverse web was chosen for the experimental investigation of a ship structural detail. This
in-tersection is a 3-d model, as shown in Fig.l. The cutout
corres-ponds to shipyard standards and was flame-cut in the same way as
normally in practice. An angle-bar was used as the longitudinal stiffener. It is used in shipbuilding engineering in the bottom
and side structures of ships such as tankers and containers. The
notches in the cutout in web inherent for such weld structures. A
further advantage of this model is that the results of present investigation could be compared with other existing test data.
The material of the specimens is shipbuilding steel GL A, see Table 1.
The dimensions of the specimens are shown in Fig.l, also. The
connection between plate and stiffener was welded by means of the manual arc process. The electrode used is of the type E
according to DIN
o
nFig. i Experiment model
3
(1) web plate 210x600x8 mm
(2) longitudinal stiffener HP 140x7 x700 mm
2.2 Numerical calulation
Present computer programs of finite element offer a large
lib-rary of finite element types. They are helpful to find out the
stresses in structure details. For the purpes of stress distribu-tian analysis, numerical calculation was performed using the fi-nite element method with the computer program ADINA [6] in this
project. The model of specimen for stress calculation is shown in Fig.2. In the figure there are two models chosen for finite ele-ment method. In (a) the load applied to the top of web is presen-ted by prescribed displacement, which is specified mainly
accor-ding to test data, corresponds to results of numerical analysis.
In (b) the load on the top of web was adopted by the actual force
acted by actuator in the loading system. Calculation models are
the half of the specimen as it is symmetric about ay axis.
Each type of finite element approaches the structural beha-viours of the details more or less accuracy. The types of ele-ments should be chosen carefully for fatigue analysis [7]. The
finite element model for transverse web is presented by
two-di-mensional solid elements -- plane stress elements with B nodes [6]. This type of elements is usually most effective for
model-ling plated structures which are loaded in-plane. According to
the possible distortional behaviour of the longitudinal stiffener
and hull plate in the test model they are idealized by plate
element to keep the number of degrees of freedom manageable. As
structures are unsyminetric about ox and oz axes their deformations may be three dimensions. The shell element is f
ormu-lated treating the shell ( or plate ) as a three-dimensional con-tinium with two assumptions used in Timoshenko beam theory, and the Mindlin/Reissner plate theory. The structural behaviour of
this element normally reflects the beam and plate characters for longitudinal stiffener and hull plate under consideration. As an
attention is focused ori stress distribution around the edge area of cutout in transverse web, in which the fatigue cracks usually
are found, much finer mesh is employed at the area near the edge of cutout as displayed in Fig. 3. Meshes of other components in
this test model are rather coarser compared with the mesh of
I, ,/í, //////2//,'Tv' /
\ \ \
li oz _-i
---Fig.
3±nite e1eent
cdeiing c
webout area.
Because c
te time cosumpti3fl,
a reasonab-e
-rìe
element mesh for the whole model should be chosen carefully.
Transitional mesh regions were introduced between the coarse mesh
and fine mesh in the middle of longitudinal stiffener and hull
plate to reduce the element number. The finite element analysis
was performed with the computer program ADINA. The CPU time used
is about 25 minutes for each loading level in static stress ana-lysis run in computer VAX-6310. Fig. 4 shows the elastic stress
distribution at whole transverse web. The retults of strains cal-culated by the FE method are in good agreement compared with that measured in the test at the edge of the cutout region as shown in
Fig. 5. It may be noticed that there are no many strain gauges
fixed in this area for measuring strain distribution around the
cutout and no measurements at higher stress level which exceeded
the yield stress of the material. All of these measurements might
be suggested in the further tests of the joint research project
to take the local elastic-plastic strains into account. The
maximum normal stress position obtained by FE calculation and
o
strain measurement are both at trie edge or cutout with 11 to the
minimum section of web, not with 50 600 to the minimum section as shown in other investigations. This may be influenced by the
shape of the cutout. A comparison between different shapes and
size of cutout for the maximum normal stress position is also
suggested in the further investigations, if it is possible.
2.3 Determination of crack initiation life
There are different approaches available for description of
the fatigue life of structures. As the local strain approach has
shown to be useful to estimate the fatigue failure of ship structural details, the so-called Smith parameter
PS
[S]
isemployed to determinate the crack initiation life N:
= .1 max Ea E (1)
where
0nax local elastic-plastic stress
Ea : local elastic-plastic strain amplitude
E : material modulus of elasticity
i1) (D U) cl 1) (J) cl- U) U) cl
r
o rl (D LY STRESS-Z Z TIME 1.000 350.0 300.0 250 . 0 200 . O 150.0 100.0 50 . O0.0
-50.0
- 100 . OÇ -
0/ '-e;-/
FEM Q 55 KN2000
r
o1000
RzO
F-r-o\
500 F 2-[N/mm
200
--
flcme cut edge
F.g.
fig.
G q z 100 10 10 iû N L/0
a log-log bilinear parameter life curve. Fig. 6 is adopted in this project which shows results of strain-controled constant amplitude fatigue tests on small scale speciments havïng
flame-cut edge under the stress ratio R=O and R=-l. Expected crack
initiation lifes N are estimated according to Fig. 6 and Smith parameters. A comparison is listed in Table 2. between the FE
cal-culation and measurements for the first two tests.
3. Experiment and results
For fatigue strength analysis of welded ship structures a
se-ries of experiments is proposed. The size and shape of cutout in
specimens are carefully chosen in order to be available to the
testing facilities in Harbin and Hamburg Shipbuilding Instuitute. Considering comparison of experimental results with the one being
obtained in Harbin the connection components between cylider of
acuator in loading system and the web plate of specimen were de-signed by Institute fur Shipbuilding, Universitat der I-Iarnburg, as shown in Fig.7. An advantage of this design is that to joint the
acuator and specimen and to apply force on the edge of web this
connection system, e.g. the flaps and tongue can be handled easi-ly. In addition, it is suitable for the facilities in the
labora-tories of both sides of the joint research project, expecially
for the test details selected with certain dimension. Moreover,
from the point of economic view it is convenient to assemble and
manufacture with lower cost. The experiment for the first two
speciments were performed in
following steps:
3.1. a specimen test
The aim in this step is to check the connection designed. On-ly few of strain gauges were used in the middle points of web to
control the deformation desired. No strain gauges were arranged
at the high stress distribution area around cutout edge in the
web. According to the experimental records during the test the
first crack initiated at a position lined with about 110 angle to
horizental axis in l.3x10° cycles and failured in 7x1O cylcies.
Compared with the expected initiation life N the result is
h
e,c. C Q cc. fl-::u-_-)V
sonable. The measured results in Table 2 demanstrated that the
connection components work well and the design is successful.
3.2. the second specimen test
In this step more strain gauges were used to record the stress
distribution around cutout area of web. Fig. 8 displays positions
at which strain gauges are fixed. DMS Typ 3/120 Lyll strain gauges were adopted for positions No. 26 to No. 32, and DMS Typ
6/120 Lyll used for positions No. 20 to No. 25. The loading range
varies from zero to 55 KN i.e. the stress ratio equals 0. Measurements were recorded step by step during static process of
loading in order to compare with FEM calculation results and that
being tested in 1-ISEI for other specimens. The service cyclic load
is applied to the experimental model with constant amplitudes in
frequency of 15 Hz. In the process of cyclic loading dynamic
mea-surements of strain values from a strain gauge arranged in the
cutout area were roughly displayed with gragh at the screen of oscilloscope in the test system in order to evaluate the crack
initiation life N0 The crack initiation and propagation measure-ments are listed in Table 3. It should be noted that a crack had
been observed with strain values decreasing another crack would
happen in the area at which higher tensen stress exists. In fact, the inspection after the test shows that the second crack
appea-red in the root of connection between web and hull plate as shown in Fig. 8.
It is possible to calculate the crack propagation according to the experimental equation proposed by Paris
d.a/dN = CKm (2)
where a is the crack lengh, N represents the number of loading
cycles, K denotes the increment of stress intensity factor, C
and m are the constants which depend on the testing environment,
its temptature , loads and their frequencies
¡ as well as the stress ratio. Here C=1.83x103, m=3 are adopted [9]. Fig. 9
dis-plays the experimental results of crack propagation and the re-sults computed using FEM according to equation (2).
/
Fig.
8posiL.ion
of crack initiation
4.,0.3/ to. 4;t-/
/
5ecc,crcç
cra.c-< ,,",, ) 70 s-0 4w 30 20 i0
ï
I. ¡ b 7 ¿. í. Cyc(. (0r 2.o .'(a) experimental crack propagation
(2)
I-5 cía. ¿ Q-2(b) calculation crack propagation
A series of investigations
According to the schedule of the joint research project 12
specimens, which are 3-D ship structural detail model consisting
of the same material and manufactured with the same weld manner
as the first two specimens, have been consigned for shipment to
Harbin, China. A series of fatigue tests of these specimens will
be carred out in HSEI. With different stress levels the intial
fatigue uf es can be obtained for each test. It is possible to
observe the crack propagation in some tests with relative high
stress level in the cutout area. All the experimental results
be-ing recorded will offer a basic data in order to compare with
others involving in this joint research project.
Summary and remarks
From the investigation of the first stage in this joint
re-search project the following conclusions can be drawn:
The higher stress ( stress concentration ) normally exists
in the cutout area of transverse web/longitudinal stiffener
in-tersection. The cracks occur near the minimum section of the
cut-out. Positions of cracks initiated will be influenced by the
notches, dimensions of the cutout and other factors.
The crack initiation life can be determined from the
elastic-plastic strain by using a damage parameter life curve and
the predicated life has a good agreement compared with the
expe-rimental results.
At a certain stress level around the cutout area
reasen-able stress distribution can be obtained by using linear FEM in
order to predicate the fatigue life of structures.
For the limited specimen test and numerical analysis there
are further work left to improve the investigation. To compute
the stress distribution in cutout area non-linearity of material due to high stress would be taken into account.
A random loading can be applied on the experimental model to
yield foundimental data to predicate the fatigue life of detailed ship structures during the voyages in their whole lifetime.
Following the development of welding Roboter using in ship
construction the dimension of cutout in web can be different from that for traditional welding. The more stress distribution
analy-sis is suggested for different dimensions and shapes of the cut-out in the web, if there is an opportunity.
Acknowl egement
It should be remarked that the investigations within this
joint research project were financially supported by the Federal German Ministry of Research and Development ( BMFT
) and by the
Structure Labourotary in Institute for Shipbuilding of Hamburg University who provided the test specimens.
The author wish to express his appreciation to Prof. Dr.-Ing H. Petershagen and his colleagues for their helpful advice and
valuable discussions.
Table 1. properties of shipbuilding steel Grad A GL
mechanical properties
16
chemical composition in % ( Oxygensteel )
C Si Mn p 5 Al
0.1
0.24
1.11
0.13
0.07
0.04
R N/mm2 Rn N/mm2 A 334 458 28Table 2. predication of crack initiation life
Table 3. crack initiation and propagation
17
test 1. F=40 KN test 2. F=55 KN
measured computed measured computed
amax, N/mm2 -- 265 349 365 xlo -- 910 948 948 N/mm2 218 223 261 266 N xlO 3 97 4 5 0 l.35 1.1 Number. of cycles xlOO length of crack mm strain observable (show in screen) 100110 --- increase
110120
---120135 --- increase obviously135136
--- decrease 148 3 158 5 168 7 188 18 193 23 198 29 203 35 208 40 213 47 218 58 218.21 60 218.615 failureReferences
[1] W. Fricke and H. Paetzold
Application of the Cyclic Strain Approach to the Fatigue failure of Ship Structural Details.
Journal of Ship Research, Vol. 31. No. 3, 1987, pp 177-185 [2] W. Fricke
Lineare und nichtlineare Strukturanalyse von schiffbaulichen
Konstru.kionsdetails am Beispiel einer
Doppelbodenunter-suchung. Institut für Schiffbau der Universität Hamburg, Report Nr. 454, 1985
[31 H. Paetzold
Beurteilung der Betriebsfestigkeit von
Längsspantdurchfüh-rungeri auf der Grundlage der örtlichen Dehnung,
Institut für Schiffbau der Universität Hamburg, Report Nr. 455, 1983
[4] Germanischer Lloyd
Rules for Classification and Construction, Part I, Ship Technology, Sec. 20, 1992
[3] W. Fricke
Guidance and Examples to the Revised Fatigue Strength Rules Germanischer Lloyd, 1992
ADINA
A Finite Element Program for Automatic Dynamic Incremental Nonlinear Analysis
Report ARD 87-1, ADINA-Engineering, Watertown Mass., 1987
Erkki Niemi
Recommendations Concerning Stress Determination for Fatigue Analysis of Welded Components
11w. Doc. XIII-1458-92
Smith, K. N., Watson, P., and Topper, T. H.
A Stress-Strain Function for the Fatigue of Metals Journal of Materials, Vol. 5 1970, No. 4, pp 767-778
11W Guidance on Assessment of The Fitness for Purpose of
Welded Structures
IIW/IIS-SST-1157-90