You are on page 1of 204

New Trends in Fatigue and

Fracture

NT2F12






27- 30 May 2012, Braov, Romania





i
Contents NT212

1. I. Popescu, R. M. Negriu, S. G. Badea, C. Besleaga, M.Stefanescu -
WAYS OF DETERIORATING THE BALLS WITH
STRUCTURAL GRADIENT FROM THE VALVES USED IN
THE OIL EXTRACTION INDUSTRY. 1
2. J. Capelle, G. Pluvinage - MODICATION OF FAILURE RISK
BYTHE USE OF HIGH STRENGTH STEELS IN PIPELINES .. 15
3. M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar, L.
Moro, M. Gh. Munteanu, F. Vecchiet - THERMO-MECHANICAL
ANALYSIS OF A COPPER MOULD FOR CONTINUOUS
CASTING OF STEEL. 25
4. E. Ghorbel, G. Casalino, A. Ben Hamida - FAILURE
ASSESSMENT OF TRANSMISSION DIODE LASER WELDED
POLYPROPYLENE. 43
5. Yu.G. Matvienko - CRACK TIP PLASTIC ZONE UNDER MODE
I LOADING AND THE NON-SINGULAR TZZ-STRESS 51
6. N. Pasca, L. Marsavina, S. Muntean, R. Negru- FAILURE
ANALYSIS OF A STORAGE PUMP SHAFT 65
7. D. Sersab, M. Aberkane - RECOVERING OF THE
MECHANICAL PROPERTIES OF A PEHD PLATE
CONTAINING A NOTCH BY THE GRINDING TECHNIQUE. 75
8. D.Danii, S.Sedmak, I.Blai - SCENARIO OF FRACTURE
DEVELOPMENT IN BUCKET WHEEL EXCAVATOR 85
9. L. Milovi, S. Bulatovi, Z. Radakovi, V. Aleksi, S. Sedmak, -
ASSESSMENT OF THE BEHAVIOUR OF FATIGUE LOADED
HSLA WELDED STEEL JOINT BY APPLYING FRACTURE
MECHANICS PARAMETERS. 101
ii
10. T. Sedmak, S.Sedmak, L.Milovi - THE APPLICABILITY OF
STRATEGY FOR RISK BASED MAINTENANCE TO A
PENSTOCK. 111
11. M. Mouwakeh, G. Pluvinage, S. Masri - CONSTRAINT
FACTORS USED IN LIMIT ANALYSIS OF POLYETYLENE
PIPES SUBMITTED TO INTERNAL PRESSURE 129
12. C. Voican - SERVICE ORIENTATION IN DISTRIBUTED
AUTOMATION AND CONTROL SERVICE 141
13. C. Voican - FLEXIBLE SERVICE BINDING IN DISTRIBUTED
AUTOMATION AND CONTROL SYSTEM 149
14. M. Simina, I. Ardelean - NUCLEAR MAGNETIC RESONANCE
PROFILING OF HIGH-STRENGTH CONCRETE SAMPLES 159
15. C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi - RESIDUAL
STRESSES AND FATIGUE BEHAVIOR OF HYBRID BUTT
WELDED JOINTS. 169
16. C. Casavola, V. Giordano, C. Pappalettere, C. I. Pruncu-
INFLUENCE OF GEOMETRIC SHAPE OF SPECIMEN IN
FATIGUE LIFE CHARACTERIZATION ON WELDED JOINT
IN TITANIUM ALLOY. 179


1


WAYS OF DETERIORATING THE BALLS WITH
STRUCTURAL GRADIENT FROM THE VALVES
USED IN THE OIL EXTRACTION INDUSTRY

Ionel Popescu
1
, Radu Mihai Negriu
2
, Sorin George Badea
2
, Cristinel
Besleaga
2
, Mihai Stefanescu
2


1
Industrial Biogas Solutions, Rokura Group, Bucharest, Romania
2
Econet Prod Bucharest, Romania
1
ic.popescu@yahoo.com;
2
econetprod@mail.com;
2
cristibesleaga@yahoo.com

Abstract: Valves (formed from ball and seat) equip pumps with pistons and
ensure the movement of the petroleum from the deposit to the surface. The balls
must withstand the complex erosive-abrasive wear determined by the corrosive
environments, while having a resistance to micro cutting and micro fatigue as
good as possible. In the case of extraction through underground combustion the
effect of high temperatures will also occur. Several experimental batches of
balls with structural gradient were made using different extremely hard alloys.
An analysis, using finite elements, of the tension states which occur during the
process was performed and the life span of the balls was estimated. We present
the analysis of the ways of deterioration caused not only by usage in the process
but also by the manufacturing defects of balls with structural gradient. The
complex analyses that were performed to determine the causes and the
mechanisms of deterioration are presented and technological ways to improve
the life span are proposed.
Key words: oil extracting pump, valves, balls, wear, extremely hard alloys, stresses,
fracture, fatigue, FEM, CAD, life span.


1. INTRODUCTION

Valves (made from balls and seats) equip P-type and TB-type piston
pumps (Fig. 1) and according to the stroke, ascending/ descending,
ensures fluid movement (oil) to the surface of the deposit. In this
hydraulic circuit for oil extraction, the balls have a major role. They are
designed to resist the thermo-mechanical state specific to the extraction
(oil type, depth of extraction, technology used, the composition of
extracted material) and the complex erosive-abrasive wear in corrosive
environments combined with the best possible resistance to micro
I. Popescu, R.M. Negriu, S.G. Badea, C. Besleaga, M. Stefanescu


cutting, micro-fatigue and, in the case of extraction through
underground combustion, the effect of high temperatures also occurs.


Figure 1. P-type and TB-type piston pumps

Based on the operating conditions mentioned, the balls are made in
accordance with API 11AX standards [1] from the materials: martensitic
stainless steel, cobalt alloys and simple composite materials WC-Co and
WC-Ni.
The ball wear phenomenon begins when the balls are lifted off the
seats under the high entrance pressure of the fluid. The fluid passing
through the valve consists of a mixture of oil, acid gases such as H
2
S,
salt water, various acids and sand in suspension derived from the
productive layer [2, 3]. This phenomenon continues at each pump and
the phenomenon of fatigue occurs. The ball can rotate so the position of
the wear is random. The seats position is fixed and a wear phenomenon
that occurs in an area worsens with each pump. In the moment the valve
is closing (it makes contact between ball and seat) the wear area is
positioned in the contact area, allowing abrasive fluid to leak and the
wear is more pronounced due to the increased speed.
3


2. TYPES OF BALLS EXAMINED
Balls were analyzed with different degrees of deterioration that were
produced [4] by the following technological procedures:
a. Balls made by powder metallurgy method from sintered metallic
carbides (WC-Co);
These balls are made of tungsten carbides witch Co matrix.
Hydrostatic pressing and the sintering is followed by mechanical
processing in order to get balls with dimensions and tolerances required
by the project. Hot isostatic pressing [5] adds substantial improvements
in the mechanical characteristics as well as a properly cost. As we will
see from the analysis of the thermo-mechanical demands, an
improvement of mechanical characteristics does not necessarily increase
the life of application, superficial wear being the main reason to stop
using the balls.
b. Reconditioned balls. These balls are produced using mechanical
processes to remove traces of wear and defects of the outer layer. If the
size obtained is in the lower dimensional range, it can be used as such.
Before starting reconditioning, balls have to be inspected through non-
destructive methods to determine the depth of the defects. If these
defects will remain in the ball with reduced dimensions, the
reconditioning situation should be analyzed carefully and generally it
was proposed to give up these balls. If the amount of material removed is
relatively small and an economic calculation shows that reconditioning is
efficient, a layer of material can be added through powder metallurgy
technologies. The layer can be of a material more resistant to wear, thus
increasing the duration of use. The technology is in course of being
implemented. This technology is demanding from the viewpoint of
compatibility between the two materials both in the fabrication phase and
in the usage phase (ex: tensions brought in the contact layer due to
differences between the thermal expansion coefficients etc.).
c. Balls made by specific powder metallurgy technologies from two or
more types of virgin powder. This technology was developed and is in
the experimental stage. Although it is more expensive, it can give
remarkable results because it combines specific features of two materials
(mechanical resistance and wear resistance). But the effects that may
occur both in production and in use due to the different characteristics of
the materials and especially the effects of diffusion between the two
materials have to be analyzed. The important problems that were solved
are the concentricity of the two layers and the constant thickness of the
outer layer. Another solved problem is the outer layer grip.
I. Popescu, R.M. Negriu, S.G. Badea, C. Besleaga, M. Stefanescu


d. Balls made by powder metallurgy technology from two or more kinds
of powder from which at least one of them is recovered. It is a
technology similar to that previously presented and it is in the
experimental stage. The great advantage is the recovery of materials.
Several adverse effects can occur (pores, cracks, large grains, micro-
cracks in grains etc). These defects can be removed through carefully
respected technology. The problems of concentricity of the two layers
and the adherence are of the outer layer are important aspects to be taken
into account.
In figure no. 2 is presented a ball made in multi-layer system from two
types of material. Inside there is a ball made of WC-12% Co and outside
a layer of WC-Co12% Ni. This ball was made through our own
technology.
3. TYPES OF DEFECTS CAUSED BY
MANUFACTURING PROCESS OF THE BALL /
SEAT VALVE AND MODES OF DAMAGE
THAT CAN OCCUR USING
Analysis of possible defects is important in order to achieve a model
that together with the macro and micro fractographic analyses allow
choosing the best manufacturing solutions [2, 3, 5-7]. In reality, two or
more deterioration mechanisms act, the superposition effect being a
reduction, sometimes drastically, of the life span. In case of
superposition, micro cracks and removal of material by erosion, no
matter the cause of their production, can evolve rapidly joining and
reaching the critical size. In these cases, the growth rate of defects is
relatively high, typical of materials with brittle behavior. The tri-axial
stress state and the tension deviator especially (with component
stretching) is the one that leads to the sudden increase of cracks.
Depending on tension levels and the way it is stressed, mechanisms of
deterioration typical to fatigue can also occur.
Damage to the ball and valves seat is statistically constant. In practice
the ball rotates so that a range of deterioration mechanisms act in
different places on the surface of the ball. On the valves seat the
mechanisms of deterioration are localized and therefore they produce
more intense effects.

5



Figure 2. Ball made in multi-layer system


The main identified damage mechanisms are:
a. Wear due to abrasive content of the fluid which is carried by the
pump.
The extracted oil contains a number of solid abrasive materials (ex.
sand). These granules, many of which have sharp edges produce a
phenomenon of abrasion. Small scratches get worse during operation and
produce a preferential flow of the fluid. In this case an accelerated wear
with undesirable effects on the sealing ability of the ball-valve seat
assembly occurs. The phenomenon can occur both on the ball and on the
valves seat. The type of material chosen for manufacturing the two
components has a major influence on the life of the assembly. The
decommissioning of the whole assembly or of a component is made
when it is no longer possible to ensure the sealing. Note that areas with
accentuated wear contain micro-cracks. In the corresponding situation of
I. Popescu, R.M. Negriu, S.G. Badea, C. Besleaga, M. Stefanescu


a state with thermo mechanical tensions, these micro-cracks can join or
increase, stable or unstable.
The most dangerous situation occurs when the "scratch/ lack of
material" on the ball and/ or the seat is in the sealing area of the ball/ seat
assembly valve. In this case the fluid that leaves the little channel has a
much higher speed than if the seal is opened. Abrasive materials in the
fluid have a greater effect in the case of flowing through the little
channel with high speed. Deterioration is also increased by the
superposition of corrosion and abrasion produced by abrasive fluid flow
with stress effect due to the contact in the sealing area. In this case
cracks/ wear open and the aggressive/ corrosive/ abrasive environment
increases the depth of deterioration. The most possible situation is
produced by exposing the tungsten carbide crystal so much that it is
pulled from the remaining binder. In the first phase, the tungsten carbide
crystal has an abrasive effect on the adjacent area.
Figure 3 shows the abrasion areas more or less pronounced.

Figure 3. The valve (ball and seat) of piston pump for oil extraction

b. Damage caused by hard material grains that are retained between the
ball and the valves seat when closing the seal. In this case, two
phenomena can occur:
b.1.by compressing a granule of hard material (sand grains, grains of
WC-Co ball ripped from the ball/ seat assembly, for example) it presses
locally in the ball and seat and can cause a local material deformation,
the appearance of cracks etc. This very small deterioration can evolve
over time due to wear effects described previously;
7


b.2. the intensification of the phenomenon described in section 1 in the
case when complete sealing cannot be achieved. In this case the space
through which fluid can flow is very small and symmetrically
misaligned. Flow velocity increases greatly and as a consequence wear is
more pronounced.
c. Deterioration through chemical corrosion due to the H
2
S present in the
oil. Generally sintered carbides resist relatively well to chemical
corrosion by hydrogen sulfide but the overlapping effects of other types
of deterioration and this phenomenon can produce a decrease in the life
span of the sealing valve assembly. However the phenomenon is highly
dependent on the concentration of H
2
S, temperature and water content of
the fluid. The phenomenon is statistically constant over the balls surface
and the valves seat surface for the area exposed to corrosion.
d. Deterioration caused by the shock that occurs at the contact between
the ball and the valves seat when closing the valve. If the closure is
made with a shock, the state of tension caused by this can accentuate the
degradation phenomena produced by the growth and unification of
micro-cracks. In the case of contact between ball/ seat valve, two
situations can be discussed:
d.1. deterioration to the balls may be much lower because the ball can
rotate so that the line of contact between the two components of the
valve may be different during successive closures;
d.2. damage caused to the valves seat can lead to a permanent
deformation which under certain conditions can lead to the apparition
and development of cracks.
In both cases the critical situation occurs when in the wear area close
to the hertzian contact between the two bodies there are defects of pre-
critical dimensions.
e. Damage caused by defects introduced by the using of improper
procedures of powder metallurgy (these defects are due to applied
technologies and can be removed either by strictly following the
procedures or through an effective control before delivering the valves
components and by eliminating the defective parts:
e.1. the appearance of a separation surface (poor sintering between layers
of manufactured when the ball is made using multilayer technology).
This type of defects can be detected by using CND so that the balls with
manufacturing defects are not introduced into use (Fig. 4); Pores occur
due to the poor preparation of the contact surfaces which allow the
occurrence and capture of gases (generally produced when the
temperature rises).
e.2. uneven thickness of the layers from different materials. This type of
defect can occur both during the pressing operation and during the
sintering operation. The uneven thickness of a layer during the thermo-
mechanical stress can lead to misaligned symmetrical tension states with
local components of significant value;
e.3. the appearance of cracks, pores, large grain areas which are initiators
of micro/ macro cracks that may occur either because of inadequate
I. Popescu, R.M. Negriu, S.G. Badea, C. Besleaga, M. Stefanescu


powders (especially when using recovered powders) or the improper
technologies during the course of fabrication;



Figure 4. Defect of the ball made by the breaking of the outer layer

e.4. pressing a layer of powder over a processed body in the case of
reconditioned balls is difficult, requiring special devices. Even in this
case, providing a hydrostatic pressing is difficult, leading to different
compacting values. During sintering different compaction leads to
different densities of the sintered and therefore to uneven properties of
the balls with negative influence on the life span.
f. Deterioration by separating the bodies in two or more pieces,
deterioration of fragile type produced by already existing cracks which
rapidly grow to critical size when the sudden rupture occurs (Fig. 4 and
5).
g. Damage caused by differential expansion that occurs in the case of
multilayer balls. If layers have different physical and mechanical
characteristics, they lead to dilatations/ different strains at the interface
between layers. If the sintering is not appropriate and at the interface
contact defects or other types of defects occur (sintering or training)
these tensions, due to different dilatations/ distortions, can lead to the
growth of defects in a dangerous way (Fig. 5). The lack of adherence can
also be explained by the low temperature that does not allow the partial
melting of the inner piece (solid) and the diffusion between the layers of
different materials.

9




Figure 5. Defect of the ball made by the breaking of the outer layer and a piece
of initial ball
4. ANALYSIS OF DEFORMATION AND STRESS
STATES IN THE BALL SUPPORT ASSEMBLY
For the finite element analysis of the tensions and strains states
produced by loads that occur during the usage of the subassembly in the
process [8, 9] a 3D model was developed with Autodesk Inventor
(Figure 6). For reasons of symmetry of the model and the loads, only one
quarter of the sub-assembly was modeled. The finite element analysis
was performed using ANSYS.
The analysis was performed for stationary conditions. The shock that
occurs at the contact between the ball and the valves seat and the fatigue
phenomenon were not taken into consideration.
The analysis of a sub-assembly composed from a ball made from one
material. The model is presented in the figure 6.

I. Popescu, R.M. Negriu, S.G. Badea, C. Besleaga, M. Stefanescu













Figure 6. The 3D model of the
valve (ball and seat)

The model was meshed into
finite elements. In areas of interest, the meshing was refined. The model
of the subassembly meshed into finite elements is presented in the figure
7.


















Figure 7. The meshing of 3D model of the valve (ball and seat)

The state of equivalent tensions calculated according to the von
Mises criterion is shown in the figure 8 and 9.

11


Figure 8. The state of equivalent tensions (with von Mises criterion)

Figure 9. The state of equivalent tensions (with von Mises criterion)
I. Popescu, R.M. Negriu, S.G. Badea, C. Besleaga, M. Stefanescu


5. CONCLUSIONS

After analyzing the deteriorated balls and the theoretical
considerations for determining the tension states, the following
conclusions are reached:
5.1. The state of thermo-mechanical stress is not the main cause by
deterioration even to a large number of cycles, when both the balls and
their seats dont present defects or at least they are smaller than the
critical size specific to the solicitation area;
5.2. The main cause of deterioration is the wear phenomenon that no
longer allows the correct closing and because of this loss of fluid occurs,
so the pumping is ineffective;
5.3. Serious deteriorations can still occur due to wear phenomena or
internal defects produced in the manufacturing process which, combined
with the thermo-mechanical state of stress can lead to breakage of the
valves components;
5.4. Manufacturing balls with several layers can bring a considerable
increase in terms of resistance to wear but it can also introduce new
types of defects if the manufacturing technology is not properly applied.
The manufacturing of multilayer balls is a complex process that does not
require special preparation and sintering conditions.

REFERENCES
1. xxx, API Specification 11AX-2011.
2. Tudor A, Dumitru V,. Negriu R.M, Proc. Tribological Congress,
Vienna 2001.
3. Tudor A., Dumitru V., Negriu R.M., An in situ wear erosion-
corrosion study of carbide and ceramic composites in ball-valve of
crude petroleum extraction pump, 2nd World Tribology Congress,
Vienna, p.464., 2001.
4. xxx, Patents RO 112660, RO 114241, RO111844, RO112609,
RO119448.
5. P. Georgeoni, N. Arnici, I.C. Popescu, and s.a., The using of
isostatical pressing at the manufacturing of the large machine parts
with high performances from sintered metallic carbides,
Metallurgical Researches, ICEM, Vol.26, page. 463 - 475, 1985,
Bucharest.
6. P. Georgeoni, I. Popescu, Considerations regarding the
manufacturing of parts from metallic carbide type WC-Co for high
pressure devices, Metallurgical Researches, ICEM, Vol.26, pag.
477-483, Bucharest, 1985.
7. A. Semenescu, I. C. Popescu, T. Prisecaru, E. Popa, L. Mihaescu,
13


V. Apostol, F.E.M analysis of some type of cracks in high
pressure-high temperature devices, International Metal.
Publication, vol. XIV, (2009), nr. 12, p. 9-15, ISSN 1582-2214,
(rev. ISI, poz. 13 CNCIS CEN APOS), Bucharest, 2009.
8. I. C. Popescu, Introduction in computer aided analysis of the
process equipments, Printech Publisher, ISBN 973- 652- 951- 7,
Bucharest, 2004.
9. I.C. Popescu, T. Prisecaru, B. Finite elements Analysis of
Pressure Equipment, Computer Aided Engineering Solutions for
Design, Analysis and Innovation, (ANSYS & FLUENT User
Group Meeting), Sinaia, 26-27 Aprilie 2007.































15

MODICATION OF FAILURE RISK BY THE
USE OF HIGH STRENGTH STEELS
IN PIPELINES
J. Capelle* and G. Pluvinage **
LaBPS - Ecole Nationale dIngenieurs de Metz et Universit de
Lorraine, 1 route dArs Laquenexy, 57078 Metz, (France)
**Fiabilit Mcanique. Conseils Silly sur Nied (France).
pluvinage@cegetel.net

Abstract: The use of new generation of pipe steels with high yield stress
increases potentially the risk of brittle fracture. In order to evaluate this risk,
safety factors associated with a surface crack and an operating pressure have
been evaluated for three pipe steels: X52, X70 and X100. This evaluation has
been made using a Failure Assessment Diagram and SINTAP procedure. This
analysis has been extended to X120 pipe steel. The use of a Domain Failure
Assessment diagram indicates that for this steel a risk of elastic plastic fracture
exists. However, for pipe steels X52, X70 and X 100, failure occurs potentially
by plastic collapse.
Key words: High strength steels, pipe line, failure risk, failure assessment
diagram

1. INTRODUCTION

At present, requirement for natural gas is rapidly increasing
internationally. Pipelines are used for natural gas transmission over long
distance. Amelioration of gas transportation capacity is possible by
increasing pipe diameters, operating pressure, gas cooling, decrease of
the internal surface roughness and increase of service reliability. Several
studies have shown that the most efficient factors on gas transportation
capacity are in a decreasing order, pipe diameter, operating pressure
distance between compression stations, compression rate and service
temperature. By increasing the operating pressure and pipe diameter, the
gas transportation capacity is increased and this results in obvious
economic advantages. Table1 summarizes the evolution of pipelines
operating pressure and diameter over the last century.
Today several pipelines are built with 1420 mm pipe diameter.
The use of this large diameter pipes needs to use high strength steels in
J. Capelle, G. Pluvinage

order to avoid thickness difficult to weld and minimize steel weight.
There are significant advantages of using higher grade line pipes, such as
X100 even X120 grade pipeline, in constructing long distance pipeline,
because it can improve transportation efficiency of the natural gas
pipelines by increasing internal transportation pressure, and material cost
can be saved correspondingly by reducing wall thickness of pipe body
and consumable for girth welding However, there are still many
transportation safety problems laying high strength pipelines. First of all,
due to line pipes laid through complicated regions, such as earthquake
region with high-risk, gas pipelines in service may endure large
displacement and stress, the maximum flexure deformation at part of the
pipeline reaches to 4%~5% when it lays through multiple-region of
earthquake and geology casualty.

Table 1: Evolution of transportations Conditions in Gas Pipelines

Secondly, the increased pressure in modern pipelines also causes the
danger of running ductile cracks as the results of the stored high energy
content of the compressed gas.
Due to combined use of high strength steel, high operating pressure and
large diameter pipe, risk of brittle failure has increased.
By comparing remaining safety factor due to presence of crack like
defects, it is the possible to describe evolution of this risk versus time
through evolution of pipe design. This is made in the following by using
Failure Assessment diagram (FAD) and particularly SINTAP procedure.

2. MATERIAL

Three pipe steels have been studied X52, X70 and X100.Chemical
compositions of these steels are given in Table 2

Table 2: Chemical composition of the studied steels.
C Mn Si Cr Ni Mo S Cu
X52 0.206 1.257 0.293 0.014 0.017 0.006 0.009 0.011
Year Operating
Pressure
Diameter

Annual capacity

Power Gas
Consumption
over 6000Km
1910 2 bar 400 mm 80 10
3
m
3
49 %
1930 20 bar 500 mm 650 10
3
m
3
31%
1965 66 bar 900 mm 830 10
3
m
3
14 %
1985 80 bar 1420 mm 26000 10
3
m
3
11 %


17
X70 0.125 1.68 0.27 0.051 0.04 0.021 0.005 0.045
X100 0.059 1.97 0.315 0.024 0.23 0.315 0.002 0.022

Tensile properties (average values) are given in Table 3 and typical
stressstrain curves in figure 1. One notes that yield stress of the studied
steel is higher than the standard requirements and elongation at fracture
is strongly reduced when yield stress increase.

Table 3 : Tensile properties of studied steels X52, X70 and X100.
Youngs
modulus
(MPa)
Yield
stress
(MPa)
Ultimate
strength
(MPa)
Elongation
at fracture
%
API 5L X52 194 000 437 616 23 .14
API 5L X70 215 000 590 712 18.3
API 5L X100 210 000 866 880 6.75
Fracture toughness K
IC
and
c
have been determined using compact
tension specimen according to French standards NF A 03-180 [2] (K
Ic
)
and NF A 03-182 [3] (
c
). Specimen dimensions are extracted from 3
different pipe as given in Table 4

Figure 1 : Stress strain curves of API 5L X52, X70 and X100 pipe steels

Specimen dimensions are extracted from 3 different pipes as given
in Table 4
Table 4 Diameter and thickness and material of the 3 studied pipes.
Steel Diameter Thickness
API 5L X52 610 mm 11 mm
API 5L X70 710 mm 12.7 mm
API 5L X100 950 mm 16 mm
J. Capelle, G. Pluvinage

One note that pre crack is along the longitudinal direction of the
pipe. Critical load has been determined using acoustic emission which
determine crack initiation (subscript i). The obtained critical load is well
correlate with the traditional offset procedure failure load. Individual and
mean values are listed in Table 5.
Table 5 : Fracture Toughness of studied steels X52, X70 and X100.

3. FAILURE ASSESSMENT DIAGRAMME AND
SINTAP PROCEDURE

In a failure assessment diagram , the basic fracture mechanics
relationship with three parameters : applied stress (
app
), defect size (a)
and fracture toughness (K
IC
or J
IC
) is replaced by a two parameters
relationship f(k
r
, S
r
). Stress and defect size are combined into the applied
stress intensity factor (K
app
)or applied J parameter ( J
app
)

and the
parameter k
r
and S
r
are non-dimensional according to the following
initial definitions:

u
app
S
r
and
K
Ic
K
app
k
r
= =

(1)


where
u
is the ultimate strength. In the plane {S
r
; k
r
}, a given
relationship k
r
= f(S
r
) delimits the safe zone and the failure zone (figure
2). Initially, the relationship between non dimensional stress intensity
factor kr and non-dimensional stress S was issued from a plasticity
correction able to describe any kind of failure continuously from brittle
fracture to plastic collapse.
A typical representation of a failure assessment diagram is given in
figure 1.On the same figure, the load safety factor F
s
is defined according
to:

OC
OB
F
s
=
(2)
The advantages to the use of Failure Assessment diagram are:

-the use of a unique tool for any critical situations (in other way, several
failure criteria need to be used from LFM, EPFM and LA)
- to get, for any non-critical situation, the safety factor F
s
.

K
I,i
(MPam)
K
I,imean
(MPam)

i

(mm)

i,mean
(mm)
API 5L X52

CT1 97,59
95,54
0,21
0,18
CT2 93,49 0,14
API 5L X70
CT1 117,99
118,59
0,102
0,112
CT2 119,19 0,123
API 5L X100

CT1 159,98
151,82
0,125
0,108
CT2 143,66 0,091


19








Figure 2. Typical presentation of Failure Assessment Diagram (FAD).
Definition of safety factor.

The SINTAP procedure is derived from the initial failure assessment
diagram. However, definitions of non-dimensional parameters are little
different: k
r
parameter is derived from the applied J
app
parameter and
fracture toughness J
Ic


J
Ic
J
ap
k
r
= (3)
and the S
r
parameter is replace by the L
r
parameter

0
ref
P
L
P
L
r
= = (4)
where P is the applied load, P
L
the limit load. The material behaviour is
assumed to follow the RambergOsgood relationship:
( )

0
0 0
n
+ = (5)
where
0
and
0
are respectively the reference strain and stress and n the
strain hardening exponent. The reference stress is given by:


0
0
P
P
ref
= (6)
where P
0
is the reference load. The applied J parameter is obtained by
assuming proportionality between J
app
and the elastic value of J
parameter J
el
. The coefficient of proportionality is derived from the
r
k
r
L 1.00
1.00
O 0.75 0.50 0.25
0.25
0.50
0.75
Interpolating curve
SAFE ZONE
FAILURE
Assessment point,
SF=OC/OA>2
*
r
k
*
r
L
A
B
C
Interpolating curve
including SF=OC/OB=2
P
l
a
s
t
i
c

c
o
l
l
a
p
s
e
Brittle fracture
SECURITY ZONE
J. Capelle, G. Pluvinage

constitutive non dimensional stress strain relationship of the material.
The relationship between k
r
and L
r
is

considered as a limit curve obtained
from numerous experimental data. This limit curve is more physically
an interpolation curve between brittle fracture representative assessment
point and plastic collapse. In this method, failure near plastic collapse is
represented by data in the tail of the diagram.

There are several similar Failure Assessment Diagram procedures i.e.
EPRI in USA; R6 in UK, RCCMR in France with small and more and
less conservative difference in the safe zone area. The SINTAP [4]
procedure is the result of a European project of a multi-disciplinary
approach in order to get an unify multi-level method useful for SME to
large companies. The level hierarchy depends on knowledge of
description of stress strain curve and fracture toughness. Lower levels
are used with simple description of stress strain curve but with higher
conservatism. The mathematical expressions of SINTAP procedure for
the lowest and more conservative (basic level) is given as below:
( )
| |
( )
| |
|
|
.
|

\
|
=
|
|
.
|

\
| +
=
(

< +
(

+
+
(

+
=

U
Y
U
U Y max
r
Y
max
r r
2N
1 N
r

2
1
r
L
2
1
2
r

1 0.3 N ,


2
1
L , ,0.6

E
0.001 min where
L L 1 L e 7 0 3 0
2
1
1
(7) 1 L 0 , e 7 0 3 0
2
1
f(L
6
r
, . .
. .
L
)
r

where ) f(L
r
,
r
L ,
max
r
L ,
Y
,are interpolating function, non-dimensional
loading parameter, maximum value of non-dimensional loading or
parameter, yield stress, respectively.

4. PIPE DEFECT AND ASSOCIATED STRESS
INTENSITY FACTOR

We have chosen to study a surface longitudinal semi-elliptical crack in
the wall of a pipe. This can of defect represent in a conservative way, the
crack-like defect approach, the most current type of defect detected in
pipe such as corrosion defects, gouges, scratches etc.
The stress intensity factor for such a crack is given by the general
formula:

(8)



21
Where p is the internal pressure, R
int
is the internal radius of the
pipe, t the wall thickness, a the crack depth, M the geometrical
factor correction and the elliptic integral of second species.

(9)

An approximate value of this elliptic integral is given by:

(10)

5. RESULTS

Three cases have been studied and corresponding to different steels.
Operating pressure is considered higher for X100 steel because it is used
for new generation of pipe lines working at higher operating pressure
and higher diameter.

Table 6 : List of the studied cases
Steel 2R
int

(mm)
t
(mm)
Operating
pressure
(bars)
Crack
depth
(mm)
Crack
ratio
(a/c)
API 5L X52 610 11 70 2.2 0.4
API 5L X70 710 12.7 70 2.54 0.4
API 5L X100 950 16 100 3.2 0.4

k
r
parameter as been determined using equation (1) and (8) and L
r
using
equation (1).For each case, an assessment point with coordinates (L
r
*,
k
r
*) and reported in a Failure assessment diagram (Figure 6). Each steel
has its own failure assessment diagram because the parameter is
different for each steel. However the difference is relatively small
particularly for L
r
< 0.8. We note that the three assessment points are in
the safe zone i.e below the failure curve given by equation (1). Then,
using the procedure described in figure 4, the safety factor is then
determined and reported in table 7

Table 7 Safety factor according to pipe steel.
Steel API 5L
X52
API 5L
X70
API 5L
X100
Safety factor 3.38 3.87 3.23
One notes that safety factors are more than 2 for all steels. According to
this conventional value, pipe is safe and defect doesnt need to be
repaired.

6) DISCUSSION
J. Capelle, G. Pluvinage


The previous results indicates that the safety factor decreases when we
change the pipe design using high strength steel like X100. In this case,
we increases pipe diameter and thickness and operating pressure
simultaneously with pipe yield stress. In order to have an idea of the
consequence of new pipe design with API 5L X120 steel, safety factor
was determined using the following data.

Table 8 : API 5L X120 steel pipe design conditions.
Diameter
(mm)
Thickness
(mm)
Operating
pressure
(bars)
Crack depth
(mm)
Crack ratio
(a/c)
1420 23 120 4.6 0.4

The diameter has been chosen as the biggest actual pipe diameter and
the thickness is compatible for the seam welding of the X120 pipe with
the submerged arc welding (SAW) method with one pass each for the
inside and outside welds, which had been employed for conventional
grades. Operating pressure has the expected value for future.
Due to unavailability of X120 pipe steel, mechanical properties (yield
stress and ultimate strength) are obtained from [6] and are reported in
Table 9. Fracture toughness is deduced from two required values of
critical CTOD
c
in base metal and in welds at temperature -20c
given in table 1. CTOD is converted into Fracture toughness using
the following LFM relationship:

(11)

Table 9 : mechanical properties of API 5L X120 steel
Yield stress
(MPa)
Ultimate
strength
(MPa)
CTOD
Base metal
(mm)
CTOD
Welds
(mm)
908 981 0.14 0.08

Required Crack Tip Opening Displacement (CTOD) was calculated on
an assumption of the existence of a surface-breaking crack 2 mm in
depth at a seam weld toe and possible shape irregularity and stress
distribution. As a result, it was concluded that a CTOD of 0.08 mm or
more was good enough. Since a defect equal to or larger than 2 mm is
detected at a non-destructive inspection and an internal defect up to 4
mm in width will be permissible under the same value of critical CTOD.
Ones notes that safety factor decreases when the yield stress of the pipe
steel increases together with diameter, thickness and operating pressure.
Evolution of failure type when increasing yield stress of pipe steels can
be predicted by using a Domain Failure Assessment Diagram (DFAD).


23
A domain failure assessment diagram is a failure assessment diagram
divided in three zones of potential failure type: brittle fracture, elastic
plastic failure and plastic collapse. A D FAD is limited by the failure
assessment curve that gives the limit of a safe and an unsafe pipe. The
safe area is divided conventionally into three zones:
Zone I: if the assessment point lies in this zone, increasing the
applied pressure leads to brittle fracture
Zone II: where increasing the applied pressure leads to elasto-
plastic fracture
Zone III: where plastic collapse occurs by increasing service
pressure.


Figure 3: Values of safety factors associated with different pipe steels.


Figure 4 : Domain Failure assessment diagram and assessment points for
the 4 studied pipe steels.

Based on Feddersen diagram [8] the limit of these three zones is defined
conventionally as follows:
J. Capelle, G. Pluvinage

Zone I 0 < L
r
< 0,62 L
r,y

Zone II 0,62 L
r,y
< L
r
< 0.95 L
r,L
Zone III 0,95 L
r, max
< L
r
< L
r,max

where L
r,y
is associated with the yield pressure and L
r,max
is the maximum
value of L
r
. In figure 4, in a domain failure assessment diagram are
reported the assessment point of the 4 studied pipe steels. One notes that
X52, X70 and X100 have a fully ductile failure potential. However, the
X120 steels as a more pronounced risk of elastic plastic failure.

7 CONCLUSION

The risk of failure for a steel pipe has been evaluated through a
conventional defect type. Under operating pressure safety factor is
always over the conventional value of 2. It can be concluded that is not
necessary from a fracture mechanics point of view to repair this defect.
The use of Domain failure assessment diagram gives in addition the
potential of brittle or elastic fracture risk. It has been seen that X120 has
an elastic plastic failure potential risk. In this case, it seems necessary to
evaluate in addition risk of brittle running crack. This risk is associated
with high stored energy due to large pipe diameter and high operating
pressure.

REFERENCES

[1] 6th EGIG report 1970 2004, , Gas Pipeline Incidents, Doc.
Number EGIG 05.R.0002, December,( 2005).
[2]Norme AFNOR : NF A 03-180, Dtermination du facteur dintensit
de contrainte critique des aciers, (1981).
[3]Norme AFNOR: NF A 03-182, Mcanique de la rupture,
Dtermination de lcartement fond de fissure (CTOD), (1987).
[4] SINTAP: Structural integrity assessment procedure. Final Revision,
EU-Project BE 95-1462, Brite-Euram Programme, Brussels, (1999).
[5] G. Pluvinage et V.T. Sapunov Fuite et rupture des tubes
endommags, Cpadus Edition I.S.B.N 2.85428.6448,(2004).
[6]Nippon steel technical report Development of Ultra-high-strength No.
90, july Line pipe, X120, (2004).
[7] G Pluvinage , J.Capelle , C. Schmitt and M. Mouwakeh Notch and
domain failure assessment diagram s as tool for defect assessment of gas
pipes. To appear.
[8 ] C.E Feddersen Evaluation and prediction of residual strength of
center cracked tension panels ASTM STP, p 50. (1970).


25

Thermo-Mechanical Analysis of a Copper Mould for
Continuous Casting of Steel

M. Ansoldi
1
, G. Bazzaro
1
, D. Benasciutti
2
, F. De Bona
2
, G. Luvar
1
, L.
Moro
2
, M. Gh. Munteanu
2
, F. Vecchiet
1


1
Centro Ricerche Danieli, Buttrio (UD), Italy
2
Department of Electrical, Management and Mechanical Engineering, University of Udine,
Italy
Email address: denis.benasciutti@uniud.it

Abstract: This work deals with the thermo-mechanical analysis of a copper mould.
In the continuous casting process molten steel flows through a water-cooled mould
which induces the solidification of the outer shell. The inner part of the mould
undergoes a huge thermal flux and large temperature gradients develop across the
copper inducing high stress and plastic deformation, especially in the region close to
the meniscus. After the operating period, a cooling to room temperature induces
residual stresses which may increase with repeated thermal cycling over a campaign.
Another source of cyclic thermal loading during operative condition is represented
by the fluctuation of melt metal level into copper mould, with a resulting variation
of the temperature peak on the surface of the mould. In this work the structural
behaviour of the mould under thermal loading condition is analyzed adopting a
three-dimensional finite-element model, with the aim of performing an accurate
evaluation of stress and strain levels. An analytical structural model is then
developed with the aim of performing a sensibility analysis in the design phase. A
simplified thermal fatigue approach has also been followed, in order to gain insights
into cyclic behaviour and improve mould life.
Key words: continuous casting mould, thermal stress, total strain range, fatigue life.
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet

1. INTRODUCTION
In the last few years, the demand of a strong improvement in term of
productivity and reliability, accompanied by cost reduction have been
fundamental requirements in the design of steelmaking equipment. The well
established practice of over sizing the most critical mechanical elements can
not be followed anymore, and the necessity to consider such complex
phenomena as plasticity, creep, low-cycle thermal fatigue, phase transition is
crucial in order to improve the life of the mechanical parts, to gain safety,
and to assure high steel quality [1, 2].
This work focuses on the mould design, a crucial component for the
process of continuous casting of steel. The mould (or crystallizer) controls
the shape and the initial solidification of steel, governing heat transfer and
the surface quality of the product. A reliable, crack-free mould within close
dimensional tolerances is a key factor to guarantee a suitable level of safety,
reliable quality and top productivity. The molten steel induces high thermal
fluxes and temperature gradients into the copper, which in turn generate high
stress levels. Subsequent sequences with start-ups and shut-downs, as well as
free-surface (meniscus) oscillations during normal service, lead to cyclic
thermal loading which may damage the mould [3, 4].
The aim of this work is to understand the mechanical behaviour of the
mould under thermal loads, in order to relate stress-strain cycles to the life of
the component and to identify the actions that can improve its durability.
2. COMPONENT DESCRIPTION
The mould is a mechanical component through which the molten steel
flows. It is designed to solidify a thin shell of metal that is continuously
withdrawn away up to a complete through thickness solidification. Different
cross sections may be adopted (square, rectangular or rounded shapes)
according the final geometry of the product (billets, blooms or slabs).

27

Molten
steel
Solidified
shell
Meniscus
Mould


Figure 1. Geometry of the component

The main function of the mould is to provide an intensive cooling of the
steel to achieve a robust shell of metal. It is required a precise control of the
shape to match the shell contraction and to guarantee the product geometry.
High dimensional stability at any operational regime is therefore a
requirement of primary relevance.
Conventional moulds consist of a copper tube surrounded by a steel
jacket: in the gap between the two elements cooling water flows [5]. In order
to improve the thermo-mechanical performances, a different design
configuration of the mould has been developed. It consists of a thicker
copper tube provided with drilled holes for cooling. In this way a high
stiffness with excellent heat transfer capacity are achieved at the same time.
This work focuses on this enhanced type of mould, schematically
represented in Figure 1.
3. EXPERIMENTAL INVESTIGATION OF A
MOULD AFTER SERVICE
Due to the presence of molten steel, the inner part of the mould is
subjected to a high thermal flux, with a characteristic profile decreasing from
top to bottom. The peak values of heat flux are found to be in the proximity
of the meniscus region, while lower values are experienced when the steel
shell becomes thicker. The meniscus zone is found to be the most critical for
the component: several works in literature account for the presence of cracks
in this location. Sometimes cracks were observed after only 2 or 3 casting
sequences [4].
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet



Figure 2. Inner mould surface with closed view of the meniscus region with
cracks

In thin-slab continuous casters the mould surface is periodically machined
to remove cracks [6], but for billet casting this practice cannot be adopted
due to restrictions in the cross section. Figure 2 shows the inner surface of a
mould after several sequences of production at 30% productivity higher than
conventional. The portion of the mould subjected to higher heat loads clearly
appears as a darker surface. As enhanced by the extended view, in this
region several micro-cracks appear after operation.



Figure 3. Manufacturing operation to obtain the tensile-test specimens.

In order to characterize the mechanical properties of the copper alloy after
use, more than 20 test specimens have been obtained (see Figure 3) in
different locations. The results of the tensile tests are reported in Table 1.

Table 1: Mechanical characteristics of CuCrZr alloy
Mean value Standard deviation
Modulus of elasticity 130 [GPa] -
29

Yield strength 260 [MPa] 21
Ultimate tensile strength 320 [MPa] 17
Ductility (Eq. 8) 0.84 0.05

The values obtained show a limited scatter confirming that the material
properties are not significantly affected by service condition.
4. THERMAL LOADING OF THE MOULD
The component undergoes two different classes of loading cycles (see
Figure 4 and Figure 5). The first one, which may be referenced as macro-
cycle, is characterized by a load cycle between the condition of uniform
room temperature and that one corresponding to the maximum heat flux
during the steady production. It represents the interval time production
between a sequence start-up and shut-down.

Time
Flux
1
0 3
2
T
0

Figure 4. Scheme of the load macro-cycle

The second loading condition, which may be named micro-cycle, is
representative of metal-level fluctuations normally occurring during casting
conditions.
Whereas the first case occurs in a quite long period of time compared to
that required to achieve the steady state condition, in the second case the
frequency is high enough to establish a continuous shift in the temperature
map across the nominal meniscus position. For this reason, whether in the
former case a static thermal analysis could be a satisfactory approximation,
for this latter condition a transient thermal analysis is compulsory, as long as
a full steady state condition is never established.

M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet

-10
0
10
0 5 10
100
200
300
Time [s]
Heat Flux
Meniscus fluctuation
Mould
Local temperature fluctuations

Figure 5. Scheme of micro-cycle

5. THERMO-MECHANICAL ANALYSIS:
NUMERICAL RESULTS AND SIMPLIFIED
ANALYTICAL MODELS
The component is characterized by 4 planes of symmetry, therefore it is
possible to adopt a reduced model as represented in Figure 6. Even if a plane
approach could be useful as a preliminary thermo-mechanical analysis, due
to the non-uniform distribution of the thermal flux in the direction of the
mould longitudinal axis, a three-dimensional (3D) model is necessary.

B
A


Figure 6. Top view of 3-D FE mechanical model

The thermal analysis is performed considering an imposed thermal flux
acting on the inner part of the mould. The outer surface is characterized by
adiabatic condition. A convective boundary condition is imposed in the inner
surface of the water cooling channels. A non-linear thermal analysis is
required in order to take into account the variation of thermal conductivity
and specific heat with temperature.
31

As the problem is uncoupled, a subsequent mechanical analysis is
performed imposing the nodal temperature distribution obtained from the
previous analysis. The in-plane and axial thermal expansion of the
component are allowed. It follows that stress-strain behaviour depends only
on the internal temperature distribution. Also in this case a non-linear
analysis is required. In fact the dependence of Young's modulus and yield
stress with temperature needs to be considered. According to [7] starting
from the measured values presented in Table 1a correlation with temperature
is imposed.
In this section a linear elastic behaviour is previously considered, in order
to gain insights in the physic of the problem. As it will be shown in the
following, in the most critical portion of the mould, stresses exceed the
elastic limit and therefore an elastic-plastic model is needed. For this
purpose a bilinear model with Von Mises plasticity and kinematic hardening
rule will be adopted. The results of the elastoplastic analysis, which differ
from the linear case only in limited localized areas, will be presented in the
following section.
Figure 7 shows the temperature distribution obtained at the maximum flux
in the steady state condition. It can be noticed that the maximum temperature
occurs in the region close to the meniscus. If a section orthogonal to the
mould axis in proximity of the meniscus is considered, it can be shown that a
relevant "radial" temperature gradient occurs up to the inner portion of the
cooling pipes. In the outer part of the mould a quite constant room
temperature is observed.
T / T
max
0
0.06
0.13
0.20
0.26
0.33
0.40
0.47
0.53
0.60
0.66
0.73
0.80
0.86
0.93
1

/
max
0
0.06
0.13
0.20
0.26
0.33
0.40
0.47
0.53
0.60
0.66
0.73
0.80
0.86
0.93
1

(a) (b)

Figure 7. Temperature (a) and stress (b) distribution during operative condition

Figure 7(b) shows the Von Mises stress distribution corresponding to the
previous thermal analysis. It can be noticed that the maximum stress occurs
close to the corner (point B in Figure 6), although higher temperatures occur
at point A. The outer part of the mould shows a uniform negligible stress
state. The stress distribution around point B can be interpreted using a simple
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet

structural model which refers to a square frame filleted at the corners
undergoing a thermal gradient across its thickness. If the temperature
variation is linear it causes a uniform bending moment and therefore
maximum stresses occur at the corners which behave as curved beams.
Figure 8 shows the "hoop" stress variation along the thickness evaluated
with the finite element (FE) model (r
e
represents the outer curved beam
radius), compared with that obtained according to Winkler theory [8]: a quite
good agreement can be noticed.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
r/r
e

m
a
x


Curved beam theory
FE model

Figure 8. Stress variation along the thickness at the corner

Even if stress in point B reaches the maximum value, it occurs in the
colder portion of the inner part of the component (see Figure 7), at a safe
distance from water. It follows that the most critical portion of the mould is
located at the meniscus mid-face (point A).
In order to develop an interpretative model of stress distribution in this
area it is useful to investigate the principal stress pattern along the thickness.
In Figure 9 are represented respectively axial, "hoop" and "radial" stresses
(i.e. the principal stresses in the axial direction and those contained in a
plane orthogonal to the longitudinal axis and respectively parallel and
perpendicular to the inner mould surface). It can be noticed that the latter
stress assumes negligible values in the whole range. In the inner mould
surface axial and "hoop" stresses show similar compressive values which
decrease almost linearly maintaining quite comparable values.
33

0 0.2 0.4 0.6 0.8 1
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
x / t

m
a
x

Radial
Hoop
Axial


Figure 9. Stress variation through the thickness in the hottest part

It can therefore be concluded that the hottest portion of the mould
undergoes a plane hydrostatic state of stress. A possible interpretative model
can be developed considering the inner surface of the mould as a semi-
infinite plane where a single point located on the surface is heated at
temperature T. As it is well know the solution in term of stresses is:

= =
1
h a
T E
, 0
r
= (1)
In other words a small heated part would freely expand but it is "laterally"
constrained by the surrounding large cold portion.
Figure 10 shows the principal stress evaluated on the mould surface along
the axial direction (y=0 corresponds to the upper edge of the mould whose
height is l ). It can be clearly noticed that a plane hydrostatic stress state still
occurs and it is directly proportional to temperature, according to Eq. (1).
From a physical point of view the component can therefore be considered as
constituted by two layers: an inner hot layer (that would expand) constrained
by a colder layer that is maintained at low temperature by water cooling.
0 200 400 600 800 1000
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
y / l

m
a
x


Hoop
Axial
Temperature
0
0.2
0.4
0.6
0.8
1
T

/

T
m
a
x

Figure 10. Principal stress on the mould surface along the axial direction.
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet


It is thus possible to conclude that a suitable interpretative structural
model of the stress and strain status in the area surrounding point A could be
that of a hollow cylinder whose external part (from a radius r
f
corresponding
to the position of inner surface of the water channel) is maintained at
constant temperature.

r
i
r
f
r
e
T(r)
A
B
Hot
region
Cold
region
T
h
e
r
m
a
l
F
l
u
x


Figure 11. Hollow cylinder constituted by a "hot" and a "cold" part.

The inner surface undergoes a thermal flux which produces a variation of
temperature according to the following relation:
( )

> =
< +
|
|
.
|

\
|
|
|
.
|

\
|

=
f f
f f
f
f
i
f i
) (
log
log
r r T r T
r r T
r
r
r
r
T T
r T
B
A

(2)
The stress distribution and the radial displacement can thus be obtained
considering the structure as composed by two parts. By following the
procedure proposed in [9] for the case of a hollow cylinder clamped at its
ends and undergoing a given temperature distribution, the solution has the
following expression:

35

( )
( )
|
.
|

\
|

+
+

+ =

2
2 1
2
2 1 1 1
1
1 i
r
C C E r ET
dr r r T
r
E r
r


( ) |
.
|

\
|

+
+

=

2
2 1
2
r
2 1 1
1
1 i
r
C C E
dr r r T
r
E r
r


( )
( )( )

2 1 1
2
1
1
z
+
+

=
EC r ET

( )
r
C
r C dr r r T
r
u
r
r
2
1
i
1
1
1
+ +

+
=


(3)

In this case, due to the fact that the temperature is described by two
functions four constants C
i
have to be determined. The values of C
i
can be
analytically obtained by imposing at the interface (r=r
i
) the compatibility
condition in terms of radial displacements and the continuity of stresses:

( )
( )
( ) ( )
( )

=
=
=
=
f
B
f
A
f
B
r f
A
r
e
B
r
i
A
r
) (
0
0
r u r u
r r
r
r


(4)
In this way the expression of
1
C ,
2
C and
'
1
C
'
2
C can be obtained
respectively for the inner and the outer part:


( )( )
( )
( )
( )
3 f
2
i
f
2
3
f 1 2
f
'
2
1
3 f
2
i
f
2
3
f 1 2
f
2
e
'
1
2
i
2 1
2
f
2
e
f
2
i
f
3
2
f
2
i
1
2
f
2
e
f 1 2
f 3
2
2
f
2
1 1
2 1
1 1
1
1
1 1
2 1
1 1
1
1 2 1
2 1
1 1 1
2 1
1
1
1 1 1
1
1
2 1 1
1 1 1 1
1
1 1
1
1 1
1
M r r
r
C
M
r M M
r
C
M
M r r
r
C
M
r M M
r r
C
r
C C
r r r r
r
M r r
M
r r
r M M
r M
M
r
C
|
|
.
|

\
|
+ +
|
|
.
|

\
|

+
=
+
(
(

|
|
.
|

\
|
+ +
|
|
.
|

\
|

+
=

=
|
|
.
|

\
|

|
|
.
|

\
|
+
+

|
|
.
|

\
|

+
+
+
|
|
.
|

\
|

|
|
.
|

\
|

+
+
+


(5)
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet

where:
( )( )
( )
( ) ( )
( ) ( )
( )
( )
f
2
e
f
3
2
i
2
f
f 2
i
f 2
f
f
2
i
i
2
i
2
f
f
2
f
f
i
f i
2
2
f
2
e
2
e
1
1
2 1
2 2
ln
2
ln
4
ln
2 4
ln
2
ln
2 1
2 1 1
r
r
r
M
r r
T
r
r
r
r r
r
r r
r
r
r
r
T T
M
r r
T
r
M
f
+ =
+
|
|
.
|

\
|
+ +

+
=



(6)
Figure 12 shows the stress in radial, hoop and axial direction. A cylinder
free to expand in axial direction is considered (this condition is the most
similar to the actual mould state) and therefore a uniform axial stress needs
to be superposed in the third of Eq. (3) in order to obtain a null resultant
force at the ends.
0.7 0.75 0.8 0.85 0.9 0.95 1
-1
-0.5
0
0.5
r/r
e
[mm]

m
a
x


Radial
Hoop
Axial

Figure 12. Stress distribution according to the proposed structural model.

By comparing Figure 12 with Figure 9 it is possible to notice a significant
similarity especially in the more stressed area. Similar results have been
obtained imposing the actual thermal flux to a hollow cylinder and
comparing the stress distribution along a path parallel to the component axis
and passing through point A.
In conclusion, from a thermal stress point of view, temperature produces a
stress distribution characterized by two critical zones, respectively, close to
the fillet and in the hottest region. In the first case this is due to the thermal
moment produced by the temperature gradient which acts on a curved beam-
like structure. In the second and most critical case the high compressive
stresses occurring in the inner surface are due to the constraining effect
imposed by the outer cold part. This behaviour can be well described by a
simple analytical model, which refers to a hollow cylinder.
37

6. THERMO-MECHANICAL ANALYSIS UNDER
CYCLING LOADS AND LIFE ASSESMENT
Figure 13 shows the stress-strain relation in point A when the mould
undergoes the two types of load cycles; only "hoop" stresses are reported,
since similar trends can be obtained considering axial stresses. In the case of
the "start-up and switch-off" cycles, after the first heating, a compressive
stress is produced which strongly exceeds the yield strength of the material.
In the subsequent cooling phase (point 3 of Figure 4) residual tensile stresses
are produced. A value of equivalent stress slightly higher than the yield
stress of the material is reached. The subsequent cycles are therefore
characterized by the typical elastoplastic hysteretic loop. In the case of the
micro cycles due to the meniscus oscillation, a similar behaviour is
produced, but in this case the yield stress of the material is exceeded only in
the first heating, therefore the following cycles occur only in the elastic
domain.
The durability of the critical zone in the inner part of the mould was then
considered. As the component undergoes cyclic strains and stresses induced
by thermal loads and plastic deformations occur, a strain-based thermal
fatigue approach can be adopted. The durability analysis of continuous
casting moulds is deeply investigated in literature [6, 10, 11, 12]. As it is
well known [4] the usual method to relate strain and life refers to the
evaluation of the plastic strain range which the material undergoes. In this
work a different approach has to be followed. In fact, as previously pointed
out, the plastic component of strain is significant only in the case of the
macro-cycles. Moreover a more general remark has to be introduced. In fact
the mechanical component here considered is characterized by self-imposed
constrained thermal expansion as a consequence of cyclic temperature
gradients. It follows that the total strain depends only on temperature
distribution and it is independent from the elastoplastic model of the
material. On the contrary, the relative amount of plastic and elastic strain is
strongly influenced by the material model implemented to perform the
numerical analysis. Therefore simplified elastoplastic models usually
adopted (kinematic or isotropic hardening, etc.) could lead to different
results in term of plastic strain. Recently, more accurate models (i.e. mixed
kinematic isotropic) have been proposed in literature [13]. When strains are
due to the combination of mechanical and thermally induced loads, these
methods are probably the only choice to obtain accurate results in terms of
plastic strain range. On the other hand, in the particular case described in this
work, such approach seems to be of less practical support. This is due to the
fact that, as the stabilized stress-strain cycle has to be obtained, a unfeasible
computational effort would be required. In addition, in the present study
cyclic (stabilized) material properties were not experimentally assessed.
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet


-4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0
-300
-200
-100
0
100
200
300

tot

m
a
x

Macro-cycle
Micro-cycle

Figure 13. Stress-strain macro-cycles and micro-cycles

According to the previous consideration a total strain range approach, as
suggested in [14, 15] is adopted. It is therefore necessary to introduce a
suitable relation between total strain range and number of cycles to failure.
Such relation is proposed in [14] and it consists of two power-law terms, one
for the plastic strain, and the other for the elastic strain:
y
f
x
f pl el tot
YN XN + = + = (7)
where
tot
is the total strain range; X, x and Y, y are the coefficient and
exponent terms relating to elastic (
el
) and plastic (
pl
) strain range
respectively; N
f
represents the number of cycles to failure.
All the coefficients and exponents proposed in Eq.(7) need to be
determined experimentally; in the case of the copper alloy considered in this
work experimental data of low cycle-fatigue test are available in terms of
plastic strain only [6]. Other works present correlation between total strain
range and life, but in different temperature-testing conditions or for different
copper alloys [12, 10].
The Universal Slopes method proposed in [16] could be a alternative
approach of practical use. In fact it relates only parameter obtained from
tensile test (ultimate tensile strength, ductility, and modulus of elasticity) to
fatigue life for a given strain range. According to this approach, for all
materials the elastic and plastic lines have slopes of 0.12 and 0.6,
respectively. One point on each of these two lines is determined considering
the intersection on the strain axis at N
f
equal to 1.0. For the elastic line, this
intersection point depends only on the parameter S
UTS
/E where S
UTS
is the
ultimate tensile strength, and E is the elastic modulus. For the plastic line,
the intersection point is related only on ductility, defined as:
39

|
.
|

\
|

=
RA
D
% 100
100
ln (8)
where RA is the area reduction in a tensile test. According to Universal
Slopes method, Eq.(1) becomes:
0.60
f
6 . 0 0.12
f
UTS
tot
5 . 3

+ |
.
|

\
|
= N D N
E
S

(9)
As previously stated, a life estimation of the inner part of the mould has
to be performed. In this region a biaxial stress state occurs, thus an
equivalent strain range has to be computed, according to:
( ) | | ( ) | | ( ) | |
2
3 2
2
3 1
2
2 1 eq
3
2
+ + = (10)
where (
i

j
) is the range of the relative difference between principal
strains
i
and
j
. The elastic part in Eq.(9) must be consequently corrected as
suggested in [14], while the plastic line remains unchanged. The curve for
total strain range is thus displaced slightly downward, the displacement
being the greatest in the region of high-cycle fatigue where the elastic
component dominates. The following relation is finally obtained:
( )
0.60
f
6 . 0 0.12
f
UTS
eq
5 . 3 1
3
2

+ |
.
|

\
|
+ = N D N
E
S
v
(11)
At elevated temperatures, where creep and environmental interaction may
occur, this method has been found to be non-conservative. As it is suggested
in [16] this is due to the fact that intercrystalline cracking essentially
bypasses the large number of cycles required to initiate a crack in the sub-
creep range. Experimental tests on a wide range of materials point out that
approximate results could be obtained by assuming that life under creep and
environmental interaction conditions could cause as much as 90% loss of
cyclic life, leaving only 10% of that calculated by the Universal Slopes
Method, thus giving rise to the development of the so called 10% rule. The
Universal Slopes Equation gives the upper-bound life, while the 10% rule
gives the lowest expected life. Median expected life is estimated to be two
times the lower bound life. The advantage of this method is its simplicity,
since only the tensile properties need to be known at the desired temperature.
Although accuracy is limited, the uncertainty related to measurements
justifies this approach.
The resulting curve is plotted in Figure 14. Values proposed in literature
for copper alloy with chemical composition that only slightly differs from
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet

that of the material used in this work are also reported. It can be noticed that
the curve proposed in [10] refers to a series of test performed at room
temperature; these data fits well with the Universal Slopes Equation without
temperature correction. In [12] experimental test were performed at 300C;
also in this case the obtained results are in good agreement with USE
corrected with the 10% rule. These comparison seems therefore to confirm
the correctness of the procedure proposed in this work.
10
2
10
3
10
4
10
5
10
6
10
7
10
-3
10
-2
10
-1
Cycles to failure, N
f
S
t
r
a
i
n

R
a
n
g
e


Univ. Slopes Eq. (Upper Bound)
10% Rule (Lower Bound)
20% Rule (Median Life)
[Marmy]
[Preston]

Figure 14. Equivalent strain range versus cycles to failure.

The results of the durability analysis are summarized in Table 2. It can be
noticed that the number of cycles to failure is strongly lower in the case of
the macro-cycle, characterized by a higher level of total strain range. On the
other hand this result could be misleading. In fact, in a single operating
sequence characterized by a start up and a switch off, a huge amount of
micro-cycles occurs, due to the high frequency of the meniscus oscillation.
Therefore, a lower number of sequences is obtained from micro-cycles.

Table 2: results of the durability analysis
frequency
eq
N cycles N sequences
Macro cycle 1/20 hours 0.0040 6358 6358
Micro cycle 0.25 Hz 0.0027 25350 1.5

In conclusion, the durability of the component has to be evaluated in term
of casting sequences; from this point of view it can be noticed that an
opposite result is obtained, i.e. in the failure analysis of the component,
strain range produced by the micro-cycle loads has to be primarily accounted
for.
41

7. CONCLUSIONS
This work deals with the thermo-mechanical analysis of a copper mould.
The mechanical behaviour of the component under thermal loading
condition is analyzed adopting a three-dimensional finite-element model,
with the aim of performing an accurate evaluation of stresses and strains
levels. Two critical zones have been detected, respectively in the hottest
region of the inner surface (point A) and close to the corner (point B). An
analytical structural model is then developed with the aim of performing a
sensitivity analysis in the design phase. A curved beam model seems suitable
to describe the stress distribution around the corner (point B). A hollow
cylinder with an imposed temperature distribution fits the hydrostatic plane
stress status observed at the mid-face surface close to the meniscus region
(point A). Using static tensile tests data performed on specimens obtained
from actual mould, durability curves have been obtained adopting the
Universal Slopes Method. The values of the total strain ranges for the two
stress-strain cycles that characterize the component operation have been
finally evaluated, estimating the lifespan of the component either in terms of
number of cycles and in terms of casting sequences.
The developed models could be a useful support for the prediction of the
residual life in real operation if the tracking of the component load history is
performed; moreover the effective evaluation of different design
improvement strategies can be also achieved.
REFERENCES
1. J. K. Park, B.G. Thomas, I.V. Samarasekera. Analysis of
thermomechanical behaviour in billet casting with different
mould corner radii, Ironmaking and Steelmaking, 29:117, 2002.
2. A. Weronski, and T. Hejwowski. Thermal Fatigue of Metals,
Marcel Dekker, New York, 1991.
3. J. K. Park, B. G. Thomas, I. V. Samarasekera, U. S. Yoon.
Thermal and mechanical behavior of copper molds during thin-
slab casting (I): plant trial and mathematical modeling.
Metallurgical and Materials transactions, 33B: 112, 2002.
4. J. K. Park, B. G. Thomas, I. V. Samarasekera, U. S. Yoon.
Thermal and mechanical behavior of copper molds during thin-
slab casting (II): mold crack formation. Metallurgical and
Materials transactions, 33B: 437449, 2002.
5. I. V. Samarasekera, D. L. Anderson, J.K. Brimacombe. The
thermal distortion of continuous-casting billet molds.
Metallurgical Transactions 13B: 91104, 1982.
M. Ansoldi, G. Bazzaro, D. Benasciutti, F. De Bona, G. Luvar,
L. Moro, M. Gh. Munteanu, F. Vecchiet

6. T. G. O'Connor and J. A. Dantzig. Modeling the thin-slab
continuous-casting mold. Metallurgical and Materials
Transactions, 25B: 443457, 1994.
7. ITER Material Properties Handbook, ITER Doc. No. G 74 MA 9
01-07-11 W 0.2, Publication Package No. 7, 2001.
8. E. Winkler. Formnderung und Festigkeit gekrmmter Krper,
insbesondere der Ringe. Der Civilingenieur, 4, 232246, 1858.
9. S. Timoshenko and J. N. Goodier. Theory of elasticity. McGraw-
Hill, 1952.
10. P. Marmy and O. Gillia. Investigations of the effect of creep
fatigue interaction in a Cu-Cr-Zr alloy. Proc. Engineering 2:
407416, 2010.
11. A. A. F. Tavassoli. Materials design data for fusion reactors. J
Nuclear Materials, 258263, 8596, 1998.
12. S. D. Preston, I. Bretherton , C. B. A. Forty. The thermophysical
and mechanical properties of the copper heat sink material
intended for use in ITER. Fusion Engineering and Design 66
68, 441 446, 2003.
13. J. H. You, M. Miskiewicz. Material parameters of copper and
CuCrZr alloy for cyclic plasticity at elevated temperatures.
Journal of Nuclear Materials 373, 269274, 2008.
14. S. S. Manson. Thermal stress and low-cycle fatigue, McGraw-
Hill, 1966.
15. J. A. Graham. Fatigue Design Handbook, Advances in
Engineering, Vol. 4, Soc. of Automotive Engineers (SAE), 1968.
16. S. S. Manson, G. R. Halford. Fatigue and durability of structural
materials, ASM International, 2006.
43

FAILURE ASSESSMENT OF TRANSMISSION
DIODE LASER WELDED POLYPROPYLENE

E. GHORBEL, G. CASALINO, A. BEN HAMIDA
1
L2MGC- Univ. Cergy Pontoise, Neuville sur Oise 95 031 Cergy Pontoise CEDEX France
2
DIMeG, Politecnico di Bari, Viale Japigia, 182, 70126, Bari.
3
Universit Pierre et Marie Curie, Bote 161, 4 place de Jussieu, 75 252 Paris Cedex 05, France
Email address:
1 elhem.ghorbel@u-cergy.fr
2 g.casalino@poliba.it
3 Ahmed.Benhamida@enis.rnu.tn

Abstract: The aim of present work is to predict using finite element analysis the
failure of polypropylene welded samples using the Laser Diode Transmission
technique. The studied material is a commercial polypropylene (PP). The
mechanical behaviour of the welded samples is modelled using elastic purely plastic
law with a von Mises yield criterion. Both the thermal degradation and the
heterogeneity of the weld zone were taken into account in the modelling, to
approach the welding conditions and the geometry of the weld. The linear
mechanics fracture was adopted to assess the failure of the welded samples. The
assumption of the plane deformations is chosen to calculate the integral of contour
of Rice J and to characterize the singularity of the stress field surrounding the
weld zone. Good agreement is. The good agreement observed between the predicted
and this experimentally obtained load failure suggests that our approach can be used
as a tool for the prediction of failure of laser diode welded specimens.
Key words: polypropylene, diode laser welding, failure assessment.
E. Ghorbel, G. Casalino, A. Ben Hamida

1. INTRODUCTION
Polypropylene (PP) is one of the most widely used polymers. The
combination of low density, chemical resistance, low cost and a balance of
stiffness and toughness allow thermoplastics to play a leading role and
replace other materials in many important applications.
Two general forms of laser welding of plastics exist: direct laser welding
and transmission laser welding. Direct laser welding usually uses CO2 laser
radiation, which is readily absorbed by plastics, allowing quick joints to be
made, but limiting the depth of penetration of the beam and restricting the
technique to film applications. The shorter wavelength radiation produced by
Nd:YAG, fibre and diode lasers is less readily absorbed by plastics, but these
lasers are suitable for performing transmission laser welding. In this
operation, it is necessary for one of the plastics to be transmissive to laser
light and the other to absorb the laser energy, to ensure that the heating is
concentrated at the joint region. Alternatively, an opaque surface coating
may be applied at the joint, to weld two transmissive plastics. Transmission
laser welding is capable of welding thicker parts than direct welding, and
since the heat affected zone is confined to the joint region no marking of the
outer surfaces occurs.
Among the properties of thermoplastics, deformation and ultimate tensile
strength have revealed as a topical preoccupation in order to utilize them
effectively in service applications, since almost every application, even the
most trivial, involves some load bearing capability.
Therefore the increase in using polymer welds for several industrial
applications leads to a strong need of developments of constitutive models.
These models are based on either phenomenological or microscopic aspects.
The purpose of this paper is to assess the influence of microstructure
heterogeneity on the mechanical behaviour of polypropylene (PP)
thermoplastics welded by diode laser.
The mechanical behaviour of flat samples under uniaxial load was studied
by numerical and experimental approaches. The ultimate strength of a
thermoplastic is naturally important in the matching of a material to an
application, as with deformational properties. Some preliminary trials
brought the authors to adopt the mechanics of fragile rupture as the most
suitable law to predict ultimate tensile strength.
2. PROPERTIES OF PROPYLENE AND ITS WELD
Both pure polypropylene and 2% carbon black filled polypropylene were
used in this investigation. The most important material properties related to
these materials are summarized in Table 1.
45

Table 1: The most important material properties related to the welding behaviour of
polypropylene. Obtained by DSC at 20C/mn (Tg : vitreous transition, Tf : fusion et
Tc : cristallinization ).
materials T
g
(C) T
f
(C) T
c
(C) X
c
(%)
PP -18 170 110 31
PP (2%C) -16
173
112 33

The laser beam is totally absorbed within the surface (interfacial) of
carbon black filled propylene. Direct contact between the parts ensures
heating of polypropylene at the joint interface. Welding occurs upon melting
and fusion of both materials at the interface. The heating and melting of the
polymer is started from absorbed energy black part.
50mmx50mmx3mm narrow plaques were used as sample geometry.
Figure 1 shows the welding setup.


Figure 1. Welding setup.

Effects of the influence of laser power density and irradiation time
(welding speed) on the weld structure was investigated as well as the
material parameters (black content) on weld soundness and heat affected
zone and bead shape were performed [1]. From microscopic observations it
appears that the cross section of the joining area has an elliptic geometry.
The centre of the ellipse is located at the interface of both polymers. The
increase of the laser power leads to a larger volume of the welding zone with
a more important depth penetration. The increase of velocity conducts to a
decrease in the volume of the welding zone that geometry remains elliptic
IR results as well as microscopic observations showed the thermal
gradients produces an evolution of the cristallinity fraction along the cross
E. Ghorbel, G. Casalino, A. Ben Hamida

section of the welding joint, a change on the morphology of the spherulites
and the occurrence of a void due to the thermal decomposition and the
vaporization of the polypropylene related to the presence of high
temperature gradients in the seam. Hence, schematically, different zones in
the welded part can be distinguished. In each zone, crystalline morphology is
related to specific thermal cycle. The aim of this investigation is to study the
influence of the microstructure and cristallinity rate on the failure of the
welded samples loaded under uniaxial tensile conditions.
Several models can be used for the prediction of polypropylene weld
failure: damage models, fracture mechanics, non-linear mechanics.
Uniaxial tensile tests were conducted on welded samples. The obtained
results show that the behaviour of the specimens is linear until the fracture
which was initiated in the proximity of the weld bead and developed into the
black part (Figure 2).
0
50
100
150
200
250
300
350
400
450
0 0,2 0,4 0,6 0,8 1 1,2
dplacement(mm)
F
o
r
c
e

(
N
)
P=3,7 W
P=2,2 watts
F
1
5
m
m
7
5
m
m
10 mm
F

Figure 2. Tensile test curve for a polypropylene weld obtained with 1,5 mm/s
welding speed.
3. MECAHNICAL MODEL OF THE WELD
Several works have dealt with the mechanical behavior of polypropylene.
Uniaxial tensile tests show that the behavior is not linear at room
temperature. The brittle-to-ductile temperature increases with increased
cristallinity, which also increases the yield stress. Mechanical properties
strongly depend on the cristallinity rate (X
c
), the spherolite size () and the
molar mass (MW) [2, 3, 4]. Table 2 shows the quasi-linear relation between
the cristallinity rate and the mechanical properties. More authors
investigated the relation between the cristallinity rate and the PP rupture. [5,
6]. Plastic yielding and fracture influence the entropy change in
47

polypropylene through a significant change in the dissipation factor. An
increase in crystallinity results in an increase in the entropy [7].
In this paper it was conceived that the mechanical behavior of the both
transparent and pigmented layers can be described using an elastic perfectly
plastic law with a Von Mises criterion (Eq. N1). Material properties of each
layer depend essentially on the crystallinity ratio X
c
(table 2).

Table : PP properties against the rate of cristallinity [VAN 98 I et II]..
X
C
(%) E (MPA)
s
(MPa)
s
(%)
31 601 19.5 17.0
38 962 25.4 11.0
43 1228 31.9 8.8
53 1650 41.0 4.0

( )
1
0
0 et : avec
eq s
e
e p p
f Tr I
E E
f f
f d



+
= = =


= = + =



For transparent PP X
c
=31% whereas different situations are considered
for the weld: the weld bead is assumed as a homogeneous and isotropic
medium with X
c
=37% and the weld bead is a heterogeneous medium
decomposed on three homogeneous domains. Each domain is characterized
by its own crystallinity ratio. The plane deformation condition was given to
the weld in the proximity of weld bead, which tackled the singularity due to
the domain change due to the different material properties. Rice integral was
calculated at the point of singularity (J). In case of straight fissure and fragile
crack G and J are connected by the following equation Ed. 2:

( )
2 2
2 2
1
2
III
I II
K
G J K K
E

= = + +

The tensile test was performed on a weld manufactured with P
laser
=40 W
et V
laser
= 3mm/s. The heat affected zone was given an elliptic shape. The
mesh was made up of 2 degree triangular elements. At the proximity of the
weld bead the elements of the mesh were square.
The weld heterogeneity was modeled by means of three wrapping
ellipsis. Each of three ellipses had different mechanical properties, as
showed in figure 3.
Figure 4 shows the geometrical model of weld. A displacement along the
x direction of 1mm during 10 seconds is imposed.

E. Ghorbel, G. Casalino, A. Ben Hamida

zone 3: E
1
= 650 Mpa, =0.35,

s
=21 Mpa;
zone 2 E
2
= 750MPa; =0.35;

s
=23 Mpa;
zone 1: E
1
= 850 Mpa, =0.35,

s
=25 MPa.


Figure 3. Ellipsis with different mechanical properties.


Figure 4. Geometrical model of weld.
4. NUMERICAL RESULTS
Figure 5 shows the force (N) against the displacement (mm) during the
simulation of the tensile test for two different set of the mechanical
properties of the weld. It appears clearly that the microstructure inside the
beam does not influence the response of the loaded welded samples.



Figure 5. Numerical simulation of the tensile test.

As a consequence the heterogeneity of the weld can be neglected. Figure
6 shows the Von Mises equivalent constraint. A bending phenomenon is
noted corresponding to experimental observations.
Deplacement (mm)
F
o
r
c
e

(
N
)

49

The highest value of von Mises equivalent stress is detected inside the
weld with a maximum located at the proximity of the seam leading to the
highest equivalent plastic strain (figure 7). These results explain
experimental observations during tensile tests consisting in the fact that
fracture was initiated in the proximity of the weld bead and developed into
the black part.



Figure 6. Von Mises equivalent constraint.

Considering that the restitution energy of the pure propylene is about 3
J/m2 [4] and that the black content reduces it, the rupture of the weld can be
placed at 0.8 mm displacement, which correspond at a load of about 400N
(see figure 8). The failure load obtained from finite element analysis is quite
equal to that obtained experimentally.

0
0,5
1
1,5
2
2,5
0 1 2 3 4 5 6 7 8
Distance partir du centre du cordon (mm)
D

f
o
r
m
a
t
i
o
n

p
l
a
s
t
i
q
u
e

c
u
m
u
l

e
0
1

Figure 7. Highest equivalent plastic strain.
5. CONCLUSION
The mechanical behavior of laser joined of polypropylene was
investigated by numerical and experimental analysis. The numerical
simulation of weld mechanical strength and rupture took into account the
material heterogeneity, which depends on weld recristallization and the
E. Ghorbel, G. Casalino, A. Ben Hamida

material degradation during the cooling phase of the thermal cycle, which
causes also material lack.
The numerical results gave the idea that that heterogeneity can play a
minor role in the mechanical strength and rupture of the weld. The
displacement at fracture was calculated by means of the Rice integral.
The obtained results can be used for a deeper understanding of the PP
weld mecahnical behaviour when deformation is due to a superimposed
displacement.
REFERENCES
[1] Ghorbel, Casalino et Abed R. Characterization of thermal and
mechanical effects of diode laser welding of Polypropylene.
MATERIALS & DESIGN, Materials and Design 30 (2009) 27452751.
[2] Van der Wal A., Mulder J.J., Thijs H.A. et Gaymans R.J. Fracture of
Plypropylene : 1. The effect of molecular weight and temperature at low
and high test speed Polymer vol. 39 No. 22, pp. 5467-5475, 1998.
[3] Van der Wal A., Mulder J.J. et Gaymans R.J. Fracture of Plypropylene:
2. The effect of crystallinity Polymer vol. 39 No. 22, pp. 5477-5481,
1998.
[4] Avella, M. DellErba, R., Martuscelli, E. et Ragosta, G. Influence of
molecular mass, thermal treatment and nucleating agent on structure and
fracture toughness of isotactic polypropylene, Polymer, Vol. 34, N 14,
pp. 2951-2960, 1992.
[5] Mareri, P., Bastide, S., Binda, N. et Crespy, A. Mechanical behaviour pf
polypropylene composites containing fine mineral filler: effect of filler
surface treatment Comp. Sci. Tech. 58, 747-752, 1998.
[6] Zebarjad, S. M., Tahani, M. et Sajjadi, S. A., Influence of fillers
particles on deformation and fracture mechanism of isotactic
polypropylene, J. of Mat. Processing Tech., article in press, 2004.
[7] T. Vu-Khanh, M. El Majdoubi. Entropy change with yielding and
fracture of polypropylene. Theoretical and Applied Fracture Mechanics
51 (2009) 111116.
51

Crack Tip Plastic Zone under Mode I Loading and
the Non-singular T
zz
-stress

Yu.G. Matvienko
Mechanical Engineering Research Institute of the Russian Academy of Sciences
Email: ygmatvienko@gmail.com

Abstract: The amplitudes of the second order terms in the three-dimensional series
expansion of the crack front stress field are the terms
xx
T
and
zz
T
which describe
in-plane and out-of-plane constraint, respectively. All previous analyses of the crack
tip plastic zone have ignored the effect of
zz
T
-stress. At the same time, the effect of
the
zz
T -stress on crack tip plastic zones is not revealed heretofore. It is therefore
very important to obtain solutions for crack tip plastic zone size taking into account
two components of the T-stresses. The present study focus on theoretical and
numerical analysis of the joint effect of the non-singular
xx
T and
zz
T -stresses on
sizes of the plastic zone in the vicinity of the crack tip under mode I loading
conditions. The three-dimensional crack tip stresses including
xx
T
and
zz
T
stresses
are incorporated into the von Mises yield criteria to develop an expression that
models the crack tip plastic zone. Calculations are performed for three thicknesses
of the CT specimen. The predicted sizes of the plastic zone in the vicinity of the
crack tip of the analysed CT specimens are bounded by sizes of the plastic zones
corresponding to two special conditions, namely, plane stress and plane strain. The
theoretical results are compared with the results computed by FEM. Theoretical
estimations of the plastic deformations zone size with provision for T-stress
components in whole shows the satisfactory results, especially on line of the crack
continuation.

Key words: plastic zone, T
zz
-stress, mode I crack
Yu.G. Matvienko

1. INTRODUCTION
The different sources of a change in in-plane constraint at the crack tip
are associated with crack size, geometry of specimen and type of loading.
The source of a change of the out-of-plane crack tip constraint is thickness.
To describe in-plane and out-of-plane constraint effects in fracture analysis,
the following parameters can be used, namely,
z
T
-parameter [1], local
triaxiality parameter h [2] and the non-singular terms in Williams series
expansion of the crack tip stress fields [3].
These parameters considerably influence on the fracture toughness [4-8].
Not emphasize attention on advantages and disadvantages of the above-
mentioned constraint parameters, we concentrate on the non-singular
components of the T-stresses at the crack tip. The second order terms
xx
T
and
zz
T in Williams series expansion are defined as T-stresses, and
they are the only non-zero and non-singular terms. It should be noted that
xx
T
has been simply referred to as T-stress.
xx
T
and
zz
T
represent the
stresses in the crack surface plane normal to and tangential to the crack front,
respectively.
In a two-dimensional (2D) crack configuration,
zz
T is related to
xx
T
by
xx zz
T T =
under plane strain conditions, where is Poissons ratio. It is
well-known that the sign and magnitude of the
xx
T
-stress substantially
change the size and shape of the plane strain crack tip plastic zone [9-11].
Therefore, the
xx
T
-stress has been used to characterize the effect of in-plane
constraint on the crack tip plastic zone.
The amplitudes of the second order terms in the three-dimensional series
expansion of the crack front stress field are the terms
xx
T
and
zz
T which
describe in-plane and out-of-plane constraint, respectively. All previous
analyses of the crack tip plastic zone have ignored the effect from
zz
T -
stress. At the same time, the effect of the
zz
T -stress on crack tip plastic
zones is not revealed heretofore. It is therefore very important to obtain
solutions for crack tip plastic zone size taking into account two components
of the T-stresses.
The present paper focuses on theoretical and numerical analysis of the
joint effect of the non-singular
xx
T and
zz
T -stresses on sizes of the plastic
zone in the vicinity of the crack tip under mode I loading conditions.
53

2. THEORETICAL ANALYSIS OF THE PLASTIC
ZONE
2.1. Modeling of the plastic zone
The general form of the linear elastic crack tip stress fields within a three-
dimensional crack problem can be characterized by the singular (the first
order term) and the nonsingular terms (second order terms) [3]
...
2
3
sin
2
sin 1
2
cos
2
+ +
|
.
|

\
|

=
xx
I
xx
T
r
K

, (1)
...
2
3
sin
2
sin 1
2
cos
2
+
|
.
|

\
|
+

r
K
I
yy , (2)
...
2
cos
2
2 + +

=
zz
I
zz
T
r
K


, (3)
...
2
3
cos
2
sin
2
cos
2
+

r
K
I
xy , (4)
0 =
yz

,
0 =
zx

. (5)
Here, r and are the in-plane polar coordinates of the plane normal to the
crack front centered at the crack tip with 0 = corresponding to a line ahead
of the crack (Fig. 1), and are the normal and shear stress respectively,
I
K
is the mode I stress intensity factor (SIF), E is Youngs modulus, is
Poisson's ratio. The singular term corresponds to the stress intensity factor.
The terms
xx
T and
zz
T are the amplitudes of the second order terms in the
three-dimensional series expansion of the crack front stress field in the x and
z directions, respectively. These terms characterize corresponding crack tip
constraint along above-mentioned axes.
The
xx
T -stress component can be calculated from equations (1) and (2) as
difference between
xx
and
yy
stresses very near the crack front. The
value of
zz
T -stress components is defined according to the following
relationship [4]
xx zz zz
T E T + =
. (6)
where
zz
is strain along the crack front.
Yu.G. Matvienko


Figure 1: Three-dimensional coordinate system for the region along the crack front

In the case of plane stress conditions (2D stress state), stress
component 0 =
zz
. For plane strain conditions, stress component
zz
is
equal to
( )
xx
I
yy xx zz
T
r
K
+

= + =


2
cos
2
2
. (7)
In the present work, the plastic zone ahead of the crack tip is determined
by the von Mises yield criterion
( ) ( ) ( )
( )
2 2 2 2
2 2 2
2 6
Y zx yz xy
xx zz zz yy yy xx


= + + +
+ + +
, (8)
where
Y
is the yield stress.
Substituting of equations (1)-(5) into the von Mises yield criterion (8), the
crack tip plastic zone size
p
r r =
can be estimated. After comprehensive
mathematical manipulations, the solution for
p
r
as a function of and the T-
stress components is given by the following formula
2 2
2
2
1
Y p
I
I
I I
p
r
K
D
A K
r

=
|
|
.
|

\
|
|
|
.
|

\
|
+

. (9)
55

The joint effect of the non-singular
xx
T and
zz
T -stresses on sizes of the
plastic zone in the vicinity of the crack tip under mode I loading conditions
is included into basic Eq. (9).
The parameters in Eq. (9) are denoted as follows
( ) |
|
|
|
.
|

\
|
+
+
|
.
|

\
|
+
|
.
|

\
|
+
=
zz
zz
xx xx
I
T
T
T T
D
2
cos 16
2
cos 8
2
5
cos 3
2
cos
2

, (10)
( ) ( ) ( ) 1 2 cos
4
3
cos 1 2 1
2
+ =
I
A
. (11)
Equation (9) can be solved to determine the angular distribution of the
plastic zone size in the vicinity of the crack tip
| |
2
2
2
2 , 1
4
4
1
) ( W U V V
U
r
p
+ =
, (12)
where parameters U, V,

W are
2
4
Y
U =
,
I I
D K V = ,
I I
A K W =
2
. (13)
Finally, solution (12) can be written in more representative form
( )
2
2
2 2
2
2
2 , 1
16 2
16
) (
(
(

=
Y
I Y I I
Y
I
p
A D D
K
r


. (14)
It can be shown that special solution for the crack tip plastic zone in the
case of plane strain conditions under mode I loading follows from general
solution (12) (e.g., [11]):
2
4
T
U =
,
I I
E K V = ,
I I
B K W =
2
. (15)
Here, the additional coefficients are
( ) ( ) ( ) 1 2 cos
4
3
cos 1 2 1
2
+ =
I
B
. (16)
( )
( ) |
|
|
|
.
|

\
|
+
+
|
.
|

\
|
+
=
xx
xx xx
I
T
T T
E

2
cos 16
2
2
cos 8
2
5
cos 3
2
cos
2
. (17)
Yu.G. Matvienko

For plane stress conditions, the angular distribution of the plastic zone at
the mode I crack tip can be calculated from (12) taking into account the
following coefficients
2
4
T
U =
,
I I
F K V = ,
I I
C K W =
2
, (18)
( ) ( ) 1 2 cos
4
3
cos 1 + =
I
C
. (19)
|
|
.
|

\
|

|
.
|

\
|
+ =
xx I
T F
2
5
cos 3
2
cos
2

. (20)
It should be noted that equation (12) has to meet certain conditions,
namely
0 4
2 2
+ W U V 0 U . (21)
If above-mentioned inequalities (21) satisfy, sizes of the crack tip plastic
zone are calculated as
U
W U V V
r
p
2
4
2
1
+ +
=
,
U
W U V V
r
p
2
4
2
2
+
=
. (23)
The value of
1
p
r
is positive in the wide range of coefficients U, V, W and
should be used in all calculations. At the same time value of
2
p
r
has a
negative sign.

2.2. The effect of thickness on the plastic zone
For the 3D model, the CT specimen is considered. The plastic zone is
analyzed for a wide range of the ratio B/W=0.25-0.5 within the limits of the
pure plane stress and pure plane strain conditions. The stress intensity factor
K
max
of experimental specimens of Steel JIS S55C is 66 MPam
1/2
,
corresponding values of the T-stress components are presented in table 1[7].
The T-stress components are calculated along the plastic zone boundary on
the line of the crack extension. According with slight variation of the
xx
T -
stresses for the CT specimens with different ratio B/W, the value of
xx
T -
stresses has been assumed to be constant and equal to 182 MPa.
The fracture tends to initiate at the specimen thickness center, the values
of the crack tip plastic zone size at specimen thickness center were chosen to
represent the characteristic intensity of these values. The predicted results of
the angular distribution of the plastic zone sizes in the vicinity of the crack
tip at the specimen thickness centre for the CT specimens with various
thicknesses to width ratio are given in Fig. 2. To determine the validity of
the plastic zone model derived above, basic equation is applied to a pure
57

mode I situation under plane strain conditions, and the results agree with
published solutions [9, 11].

Table 1: The T-stress components on the line of the crack extension (boundary of
the plastic zone)
B/W 0,25 0,40 0,50
T
xx
, MPa 186.59 182.36 176.28
T
zz
, MPa -159.47 -106.81 -84.97



Figure 2. The angular distribution of the plastic zone size at the specimen thickness
centre of the CT specimens

The predicted sizes of the plastic zone in the vicinity of the crack tip of
the analysed CT specimens are bounded by sizes of the plastic zones
corresponding to two special conditions, namely, plane stress and plane
strain. Moreover, the shape and size of the plastic zones tend to typical
plastic zones for plane strain conditions, when specimen thickness increases.
Thus, the results confirm the necessity to take into account the constraint
effect at the crack tip on the plastic zone by means of both non-singular
stresses
xx
T and
zz
T .

Yu.G. Matvienko

3. NUMERICAL MODELLING THE PLASTIC
ZONE
3.1. Finite Element Analysis
To demonstrate the validity of the plastic zone model, finite element
analysis was conducted.



Figure 3: The geometrical model of the CT specimen

It is well-known that the creation of the calculation model lies in the basis
of numerical experiment. Primary tasks, which are being solved for this
purpose, are the following. First of all, the solid-state geometrical model of
the CT specimen (fig. 3) is created in the modern CAD system. Three
geometrical models with the ratio B/W = 0,25, B/W = 0,40 B/W = 0,50 are
prepared for an analysis of the effect of the T-stress components on the
plastic zone. Deviation from the figure was that the crack length a, was set at
the nominal value of 12.5 mm (a/W= 0.5) for all cases.
The finite element model of the CT specimen is represented in Fig. 4. For
the reason of minimization of the required computing resources finite element mesh
was created on the one half of the geometric model (fig. 4). For more detailed
account of the stress and strain distribution at the crack tip the local mesh
concentration with diameter 10 mm was created around crack tip.
Spatial finite element mesh was created in one of the modern CAE system
by means of decomposition of the internal volume of the geometrical model
to the finite number of the small calculation elements having spatial polygon
shape. Parabolic elements (the second order elements) with tetrahedral shape and
one intermediate node along each side were used in this numerical analysis. The use
of such elements allows achieving greater calculation accuracy due to more accurate
59

reproduction of the curvilinear surfaces of the geometric model, as well as more
accurate shape function which connects displacement of the arbitrary point of the
calculation element with displacement of its nodes.



(a) Two- dimensional global mesh


(b) Three-dimensional global mesh

Figure 4: Finite element model of the CT specimen.

The creation of the load system is made in the modern CAE system.
Fixed values of the vertical forces (force P) were added to the nodes situated
on the cylindrical surface of the specimen hole. The value of K
max
in the table
2 was obtained as the stress intensity factor corresponding to the maximum
load P
max
from the well-known equation in ASTM E399.

Table 2: Loading conditions of the CT specimen
B/W=0,25 B/W =0,40 B/W =0,50
P
max
, kN 6,0 9,6 12,0
K
max
,
MPam
1/2

66,0 66,0 66,0
Yu.G. Matvienko


Finite element calculations allow analyzing size and shape of the plastic
zone in the vicinity of the crack tip by means of diagrams of the von Mises
equivalent stresses.

3.2. Algorithm for the processing of results of the numerical modeling
Sizes and shape of the plastic zone were analyzed using the image data of
the distribution of equivalent von Mises stresses in the vicinity of the crack
tip.
Estimation of the influence of the T-stress components on the plastic zone
size is made in the mathematical package, for what special algorithm is
created for processing the computed diagrams (Fig. 5). These diagrams must
satisfy certain graphic conditions. The diagram must present itself as
monochrome images of the equivalent stress fields with clear gray gradation
of stress from white (stress level equal to yield stress
Y
) to black (zero stress
level). In according with the main idea of algorithm the special procedure
realizes consequent selection of pixels, which belongs to the image of the
current diagram, along radius from the crack tip with appropriate step. If the
color of current pixel differs from white, the procedure of pixel selection is
stopped and current values of radius and angle are saved. This radius
corresponds to the boundary of the plastic zone in the vicinity of the crack
tip.


Figure 5: Typical image data of the distribution of equivalent von Mises stresses in
the vicinity of the crack tip of the CT specimen
4. DISCCUSSION
In according with the above-mentioned methodology for theoretical
estimation of the plastic zone size, the predicted results are compared with
the results computed by FEM. Comparison of theoretical and FEM results of
the plastic zone size estimation at the CT specimen thickness is presented at
Fig. 6.
61



Figure 6: Comparison of theoretical and FEM results of the plastic zone estimation
at the CT specimen thickness

It is observed that deviation of the theoretical plastic zone sizes from the
FEM results does not exceed 20% in angular intervals (0, 3045) and
(90100,135145) (Fig. 7). Deviation achieves maximum and exceeds
40% in angular intervals from 135145 to 180 that can be connected
with particularity of the algorithm which is used during processing of the
equivalent von Mises stress diagrams.


Figure 7: Deviation of the theoretical plastic zone sizes from the FEM results
Yu.G. Matvienko


Probably, so high divergence between results of the numerical experiment
and analytical calculation in interval (0145) can be explained that
actually -stresses are not a constant and depend on angle . This
assumption is corroborated by the results of the FEM analysis (Fig. 8, Fig.
9).


Figure 8: Finite Element estimation of angular distribution of the T
xx
-stress along the
boundary of the plastic deformations zone at the specimen thickness centre


Figure 9: Finite Element estimation of angular distribution of the T
zz
-stress along
the boundary of the plastic deformations zone at the specimen thickness centre
63


We can receive evidence that dependence between T-stress components
and angular coordinate exists if we will express K
I
from equation (2) and
substitute received relation to the equation (1). As a result we will receive
angular distribution of
xx
T :
|
|
|
|
.
|

\
|

+
=
2
5
cos
2
cos 5
2
5
cos
2
cos 3



yy xx xx
T
. (27)
In similar manner we can receive angular distribution of
zz
T :
|
|
|
|
.
|

\
|
+
|
|
|
|
.
|

\
|

=
xx yy zz zz
T T
2
5
cos
2
cos 5
2
cos 8


. (28)
Received relations confirm made early assumptions about necessity of the
introduction into asymptotic formulas (1) and (3) angular distributions of
components of the nonsingular stresses.
5. CONCLUSIONS
The theoretical analysis of the joint influence of the nonsingular
components of the T-stresses on the size of the plastic deformations zone at
the tip of the mode I crack is carried out with attraction of asymptotic
formulas, which taking into account triaxiality of the stress state at the tip of
the mode I crack and Mises yield criterion.
The size of the plastic zone at the middle surface of the specimen
decreases during increasing of the specimen thickness. It reflects the
increasing of the deformations constraint degree at the crack tip by means of
increasing of nonsingular stress
zz
T .
Theoretical estimations of the plastic zone size with provision for T-stress
components in whole shows the satisfactory results, especially on line of the
crack continuation. This is especially important in case of estimation of the
reliable values of the fracture toughness. However, in some cases divergence
between analytical calculation and FEM results exceeds 20%. For more
correct determination of the shape of the plastic deformations zone it is
necessary to take into account angular distribution of the nonsingular T-
stresses at the crack tip.
Yu.G. Matvienko

Certainly, accurate analysis and estimation of the plastic zone sizes at the
mode I crack tip will promote the development of more correct criterions of
validity in estimation of the fracture toughness of engineering materials.
REFERENCES
1. W. Guo. Three-dimensional analyses of plastic constraint for through-
thickness cracked bodies. Engineering Fracture Mechanics, 62: 383-
407, 1999.
2. B.S. Henry and A.R. Luxmoore. The stress triaxiality constraint and the
Q-value as ductile fracture parameter. Engineering Fracture Mechanics,
57: 375-390, 1997.
3. T. Nakamura and D. M. Parks. Determination of elastic T-Stress along
three-dimensional crack fronts using an interaction integral. International
Journal of Solids and Structures, 29:1597-1611, 1992.
4. W.A. Sorem, R.H. Dodds and S.T. Rolfe. Effects of crack depth on
elastic plastic fracture toughness. International Journal of Fracture, 47:
105126, 1991.
5. S. Liu and Y.J. Chao. Variation of fracture toughness with constraint.
International Journal of Fracture, 124: 113-117, 2003.
6. X.K. Xhu and Y.J. Chao. Specimen size requirements for two-
parameter fracture toughness testing. International Journal of Fracture,
135: 117136, 2005.
7. T. Meshii and T. Tanaka. Experimental T
33
-stress formulation of test
specimen thickness effect on fracture toughness in the transition
temperature region. Engineering Fracture Mechanics, 77: 867-877,
2010.
8. H.M. Meliani, Yu.G. Matvienko and G. Pluvinage. Two-parameter
fracture criterion (K
,c
-T
ef,c
) based on notch fracture mechanics.
International Journal of Fracture, 167: 173-182, 2011.
9. J.R. Rice. Limitations to the-scale yielding approximation for crack-tip
plasticity. Journal of the Mechanics and Physics of Solids, 22: 17-26,
1974.
10. S.G. Larsson and A.J. Carlsson. Influence of non-singular stress
terms and specimen geometry on small-scale yielding at crack tips in
elastic-plastic materials. Journal of the Mechanics and Physics of
Solids, 21: 263-278, 1973.
11. Q. Nazarali and X. Wang. The effect of T-stress on crack-tip plastic
zones under mixed-mode loading conditions.
Fatigue and Fracture of Engineering Materials and Structures,
34: 792803, 2011.

65

FAILURE ANALYSIS OF A STORAGE PUMP
SHAFT

Niculai PASCA, Liviu MARSAVINA, Sebastian MUNTEAN, Radu
NEGRU
Affiliation: 1) Politehnica University Timisoara, 2) Romanian Academy , Timisoara Branch
Email address: nicu_pasca21@yahoo.com
Abstract: The shaft is an important component in hydraulic pumps used to transmit power
from electrical motor to the pump impeller. The paper presents failure analysis of a shaft,
from a double suction hydraulic pump in operation of approximately 30 years, in a storage
station. The shaft material is a Romanian steel, equivalent after their chemical composition
and mechanical characteristics with American SAE 4340 steel for which in literature are
available necessary data for failure analysis. The paper is structured in two parts: finite
element analysis of the pump shaft and an analytical failure analysis for a circumferential
crack type, using the failure assessment diagram (FAD). In a modeling program for stress and
strain analysis the 3D shaft was loaded in torsion. The numerical results show the maximum
stress zones, the stress concentration effect, and the possibility of crack occurrence. For
circumferential crack type the failure assessment diagram for Mode III loading were plotted
using the stress intensity factor solution. The results indicate the unsafe zone respectively the
critical circumferential crack depth, where the shaft cannot operate with defects. This study
presents an opportunity related to safe operation condition and remaining life estimation for a
storage pump.
Key words: finite element analysis, stress, storage pump shaft, failure assessment diagram
(FAD)
N. Pasca, L.Marsavina, S. Muntean, R. Negru

1. INTRODUCTION
Storage pumps all over the world are integrated equipment in
hydropower facilities. These pumps are required to allow primarily large
flow rate, in addition to operate at high efficiency and tolerant cavitation
conditions. In Figure 1 is shown the cross section and the storage pump
components: 1 impeller, 2 suction elbow, 3 volute, 4 shaft, 5 radial bearing ,
6 radial-axial bearing.


Figure 1. Storage pump cross section.

Table 1: Technical characteristics of the storage pump
Parameters Symbol Value
Pumping Head H [m] 197
Flow rate Q [m
3
/s] 4
Hydraulic Power P [kW] 8840
Efficiency [%] 87.5


The pump shaft fails after approximately 30 years of operating time. In
Figures 2 and 3 are presented the shaft fracture surface zone compared with
theoretical aspect of a shaft failure in torsion loading condition [1]. This
leading to hypotheses that shaft failed in torsion loading condition. In
literature other studies present aspects of shaft failure in operating condition
[2], [3], and [4]. In [2] the authors present a numerical turbine shaft failure


67
analysis with a crack propagation analysis using ANSYS and Afgrow
software. An analysis of the premature failure of two counter shafts used in
centrifugal pumps for lifting slurry has been carried out by Das et al. [3]. A
collection of pump shaft failures like, corrosion failures on shafts, fatigue
failures on shafts and shaft sleeve failures are analyzed [4]. However, these
papers have not tackled the failure analysis using failure assessment diagram
in mode III.



Figure 2. Fracture surface of the
investigated shaft.
Figure 3. Sketch of fracture surface in
torsion loading condition [1].

This paper is structured in five parts which include a problem
introduction, the shaft material, a stress and strain numerical analysis, and
mode III failure diagram for a circumferential crack in a shaft. The
conclusions are drawn in last section.
2. STORAGE PUMP SHAFT MATERIAL
The shaft material is a Romanian steel named after STAS 791-66,
34MoCN, with the following chemical composition presented in table 2.

Table 2 Chemical composition and mechanical characteristics according to STAS
791-66 [5] for storage pump shaft material.
Steel name Chemical composition [%]
34 MoCN 15
C Si Mn Cr Ni Mo
0.3-0.38 0.17-0.37 0.4-0.7 0.6-1.7 1.4-1.9 0.15-0.3
Mechanical characteristics
Yield Stress
y
[MPa] Tensile strength
r
[MPa]
882.9 1079.1

N. Pasca, L.Marsavina, S. Muntean, R. Negru

Material properties and specially fracture toughness are necessary for
failure analysis. Because for the Romanian steel these parameters are not
presented in literature, and equivalent steel was searched. In table 3 are
presented the material characteristics of the equivalent steel SAE 4340 [5]. It
can be observed the chemical composition for both materials are very close.

Table 3 Chemical composition and mechanical characteristics for equivalent steel
SAE 4340 [6]
Steel name Chemical composition [%]
SAE 4340
C Si Mn Cr Ni Mo
0.37-0.4 0.23 0.7 0.7-0.9 1.83 0.2-0.3
Mechanical characteristics
Yield stress
y
[MPa]
Tensile strength
r
[MPa]
834 931

The fracture toughness for mode III of 4340 steel which is used in failure
assessment diagram for mode III construction K
IIIc
=68.235 MPam is
presented [7].

3. STRESS AND STRAIN ANALYSIS
The stress and strain analysis for 3D geometry obtained in modeling
software has been performed. The 3D shaft geometry is shown in Figure 4.
Using the numerical software ANSYS has been performed a finite element
analysis for the shaft.


Figure 4. 3D shaft geometry


The shaft model has been meshed using tetrahedral elements type 3D
Solid 187, defined by 10 nodes having three degrees of freedom at each
node: translations in the nodal x, y, and z directions. The element has
plasticity, hyperelasticity, creep, stress stiffening, large deflection, and large
strain capabilities. In radial bearing close the electrical motor, where the
shaft has an important stress concentrator, a fine mesh was performed, using
2 mm elements size. For this type of mesh the finite elements number was


69
215045 and 361108 nodes. A detailed stress concentrator zone with mesh
refinement is presented in Figure 5.


Figure 5. A detailed stress concentrator zone with mesh refinement

The boundary conditions have been made according to real loading form
technical data sheets. A torsion loading was imposed from electrical motor
which is transmitted to pump impeller located on the shaft middle, marked
M
t
torque. Another condition was the impeller weight marked with G
r
and
bearings in shaft zones indicate in Figure 6.


Figure 6. The problem set-up

A convergence study for maximum equivalent stress was performed. For
the fine mesh with 428468 elements and 719492 nodes the solutions
difference are 0.43% compared with the solution with lower elements and
nodes number. In Table 4 are presented the results for 7 meshes with
different number of elements and nodes, respectively.

Table 4. Solution convergence study
Simulation
case number
Maximum equivalent
stress [MPa]
Elements
number
Nodes
number
Error
[%]
1 487.79 215045 361108
2 487.42 224145 376389 1.46
N. Pasca, L.Marsavina, S. Muntean, R. Negru

3 487.32 265452 445752 1.03
4 487.29 304121 510686 0.79
5 487.24 352145 591329 1.11
6 487.22 398456 669096 0.72
7 487.16 428468 719492 0.43

The stress and strain distributions on the pump shaft were obtained using
finite element method. As a result, the shear stress (
xy
,
xz
,
yz
), maximum
equivalent stress and deformation are presented in figures 7, 8 and 9,
respectively.
Results obtained and the fracture surface aspect indicate in Figure 2,
show that the fracture occurred in bearing zone after application of shear
loading where there is a stress concentrator. The stress results show a
maximum in that zone, the maximum equivalent stress as the normal and
tangential components is 483.97 MPa. The shear stresses occurring in the
concentrator are
xy
=131.45 MPa,
yz
= 278.78 MPa while
xz
is negligible.



Figure 7. Equivalent stress results on the pump shaft



Figure 8. Shear stress results: general view and zoom of the region with
maximum value (A region)



71


Figure 9. Deformation distribution on the shaft

4. FAILURE ANALYSIS DIAGRAM (FAD)
The failure (or fracture) assessment diagram is an attempt at combining
failure criteria based on linear elastic fracture mechanics and plastic
collapse. This failure criterion is a function of two parameters K
r
and S
r
,
which respectively quantify the elastic fracture component and the plastic
collapse component. K
r
and Sr for failure mode III were calculated using eq.
(1). Curve that represents a limit of safe design is plotted in the coordinate
system (Kr, Sr). The safe zone is considerate, the zone below the curve
provided Kr and Sr in FAD diagram.
IIIc
K
III
K
r
K = and
c
r
S

= (1)
where K
IIIc
[MPa m
0.5
] is the material fracture toughness in mode III [3] and
K
III
[MPa m
0.5
] represents the stress intensity factor in mode III at different
crack length calculated with eq. (2):
] m MPa )[ b / a (
III
F c
III
K = and (2)
] MPa [
) c D (
t
M
2
3
2 -
2

(3)
where F
III
(a/b) is the factor which consider the b=D shaft diameter 170 mm
and a uncracked zone radius [8], c crack depth for circumferential crack,
shear stress eq. (3), M
t
torque moment,
c
yield shear stress
y
.
c
577 0 =

[9].
The failure criterion for mode III is presented by 3 models function of
plastic zone shape near the crack: Hult-McClintock, Cherepanov and Bilby-
Cottrell-Swinden.
2 / 1 2
r r
)]} A B 1 ( ) S ( [ 1 { K + - = (4)
where B/A is ratio of the semi minor axis to the semi major axis of the
elliptic plastic zone.
Eq. (4) is a generalization of this failure criterion, for a customization we
have different type of plastic zone near the crack (circle when B/A=0,
elliptic when B/A=1). In case of B/A=1, result Hult-McClintock failure
N. Pasca, L.Marsavina, S. Muntean, R. Negru

criterion for mode III. For B/A=0, eq. (3) results Cherepanov failure
criterion. Another failure criterion for mode III, based on the Bilby-Cottrell-
Swinden customization has the same form as Dugdale model as eq. (5) [10]:
2 / 1
r
2
r r
)] 2 S sec( ln ) 8 [( S K = (5)
Based on these failures criterion and the concept presented above were
carrying out different shaft circumferential crack depth loaded in mode III.
The failure assessment diagram represented in K
r
and S
r
coordinates and
obtained from eq. (1) is presented in Figure 10.
From the failure assessment diagram is observed that for a circumferential
crack depth in shaft loaded in mode III, for all 3 failures solution critical
depth is 33 mm, after this circumferential crack depth the shaft is in unsafe
zone.


Figure 10. Failure assessment diagram in mode III

For crack depth between 33 mm and 35 mm after Bilby-Cottrell-Swinden
and Hult-McClintock failure criterion the shaft is in a safe zone.



73
5. CONCLUSIONS
The paper investigates the failure of the shaft from double suction storage
pump using numerical analysis and FAD diagram. The numerical analysis
has shown a stress field as well as the deformation distribution. A stress
concentration near to the lubrication channel of the bearing is highlighted by
shear stress and equivalent stress distribution. The fracture zone aspect
indicates a mode III loading produced by torque in order to pass the power
from electrical motor to the pump impeller.
The FAD diagram for mode III was developed taking into account three
different plastic zone shapes and was applied for circumferential crack. As a
result, the three models predict different critical crack depths. However, the
value of the maximum crack depth of 33 mm is could be considered the
maximum allowable crack depth. The inspection periods could be estimated
based on the previous data in order to prevent a catastrophic accident.

ACKNOWLEDGEMENT
This work was partially supported by the strategic grant POSDRU
ID77265 (2010), co-financed by the European Social Fund Investing in
People, within the Sectoral Operational Programme Human Resources
Development 2007-2013. Dr. Sebastian Muntean was supported by
Romanian Academy program.


REFERENCES

1. Rusu O., Teodorescu M., Lascu-Simion N., Material fatigue. Calculus
guide, 1, Ed. Tehnic, Bucharest, 1992. (in Romanian)
2. Bordeasu I., Popoviciu M., Marsavina L.,Voda M., Negru R., Pirvulescu
L., Numerical simulation of fatigue cracks initiation and propagation for
horizontal axial turbines shafts, Annals of DAAAM & Proceedings, 407- 408,
2009.
3. Das G., Sinha

A.N., Mishra S.K., Bhattacharya D., Failure analysis of
counter shafts of a centrifugal pump, Eng. Fail. Anal., 6:267-276, 1999.
4. Berndt F., van Bennekom A., Pump shaft failures - a compendium of case
studies, Eng. Fail. Anal., 8:135-144, 2001.
5. *** Alloy steels and high alloy steels for machine building, STAS 791-
66, 1966 (in Romanian)
6. Davis J.R. (Editor), ASM Specialty Handbook - Carbon and Alloy Steels,
ASM International, Metals Park, OH, 1996.
N. Pasca, L.Marsavina, S. Muntean, R. Negru

7. Liu Shu, Yuh J. Chao, Xiankui Zhu, Tensile-shear transition in mixed
mode I/III fracture, Int. J. of Solids and Struct., 41:61476172, 2004.
8. Khoo S. W., Karuppanan S., Stress intensity factor for cracks emanating
from a shaft, J. of Appl. Science, 11(10):1839-1844, 2011.
9. Susmel L., Multiaxial notch fatigue, CRC Press, USA, 2009.
10. Unger D., Analytical fracture mechanics, Academic Press, USA, 1995.

75


RECOVERING OF THE MECHANICAL
PROPERTIES OF A PEHD PLATE CONTAINING
A NOTCH BY THE GRINDING TECHNIQUE

D. Sersab, M. Aberkane

Laboratoire d'Elaboration de Caractrisation des Matriaux et Modlisation LEC2M.
Mouloud Mammeri University of Tizi-Ouzou. PO BOX 17 RP. Tizi-Ouzou 15000
Email address: sersabdjamel@yahoo.fr,
meziane_aberkane@yahoo.fr


Abstract : In this work, we present a repair approach of HDPE pipes. We use for
this goal, a single Edge notched tension (SENT) samples with different notches
lengths. We observe, in the first part, that the shapes of the force displacement curve
change when the specimen contains a notch comparing to the tensile test of the
plates without a notch and in the second part that the area value of the specimens
decreases with the increase of the notch length. This area under the force-
displacement curve represent the material deformation potential and the decrease of
this energy indicate that the material loss his deformation potential in the presence
of a notch. The aim of our contribution is to propose a rehabilitation method of the
material deformation potential based on the mater removal around the notch by the
grinding technique. We show that the recovering of the deformation potential of the
material is possible by using this method. We, also, show that the mechanical
properties of the material are better when the grinding well is bigger. To finish, we
propose a recovering efficacy criterion based on the determination of the optimal
value of W a (where a is the notch length and W is the width of the sample), at
eight percent (80%) of mechanical properties recovering, beyond witch the
rehabilitation is ineffective.

Key words: HDPE, repair, notch, tension, recovering.







D. Sersab, M. Aberkane

1. INTRODUCTION

Transport pipe is an enclosure under pressure transporting of the fluids at
long distances. It is known as an adduction when it is intended for the
transport of the large flows. It must resist to the internal pressures of the
transported fluid and with the constraints of the ground in which it is
hidden. The transport pipes are manufactured with concrete, cement
asbestos, cast iron, steel, PVC or HDPE. The canalization safety is one of
the public interest and a question of the first importance. The number of
fractures per year is a criterion of the pipe reliability. The presence of the
defects in a pipe is unavoidable. These defects can be internal such as
microscopic cracks, pores, the brittle particle's inclusions, or external due to
the presence of notch (macro cracks) resulting from the process of
construction, a corrosion or a damage caused by thirds . When the state of
the defect is not sufficiently serious to lead to a complete replacement of
pipe, and if the conditions of flow remain satisfactory, the rehabilitation of
the pipes is justified. The state of the canalization, the possibility of action
and the cost of a probably reparation are the element which are considered
in the choice of the rehabilitation technique:

2. ANALYZE EXISTING DEFECTS IN A PIPE

The analysis of the defects met in a pipe, enables us to use the following
working hypotheses:

1. The defects, whatever their natures, are a geometrical
discontinuities being able to be described simply by a notches
which can be characterized by three parameters, the length a, the
ray and the angle of notch .
2. Under solicitation, the notches lead to a local stress concentration.
In the case of HDPE pipes, the defects are induced by stress
concentrations generated by the presence of external or interior defects
(scratches, stripes, scratches and inclusions) caused during the process of
extrusion or at the time of the implementation of the canalization by wrong
movement. For example, a wrong movement during the handling which can
cause a geometrical discontinuity at least visible on the external surface of
the pipe [1 ; 2]. These stress concentrations generate, according to
conditions' of temperature, loading, depth of notch and environment various
types of propagation of crack, such as the slow propagation of crack
(SCG:slow fox trot ace growth).This type of progression, which results from
the creep or the loading of tiredness, is wished the least because it does not
show any sign before the rupture of the tube [ 3 ].It can be ductile or
fragile.Lu and Brown [ 3 ] suggested that the two processes occur
simultaneously;and the rupture final depends on which process is faster
under the loading, the temperature and the depth given of notch. To
evaluate the behavior of the PEHD under the quoted conditions known
77

several work is completed [ 3]-[4 ] and [ 5 ]; according to American Water
Works Association the scratches which have a depth of 10 % thickness of
wall of the tube in PEHD are tolerable [ 5 ] and above these 10% the
product should not normally be used [ 6 ]. To give in service the tubes in
PEHD presenting of the defects the PPI (Plastics Institute Pipe) published a
technical note describing the various methods of repair according to
characteristics' of the defect [ 7 ].
In this presented work we use the plates instead the pipe for obvious
practical reasons. We chose the sample shape with Single Edge Notched
Tension (SENT) because of the test simplicity and the possibilities that this
geometry permit us in terms of repair by grinding or removal of matter.

3. PROPOSAL FOR TECHNICS OF REPAIR OF THE
HDPE PIPE

A. Grinding reparation
The problem of stress concentration being related to the notch effect, we
propose in our study a solution based on the modification of the shape of the
notch. This solution consists on the reduction of the strong stress
concentrations by the removal of adjacent matter at the least forced zones.





Figure1. Removal matter technics schematization [8]

This geometry treatment improves the profile of the defect by
eliminating the existing cracks and by reducing the related stress
concentration.
4. EXPERIMENTAL METHOD
The samples used are rectangular (see figure 2), They are tested in
uniaxial traction for the law of behavior of the virgin sample (1), the
notched ones (2) and the repaired ones by the grinding technics (3). These
samples are taken on HDPE pipe, with external diameter of 600 mm and
thickness of 30 mm in the circumferential direction (see figure 3), The notch
is practiced in the direction of extrusion, Sample's dimensions are in
conformity with standard ESIS [9] which stipulates that in the goal to have a
D. Sersab, M. Aberkane

planes stress in the ligament, it is necessary to observe the double following
condition:

3
3
W
l b (1)

Figure2. sample used for the tensile test ;(1) : virgin sample;
(2) : notched sample; (3) : sample repaired by grinding



Figure 3. Sampling direction



Figure 4. SENT Sample [4]


79

Table 1: Samples Dimensions
N
samples
Lengt
h l(mm)
Wid
th
W(mm)
Thickne
ss b(mm)
Leng
th of
notch
a (mm)
1
150 45 6
5
2 10
3 15
4 20
5 25

The tests are carried out on a IBERTEST tensile testing machine of +/-
100 kN load capacity. The forces and the displacements are measured by a
load cell integrated into the machine and by a grip sensor directly posed on
the sample respectively. The load-displacement curves are then collected
and integrated by software into the machine in the form of Excel files.

5. RESULTS AND DISCUSSION
5.1. NOTCH'S INFLUENCE ON THE MATERIAL TENSILE
BEHAVIOR
The notches carried out on the samples plainly change the material tensile
behavior. We note that in the case of the virgin sample (see figure 5) we
find the known behavior of the HDPE characterized by the succession of
three stages which are:
1) The elastic phase with the thresholds of flow low and high;
2) The phase of flow elastic or viscoelastic;
3) The phase of hardening of material before its break off
In the case of the notched samples the viscoelastic phase disappears (see
figure 5) and the elastic phase followed immediately the material break off.
This indicating a great material embrittlement. This fact induces, also, a loss
of a material deformation potential proportionally to the notch length.










Figure5. Curve load displacement of the samples not notched (1) (black
curve)and notched (2)(color curves)

D. Sersab, M. Aberkane

5.2. PROPOSITION OF THE RECOVERING
TECHNIQUE OF THE ELASTIC FLOW
POTENTIAL
The material embrittlement observed previously is a problem and a
proposed solution to avoid this situation is consisting on the total or
partially recovering of the loss energy by rehabilitation or a repair
technique. In our study we were inspired by the Eckstein [5] work who
proposed a rehabilitation solution based on the removal of matter thanks to
an operation of grinding of the cracked part (see figure 6).


Figure 6. principle of the grinding technique used

The principle of rehabilitation is very simple; it acts to grind the cracked
area thanks to a wheel. To optimize this repair we use several diameters of a
grinding wheel so as to be able to reach repair optimal. After the matter
removal, we carry out tensile tests on the repaired samples and we compare
the tensile diagram of the notched sample with that of the repaired samples
(see figures 7 and 8).


Figure 7. Curve load-displacement of the samples repaired with radius of 5, 10
and 30 mm for notchs depth a = 5 mm

81


Figure 8. Curve load-displacement of the samples repaired with radius of 25,
30 and 40 mm for a = 20 mm

We measure the material viscoelastic potential by the area under the
tensile curve which is considered as the total energy dissipated during the
notched sample fracture. We compare, then, this energy with the one
dissipated during the tearing of the virgin sample. The test results are
regrouped in a table which contain for each sample tested the notch length a,
the grinding ray R, the maximum loading Fmax and the value of the area
under tensile curve Wr.

Table2. Comparison of F
max
and W
R
before and after repair by
grinding
a
[mm
]
R
[mm
]


W
r
[J]


F
max
[KN]
Before
Reparation
after
Repara
tion
before
Repara
tion
After
Repara
tion

Without notch

1240 1240 5.78 5.78

5
5 235.12 5.22
10 183.37 274.78 5.18 5.46
30
1192.5
0
4.69

10
5 210 4.73
10 132.09 333.52 4.35 4.45
25 373.91 4.60

15
5
92.32
130.56
3.54
3.54
25 309.40 4.14

20
25 140.05 3.25
30 62.65 625 2.72 3.35
40 617 3.37
25 306.58 2.7
D. Sersab, M. Aberkane

25 30 39.36 580.57 2.03 2.58
40 1022.1 3.05
.
We notice in table 2, on the one hand, that each time repair is undertaken
the Wr energy dissipated during tearing increases proportionally with the
grinding ray, what indicates that grinding has a positive effect on the HDPE
potential of deformation, in addition, we notice that the value of Wr drops
proportionally with the length of the defect, These two observations, allow
us to propose an optimum of repair which would take into account these two
parameters.
1. REPAIR OPTIMIZATION
To account for the depth to which repair the notch could be effective we
standardize all the values of energy of repair of the samples presented in
table 1 at 60 mm of displacement, just at the necking end, we compare, then,
the energy dissipated under the traction curve for the sample without notch
(see figure 9) with the energy of the samples repaired, The results of these
measurements are gathered in table 3:










Figure 9. Measure standardized energy of the healthy sample

Table 3: Table gathering the values of standardized energies of repair,

a
[mm]
[mm]
L
[mm]
W
rn
[J] standardized to 60
mm of displacement [J]
0 0 0,00 267,84
5 30 33,17 207,41
10 10 20,00 200,11
10 25 40,00 210,13
15 25 45,83 182,77
20 30 56,57 143,45
20 40 69,28 150,89
25 25 50,00 122,83
25 40 74,16 131,60
83

To find the optimal value of repair, we plot a graph connecting the
fracture energies standardized to a geometrical parameter which is
report/ratio length of notch a over the ligament length W( a/W) as
represented on figure 10,
0 0.1 0.2 0.3 0.4 0.5 0.6
0
20
40
60
80
100
a/w
T
a
u
x

d
e

r
e
c
o
u
v
r
e
m
e
n
t

d
e

l
'

r
g
i
e

i
n
i
t
i
a
l
e

(
%
)

Figure 10. Recovery rate of energy dissipation as a function of aspect ratio
( W / a )

In Figure 10, we note that the fracture energy of the material
decreasesW
rn
=267,84J corresponding to 100% of the energy is dissipated in
the breakdown of the virgin sample W
rn
= 207 J representing about 80 % for
samples rehabilitated it then stabilizes between a ( a/W) ratio ranging from
0.1 to 0.22 and then begin to decline until a rehabilitation rate of 50% of the
energy dissipated. To optimize the repair we recommend a rehabilitation
rate of energy dissipation of 80% which corresponds to an optimum aspect
ratio a/W=0,22 beyond which the repair is not recommended and
replacement of the section notched is recommended.
6. CONCLUSION
This work allowed us to shed light on the vulnerable behavior of HDPE
in the presence of notches. It also allowed us to propose a solution for the
rehabilitation of this material by the technique of grinding. The tests helped
to highlight the influence of the grinding radius on the repair. It is shown
that the material rehabilitation is better when the grinding radius is the
largest. The close relationship between the ability to rehabilitate the material
expressed by its ability to recover its qualities of plastic deformation
expressed by its potential for dissipation of fracture energy normalized with
the geometric properties of the samples expressed as the ratio ( a/W) has
determined an optimal value of this parameter at which the grinding repair
method of HDPE is more efficient.
D. Sersab, M. Aberkane

REFERENCES
1. A. Benhamena , L. Aminallah , B. Bachir Bouiadjra, M.
Benguediab , A. Amrouche , N. Benseddiq J integral solution for
semi-elliptical surface crack in high density poly-ethylene pipe
under bending Materials and Design 32 (2011) pp.25612569.
2. R. Schouwenaars, V.H. Jacobo, E. Ramos, A. Ortiz Slow crack
growth and failure induced by manufacturing
defects in HDPE-tubes, Engineering Failure Analysis 14 (2007)
pp.11241134
3. J. Zhang, Experimental Study of Stress Cracking in High Density
Polyethylene Pipes, a Thesis Submitted to the Faculty of Drexel
University by in partial fulfillment of the requirements for the
degree of Doctor of Philosophy November 2005; pp.5-7.
4. D. Eckstein, Buried Plastic Pipe Technology, 2nd Volume,
ASTM Special Technical Publication 1222 (1994).
5. Thomas S. Walsh, Plastic Pipe and Fittings: Past, Present, and
Future, Journal of ASTM International, Selected Technical Papers
STP1528 (November, 2011).
6. Plastics Pipe Institute, Frequently Asked Questions; HDPE Pipe for
Water Distribution and Transmission Applications, Technical Note
27 (2009), pp 13; question 68.
7. Plastics Pipe Institute, General Guidelines Repairing Buried HDPE
Potable Water Pressure Pipes, Technical Note 35 (2006).
8. A. Belalia, M. Aberkane et Z.Azari La rhabilitation des
canalisations de transport endommages Rapport CMEP tassili.
Dcembre 2010.
9. M. Aberkane, M. Ould Ouali Fracture Characterization of ST37-2
Thin Metal Sheet with Experimental and Numerical Methods, Key
Engineering Materials. Trans Tech Publications inc. Volume 473,
2011, pp. 396-403.

85

SCENARIO OF FRACTURE DEVELOPMENT IN
BUCKET WHEEL EXCAVATOR
Darko Danii
1
, Stojan Sedmak
2
, Ivo Blai
3

1
Kolubara metal, Lazarevac, Serbia
2
University of Belgrade, Faculty of Technology and Metallurgy, Belgrade, Serbia
3
Military Technical Institute, Belgrade, Serbia
e-mail darko.danicic@kolubarametal.com

Abstract: Continued service of BWE in the severe working conditions may result
in machine failure. In case of important structural elements it might lead to a large-
scale breakdown. The allocation of structurally significant elements is of vital
importance for the lifecycle of BWE, starting from design and construction, through
operation and maintenance. The example of recent failure of one BWE caused by
complex fracture, consisting of fatigue and final fast fracture is considered. Possible
scenario indicated that fracture started from a small element due to defects
introduced by poor quality of welded joints which significance in structure was
underestimated in given operating condition and applied loading.

Key words: welding joint; stress distribution; structural integrity; maintenance







D. Danii, S. Sedmak, I. Blai
1. INTRODUCTION
In-service behavior of bucket wheel excavator (BWE) and other
equipment operating on surface mines (bucket chain excavators, spreaders,
belt wagons, stackers), aimed to supply electrical power plants by coal,
depends on the design, capacity, manufacturing quality, applied loading and
typical mining conditions (stability of mining area, strength of overburden
mass, allowable soil loading). BWE development and application include
design and manufacturing, acceptance for service after performed requested
testing, operation, in-service inspection and maintenance. Structural integrity
can be endangered in each of these steps, and this requires the decision to
continue the operation or undertake the repair of damaged components. In
spite of strictly obeyed prescribed rules and sequences, premature damages
and failures of surface mine equipment occur in service, causing significant
costs. Such failures are experienced also in opencasts surface coal mines in
Serbia, and one of them is selected for a detailed consideration. In addition
to direct costs, the losses due to downtimes caused by failure disturb
electricity production, and for that they are important.
Performed failure analysis revealed that wheel with buckets and boom are
most critical parts, requiring in some cases to improve the design /1-5/. One
important aspect of excavator design is fatigue and fracture behavior of
welded steel structures /6-8/. Welded joints, due to imperfections caused by
manufacturing and heterogeneous microstructure (parent metal - PM, weld
metal - WM, heat-affected-zone HAZ) are most critical parts regarding
crack initiation and growth, requiring special attention when exposed to
variable loading and fatigue, as it is the case on open surface mines.
The service problems of equipment operating on surface mines attracted
the attention and in many papers they had been considered /9-12/.

2. FAILURE OF A BUCKET WHEEL EXCAVATOR
To avoid unexpected failures of bucket wheel excavators (BWE) and
save their structural integrity in service, necessary care during operational
life, monitoring and diagnostics of all vital elements of supporting structure,
sometimes also repairing and redesign, are required.
BWE SchRs 1760, unexpectedly and with no warning catastrophically
failed in 2004, after 17 years of regular service, on an open surface mine in
Serbia (Fig. 1). Fatigue fracture, initiated in welded joint and developed in
lugs of counterweight holder (Figs. 2, 3), followed by final fast fracture, had
been the cause. Cracks initiated in the sites of stress concentration and
inhomogeneous microstructure of welded joints, primarily in HAZ, under the
effect of external loads and residual stresses /13/.
87


Figure 1. Collapse of bucket wheel excavator SRs 1760

The fracture surfaces of left lug, pos. 68 (Fig. 3.a) and right lug, pos. 62
(Fig. 3.b) are substantially different. Right lug fractured in brittle manner,
due to an overloading. Brute fracture in left lug took place when the loaded
cross section area was significantly reduced after extended fatigue cracks on
both sides of welded rib (Fig. 3). Flat fatigue crack growth in pos. 68 had
been interrupted by stable crack growths, with visible shear lips.


Figure 2. Fracture of two lugs on counterweight holder

a.

b.
Figure 3.a. Lug (pos. 68) fracture (fatigue, stable crack growth and brute fracture)
b. Brittle fracture of lug (pos. 62), indicating single initiation point from both sites

pos. 68
pos. 62
D. Danii, S. Sedmak, I. Blai
2.1. Analysis of applied load and stress distribution

The loading acting on the bucket wheel is stochastic. It was limited
across the load cases given in the standard DIN 22261-2 /14/; sizing of
individual components is done based on this standard. Based on the static
load and center of gravity position, it can be seen that the anchor rope is
exposed to the maximum load when the BWE is ready to work (load case
in standard) with a horizontal boom. At the time of excavation, in any floor,
the values of digging forces reduce, while the coupling slipped. Variable
load and stress will be ranged between maximum and minimum values, i.e.
between static load and minimum value, corresponding to maximum applied
digging load in each digging contact between the bucket and the ground.
This is confirmed by the strain alteration, recorded by strain gages (Fig. 4)
/15/ in regular operating condition of BWE. When applied digging force
exceeds nominal value for 50%, drive would be switched off automatically.
Analyzing the lugs in Fig. 2 and design scheme (Fig. 5.a) a symmetrical
distribution of load and stress in two lugs could be assumed as reasonable.
This was not the case, as it is possible to conclude analyzing the fracture
surfaces on lugs 62 and 68 (Fig. 3).
Two effects could contribute to induced asymmetry. First one is an error
in the design; the second one is low quality of welded joint, which allowed
high level of stress concentration, induced by welding imperfection and
defects. In this aspect it is necessary to explain (1) how cracks initiated and
transferred from welded joint to lugs parent metal and (2) how variable
loading and stress did not affect the behavior of lug, pos. 62 (Fig. 3.b).


0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00
Time [Sec.]
-1.0

-0.5
0.0

0.5

1.0


Figure 4. Strains recorded in typical locations of BWE welded structure in operation
89





Figure 5.a. Design scheme of anchor ropes. b. Presentation of connection of the
main rope and a structure of the counterweight

The design details of welded joint between rib (pos. 60) and the lug (pos.
68) are shown in Fig. 6. The same design is used for other lug, pos. 62.
In the case of BWE SchRs 1760, originally applied design of connection
of the main rope to counterweight structure had more than one unfavorable
solution. Double lugs design is used (Fig. 5) and the load distribution in
structure is non symmetric regarding direction of the load. From Fig. 5.b it is
clear that the tension force is transferred to the positions 62 and 68 through
the ribs, pos. 60. In this way the distribution of load to the lugs is
unbalanced. Ribs, although transferred the load, were treated as auxiliary
elements and strict requirement for welding inspection was not prescribed.
Two facts have to be noticed. At first, the fatigue crack started from
welded joint between the positions 68 and 60 (Fig. 7 left), on the defect
location 1, produced by welding. Variable loading had been transferred as
fatigue loading completely to position 68 (Fig. 3.a). On the other side, on the
position 62, although similar defects had been detected (Fig.7 right), fatigue
crack did not developed (Fig. 3.b). The explanation can be found in the
design of the lugs: position 62 is fixed to the main structure (Fig. 5.b) while
position 68 is free to swing.
pos. 62
pos. 68

pos.60
D. Danii, S. Sedmak, I. Blai

Figure 6. Design scheme of welded joint between lug (pos. 68) and rib (pos. 60)



Figure 7. Welded joints between rib and lugs (pos. 68, left, pos. 62 right)

Having in mind long time period for fracture, it is interesting to assess, at
least roughly, the time consumed for different steps. For that, the data
recorded for BWE till its collapse is taken from the diary of work. They have
shown total of 60 600 working hours with following data:
The frequency of load change 58.2/60 = 0.97 sec.
The total number of load changes 60.600x3600/0.97= 2.25x10
8

The nominal circumferential force 373.8 kN
The switch off force 560.7 kN
Maximum calculated force in anchor rope 3874 kN
pos. 60
metal seam
pos. 68
cross section surfice
view 1
view 2

91

2.2. Quality of welded joints
The design details (Fig. 6) indicate that fillet welds applied between rib
(pos. 60) and the lug (pos. 68) had been performed without penetration,
since the joint had been considered as an auxiliary one. These welded joints
had not been manufactured according to welding procedure specification
(WPS) and were not inspected properly. Heterogeneous microstructure
(Fig. 8), and presence of defects and imperfections, (Fig. 9), discovered by
investigation after collapse, are typical.
Figures 7 to 9 allow to conclude that for both lugs the welding quality is
poor, causing eventually the high level of stress concentration, similar in
joints of both lugs, pos. 68 and 62. to the ribs.




Figure 8. Welded joint section,
normal to fracture of lug, pos.
68
Figure 9. Crack in the fillet weld root (up);
incomplete root penetration (down)
D. Danii, S. Sedmak, I. Blai
3. PROBABLE SCENARIO OF FRACTURE DEVELOPMENT

3.1. Crack initiation in welded joint
In BWE welded structure with defects (Fig. 7 and Fig. 9), and
inevitably heterogeneous microstructure (Fig. 8), exposed to load
characterized by maximum initial static value and operational variable
load, crack might be locally initiated by brittle fracture in microstructural
region of low fracture toughness, by combined low cycle fatigue and crack
tip blunting followed by stable crack growth in the more ductile region or
by high cycle fatigue due to variable loading. Since the change in
microstructure is sharp, crack initiation can include in both lugs all three,
brittle, ductile and fatigue cracks, Fig. 7. The stress concentration,
constraint and residual stresses, in addition to the variation of mechanical
properties due microstructural heterogeneity can govern the mode of crack
growth. Interrelation between influencing factor is decisive for the
dominant type of a crack, but the fatigue crack will be surely present at the
transition of welded joint to a lug, as can be concluded from Fig. 10.
Time period for crack initiation and propagation through weld metals in
two runs (about 3,6 and 3,8 mm, Fig.) up to the HAZ of lug (2,2 mm) is
difficult to evaluate because this process is complex.

Figure 10. Initiation of dominant fatigue cracks in lug, pos. 68: details in locations
A and B, LZ- ratchet marks, LO- beach marks, critical region C

93

3.2. Fatigue crack transfer to a lug

For the scenario of fracture in Fig. 3.a more interesting is development
of fatigue fracture in lug, pos. 68. More details about fatigue crack
transition to the lug, pos. 68, are visible in Fig. 10. They might be used in
explanation of fatigue crack initiation in this lug. Variable loading exerted
its effect only to pos. 68, and not in pos. 62, what is attributed to fixation
of the later to main structure.
Ratchet marks, beach marks and striations, as typical for fatigue, can be
recognized on fracture surface (Figs. 7 and 10), which entered in lug as
two separated fatigue cracks 1 in an early step and merged in a dominant
fatigue crack 2 (Fig. 11). It is interesting to note that cracks 1 started in the
HAZ of low fracture toughness in lug material, and crack 2 developed
through the material of more homogeneous microstructure of parent metal.

Figure 11. Merged early fatigue cracks 1, points A and B, in a dominant crack 2

3.3. Advancement of fatigue crack

Continuous fatigue crack, above 300 mm long (Fig. 3.a), developed
through homogenous microstructure, uniformly for every sequence of
variable loading. In this segment of fracture Paris law can be applied for
determination of time spent for fatigue crack growth.
Increasing fatigue crack continually reduced bearing cross section area
of the lug, pos.68, to a size critical for fast fracture It is reasonably to
assume that fatigue crack grew uniformly from crack 2, according to
applied load, Figs. 3.a and 10, on both sides of welded joint, Fig. 7 left.
Applied variable stress increased by increasing fatigue crack, due to
continuously reduced bearing cross section area. This allowed fracture
mode transformation from high to low cycle fatigue first, as shown in
region IP and in magnified view in Fig. 12.
D. Danii, S. Sedmak, I. Blai
3.4. Stable crack growth by ductile fracture

Three regions of stable crack growth by tearing fracture had been
registered in Fig. 3.a. They occurred when applied maximum stress in
analyzed component locally exceeded yield stress of material, the energy
had been consumed for tearing fracture and crack tip blunting, followed by
shear lips and development of stretch zone, according to maximum shear
stress, under 45 to acting force. But among these three regions there is the
difference in starting condition. In the case on the left side (Fig. 3a) high
cycle fatigue preceded stable crack growth, which had been arrested by final
stretch zone (FSZ) in the region ZP. At the end of FSZ final fast fracture
took place. On the right side, in the region IP (Fig. 12) low cycle fatigue
preceded the tearing segment CA, which had been arrested by FSZ (Fig. 13).
After this new FSZ, crack had to continue to grow, again by tearing fracture,
up to the formation of FSZ before final fast fracture.
It is to notice that in all three cases plane stress dominated and shear lips
are present. However, the sequence of occurrence individual tearing fracture
regions is not clear and request further analysis.


Figure 12. Transition from flat fatigue
fracture surface, with visible beach marks
and less visible striation, to stable tearing,
region IP in Fig. 3
Figure13. Final stretch zone after
developed shear lips prior the next
stable growth of ductile crack,
corresponding to region CA in Fig. 3

95

After stress redistribution, fatigue crack continued to grow continuously,
by low cycle process beyond final stretch at CA and by high cycle process
on the other side, up to simultaneous transition to tearing mode on both
sides. Fracture process ended by fast fracture on both ends, after probably
simultaneous development of final stretch zones (FSZ).
The region of transition from fatigue into static fracture is presented in
Fig. 14. In addition to beach marks, static splitting fracture had been detected
in the fatigue process, in crack tip blunting region, where structure with low
cohesive strength are present along the plate thickness. The presence of
static fractures on the fatigue fracture surface indicates the faster rate of
propagation of the fatigue crack.
The presence of chevron patterns in the static fracture portion indicates
the initiation of unstable growth, in the middle of plate thickness where
plane strain condition dominated. Partial crack arrest (Fig. 15), along with
unstable fracture, indicated that the fracture occurred above nil-ductility
transition temperature (NDTT), in the brittle range. According to Pellini /16/,
a material can arrest the crack only if the applied stress is lower than R
p0.2
.
In this case, this is 185 MPa, well bellow allowed operating stress.


Figure 14. Transition from fatigue to
static fracture, PNL- final stretch zone
Figure15. Crack arrest zone, ZP crack
arrest

3.5. Final fast fracture of two lugs

For final fracture it was necessary that applied stress exceed ultimate
tensile stress of material. The process of cross section area reduction took
place during fatigue, but in fatigue stage only cross section of lug, pos. 68,
had been affected, and not of pos. 62. During stable crack growth, which
D. Danii, S. Sedmak, I. Blai
preceded the final fracture, the reduction of cross section continued, still
only of pos. 68. When total residual area of bearing cross section in both
lugs had been reduced to a minimal value instantaneous fast fracture
occurred, but by tearing mechanism. For the balance of applied load between
different residual cross sections of two lugs, the load had to be redistributed,
that means pos. 62 accepted greater part of the load than pos. 68.
This process had ended by development of first local final stretch zone
CA. Simultaneously fatigue process might continue on lug other side up to
the crack arrest and second final stretch zone ZP produced. When the
stresses on both sides were balanced after strain energy had been consumed
by final stretch zones development, last ductile stable crack had grown
beyond CA location. This step finished by reached third final stretch zone. In
this very moment bearing total cross section area of both lugs, pos. 68 and
pos. 62, was reduced to the value critical for instant fast fracture.
The examination of the fracture surface of the position 62 showed the
chevron pattern indicating that the fracture started in the weld metal, in the
area between the two passes, Fig. 3.10. Generally, the occurrence of the
unstable crack growth is typical for all the constituents of the welded joint
and the parent material. Since no cracks have been arrested during the
fracture of the position 62, one can say that, based on Pellinis analysis /16/,
the stress applied in the cross section was above R
p0.2
. This fracture has
another typical characteristic: there are no shear lips on the fracture surface.

4. DISCUSSION
From theoretical point of view it is interesting to analyze the levels of
applied stress and material response in different stages of considered fracture
and the rates of crack growth, going back from final separation. This might
help to better understanding of fracture mechanism and process duration.
Material will be separated if the applied stress exceeds ultimate tensile
stress (UTS). However, it can happen at global level, for average value of
applied stress, but also it can be sufficient if it happened locally, where stress
concentration increases the value of average stress well above UTS. For
tearing fracture the applied stress has to be in average higher than yield
strength (YS) of material, accounting with stress concentration again. If the
maximum value of applied variable load is slightly lower than YS, low cycle
fatigue is most probable mode of cracking. The value of stress for high cycle
fatigue should be lower than YS.
Rate of crack advancement also corresponds to fracture mode. For
unstable fracture the rate is very high, the fracture occurred in very short
time period, almost instantaneously. The rate of stable crack growth by
tearing is lower, and for final separation, after final strength zone formation,
97

determined time is necessary. Low cycle fracture occurs in the range of cycle
of variable loading (e.g. 50 000 to 100 000 cycles. For high cycle fatigue
long time is required for crack development, measured by more than one
million cycles.
It is also important to consider local microstructure for the evaluation of
time necessary for individual steps of fracture.
The more detailed analysis of applied stress level, crack growth rate and
microstructure investigation of welded joint is necessary for the evaluation
of time period required for individual steps in presented fracture.
Roughly, of 17 years of BWE operation, it might be proposed that 70%
of time had to be spent for high cycle fatigue, and the rest for crack initiation
and early development in weld metal.
In the case of BWE fracture more interesting is practical engineering
aspect of considered fracture, directed to the possibility how to avoid crack
occurrence and how to prevent so difficult collapse and its consequences.
Fracture initiated from cracks in critical welded joint. The problem would
not appear if this welded joint is not introduced in a structure of the
counterweight. It is likely that, after experience with BWE failure, the design
solution without welded joint of that kind is possible.
Next question is if the fatal collapse could be prevented by proper action
during BWE design and exploitation.
By introduction of ISO 9000 series standards for quality assurance and
ISO 9001 and ISO 9002 for quality systems welding is defined as "special
process" because welded joints can not be fully inspected according to the
standard requirements for a complete verification. In the case of welding, the
quality can not be verified on the product but has to be built-in in the
product. This generally accepted approach is dictated by the nature of fabri-
cation in welding. Anyhow, the quality of welded joint can be endangered:
(1) by defects induced during manufacturing and service and (2) by
inevitable heterogeneity in microstructure and mechanical properties, corres-
ponding to the nature of welding process, consisting of applied heating -
cooling cycles.
In design stage most important was the underestimated significance of
this welded joint, classifying it as auxiliary, and quality assurance system
was not applied, neither in fabrication, nor in inspection. As a consequence,
many detected defects and imperfections induced during manufacturing of
welded joint contributed to high stress concentration, and critical welded
joint was not available for inspection during service. In addition, misbalance
of load and reduction of lug thickness from 40 mm to 20 mm by design
solution produced inconvenient stress distribution. So, the structure
contained the cracks, developed in fabrication or during operation under
variable loading. For sure they could be detected in a proper inspection
system, but it was not provided. Time period of 17 years of operation
D. Danii, S. Sedmak, I. Blai
indicates that in corresponding inspection and maintenance system the
failure might be prevented.

5. CONCLUSION
Considering the importance of safe operation of BWE and other
equipment operating on surface mines, complex systems for safety have
been established, mainly as governmental agencies. It includes also welding
procedures.
This approach is the direct contribution of new inspection and
maintenance approach, especially proactive maintenance /17/, because the
time estimates based on the crack can set a time frame in which the cracks
can be detected and repaired to avoid a collapse of the structure. Also, during
the inspection may be extended, reducing maintenance costs. Concerning
this approach set a time in which critical elements of the structure should be
revised so as to prevent damage, applying correct non-destructive
examination. It means that corrected time of steel structure inspection can be
applied basing on this kind of estimates and make more efficient inspection
and maintenance.
Given the levels of assessment, definitely oblige to do the experimental
confirmation of the hypothesis that would be a continuation of this work.

6. REFERENCES
[1] M. Arsi Korelacija zamorne vrstoe i praga zamora zavarenih
spojeva (Correlation between endurance and fatigue threshold of welded
joints), Ph. D. thesis, Pritina, 1995.
[2] M. Arsi, S. Sedmak, M. Sarvan:Analiza uzroka pojave prslina na
zavarenim spojevima konstrukcije rotora bagera SRs 1300.26/5.0,
International conference Welding 96, Beograd, 1996.
[3] D. Danii, T. Maneski, D. Ignjatovic: Diagnostic approach to steel
structure maintenance to prevent mining machine fractures, New Trends in
Fatigue and Fracture, Metz 2010.
[4] D. Danii, T. Maneski, D. Ignjatovi: Structural Diagnostics and
Behaviour of Bucket Wheel Excavator, IVK 1-2010,53-60.
[5] D. Danii, T. Maneski: Structure Failure of the Discharge boom of
BWE C 700 S due to Dynamic Effects, IVK 1-2012, 43-46
[6] M. Berkovi, Numerical Methods in Fracture Mechanics, IVK 2-2004,
63-66
[7] M. Berkovi, S. Maksimovi, A. Sedmak, Analysis of Welded Joints by
Applying the Finite Element Method, IVK 2-2004, 75-84
99

[8] T. Maneski, D. Ignjatovi, Dijagnostika vrstoe konstrukcije - Structural
Performance Diagnostics, IVK 1-2004, 3-8
[9] T. Maneski, D. Ignjatovi, Sanacije i rekonstrukcije rotornih bagera -
Repair and Reconstruction of Bucket Wheel Excavators IVK 1-2004,9-28
[10] T. Maneski, D. Ignjatovi, Sanacije i rekonstrukcije transportera i
odlagaa - Repair and Reconstruction of Belt Wagons and Stackers, IVK 1-
2004, 29-38.
[11] M. Arsi, S. Sedmak: Spektar optereenja za zamorna ispitivanja
zavarenih spojeva rotornih bagera, Meunarodno savetovanje Zavarivanje
94, Novi Sad, 1994.
[12] D. Danii: Diagnosis of Conditions and Behavior for Steel Structures
of Mining Machinery, PhD Thesis, Belgrade 2010.
[13] S. Sedmak, V. Grabulov, D. Momilovi, Chronology of lost structural
integrity initiated from manufacturing defects in welded structures, IVK 1-
2009, 39-50
[14] Standard DIN 22 261-2:2006-12, Bagger, Absetzer und Zusatzgerte in
Braunkohlentagebauen - Teil 2: Berechnungsgrundlagen
[15] N. Zuber, H. Lien, A. Klanja-Milievi: Applied Remote Condition
Monitoring of the Bucket Wheel Excavator, Istraivanja i projektovanja za
privredu, Nauno-struni asopis, Beograd 2009.
[16] W.S. Pellini, Guidelines for fracture-safe and fatigue-reliable design of
steel structures, The Welding Institute, Abington, Cambridge, 1983.
[17] P. Jovani, D. Ignjatovi: Proactive monitoring system for main
mining mechanization at open cast mines, IVK 1-2010, 11-20.



101


ASSESSMENT OF THE BEHAVIOUR OF FATIGUE
LOADED HSLA WELDED STEEL JOINT BY
APPLYING FRACTURE MECHANICS
PARAMETERS

Ljubica Milovi
1
, Sran Bulatovi
2
, Zoran Radakovi
3
, Vujadin Aleksi
4
,
Stojan Sedmak
5


1
University of Belgrade, Faculty of Technology and Metallurgy, Serbia
2
Yugoslav River Shipping, Belgrade, Serbia
3
University of Belgrade, Faculty of Mechanical Engineering, Serbia
4
Institute for Testing of Materials, Belgrade, Serbia
5
Society for Structural Integrity and Life (DIVK), Belgrade, Serbia

Email address: acibulj@tmf.bg.ac.rs


Abstract: In discussion of the welded joints from fracture mechanics point of
view, it is assumed that the welded joint is pre-cracked and that therefore fatigue life
of the pre-cracked structure is determined by the period of crack growth under
variable loading. Experimental data obtained by testing provide a substantial
foundation for better understanding and explanation of the phenomenon of fatigue of
a material. Low-cycle fatigue (LCF) occurs during charging and discharging of the
reactors, pressure vessels and pipelines; it can be accelerated by additional negative
effect of temperature variation and aggressive effect of vessel contents during
exploitation of the equipment in processing industry.
In present paper, the results of measurement of J-integral of weld metal made of
low-alloy high-strength steel, used for manufacture of the pressure vessels in the
process of powder welding, at LCF have been presented.

Keywords: LCF, J-integral, HSLA steel, process equipment




L. Milovi
1
, S. Bulatovi, Z. Radakovi, V. Aleksi, S. Sedmak

1. INTRODUCTION
Material fatigue can be clarified to a large extent using the results
obtained by experimental examination, particularly so when one should
understand the behaviour of a crack in a material with heterogeneous
structure such as welded joint. Thus, we should conduct fatigue testing of
fracture mechanics of the specimens with notch and crack for determination
of the stress-intensity factor, K
I
and crack opening displacement, COD, or
for determination of energy parameter, J-integral. In addition, one should
compare conditions for crack propagation at high-cycle fatigue (HCF) and
low-cycle fatigue (LCF) on one hand, and behaviour of the welded joint on
the other hand. Based on it, one can get a picture of behaviour of the welded
joint affected by fatigue loading, and the possibility of application of J-
integral as a universal parameter of elastic and plastic behaviour of a
material with a crack, and their effect on the problem of fatigue-crack
propagation.
In present paper, measurements of J-integral at LCF for the specimens
made of low-alloy high-strength HSLA steel (PM) and weld metal (WM) of
their welded joints have been presented.

2. MATERIAL
For these tests, HSLA steel of NIONIKRAL 70B designation with
welded joints made in submerged arc-welding (SAW) process with US-80B
wire was chosen.
Chemical composition of tested material has been presented in Tab. 1 and
chemical composition of the wire for the SAW process in Tab. 2,
respectively.
Table 1 Chemical composition of tested batch of NIONIKRAL 70B steel
(weight %)
C Si Mn Cr Ni Mo P S
0.19 0.4 1.11 1.06 2.59 0.25 0.019 0.024
Table 2 Chemical composition of US-80B wire for SAW of NIONIKRAL 70B
steel (weight %)
C Si Mn Cr Mo P S
0.09 0.19 2.15 0.49 0.84 0.014 0.013

Tensile properties of HSLA steel of NIONIKRAL 70B designation have
been shown in Tab. 3 and tensile properties of WM of the welded joint weld
in the SAW process with US-80B wire have been shown in Tab. 4.
Table 3 Tensile properties of NIONIKRAL 70B steel
Ultimate
tensile strength,
MPa
Yield stress,
MPa
Elongation A
5
,
%
Reduction of
cross section Z,
%
842 707 16 56.5
103

Table 4 Tensile properties of WM of tested welded joint
Ultimate
tensile strength,
MPa
Yield stress,
MPa
Elongation A
5
,
%
Reduction of
cross section Z,
%
848 701 17 46.2

3. PLAN OF THE EXPERIMENT
Welded plates, specimens to be cut from the plates, order of cutting of the
specimens and their testing are defined by the work plan.
Experimental examination of the behaviour of NIONIKRAL 70B and
WM of its welded joint included the following:
1. determination of the properties of PM and WM (the results have been
shown in Tabs 3 and 4);
2. determination of fatigue-crack growth at HCF;
3. measurement of compliance of a specimen and determination of
dependence of compliance and crack length;
4. establishment of dependence of fatigue-crack growth rate (increase of
crack length per cycle) and range of stress-intensity factor;
5. determination of J
R
-curve and critical value of J
IC
-integral, and
6. monitoring of J-integral value at LCF.
According to the plan for cutting shown in Fig. 1, from two welded plates
dimensions of which were 550 x 330 x 12 mm round specimens for tensile
tests and CT specimens for testing of fracture mechanics were made.
Figure 1 Plan for cutting of CT specimens and specimens for tensile tests cut from
PM and WM of SAW welded joint
In this paper, the results of monitoring of J-integral value at LCF will be
presented; other results can be found in [1].
L. Milovi
1
, S. Bulatovi, Z. Radakovi, V. Aleksi, S. Sedmak

4. THE RESULTS OF TESTING OF J
R
-CURVES WITH LCF
PHASES
For application of HSLA steel in welded structures, its behaviour when
affected by LCF is important, especially when a part of the structure contains
a crack, [2-6].
4.1. TESTING OF SPECIMENS FROM PARENT METAL
CT specimens (B=11.85 mm; W=86 mm; a
0
=32 mm) from PM were
tested for resistance to cracks based on J
R
-curves with phases of one-
direction LCF for analysis of the effect of LCF on shape of J
R
curves and
value of J
IC
. The data on testing of the subject specimen have been shown in
Tab. 5.
As one can see from Tab. 5, testing was conducted in a few phases shown
in Fig. 2. Taking into consideration noticeable proportion of LCF, in these
tests residual strain, measured by the magnitude of residual COD, was
monitored.
Figure 2 Fracture surface of CT specimen cut from PM
105

Table 5 Data on complex tests for determination of J
R
curves with phases of LCF
for CT specimens taken from PM

FATIGUE
COMPLIANCE

CRACK OPENING
DISPLACEMENT
(COD)

T
e
s
t

p
h
a
s
e

U
p
p
e
r

f
o
r
c
e
,

F
m
a
x
,

k
N

L
o
w
e
r

f
o
r
c
e
,

F
m
i
n
,
,

k
N

R
=
F
m
a
x
/
F
m
i
n

F
r
e
q
u
e
n
c
y
,

f
,

H
z

I
n
c
r
e
a
s
e
,

N
o
.

o
f

c
y
c
l
e
s
,

N
,

c
y
c
l
e

T
o
t
a
l

N
o
.

o
f

c
y
c
l
e
s
,





N

c
y
c
l
e

I
n
c
r
e
a
s
e
,

c
r
a
c
k

l
e
n
g
t
h
,

a
,

m
m

T
o
t
a
l

c
r
a
c
k

l
e
n
g
t
h
,


a
,

m
m

M
a
x

s
t
a
t
i
c

f
o
r
c
e
,

k
N

R
e
s
i
d
u
a
l
,

m

C
u
r
r
e
n
t
,

m

T
o
t
a
l
,

m

C
o
m
p
l
i
a
n
c
e

C
,

m
/
k
N

A

32 50

408 408 8.16
32 70 576 576 8.23
32 30 67 244 311 8.13
32 50 67 408 475 8.16
B
36.1 5.1 0.14 26 5000 5000 2.7 34.7 70 67 597 664 8.53
50 5.7 0.11 1 300 5300 1.2 35.9 50 96 435 531 8.7
36.1 5.1 0.14 28 10
4
15300 2.02 37.92 30 96 280 376 9.33
C
4.18 42.1 89 96 3808 3904
D
40 2816 504 3320 12.6
32 4.8 0.15 21 2000 2000 0.5 42.6 40 2816 512 3328 12.8
32.2 5.6 0.17 21 5000 7000 1.8 44.4 40 2816 608 3424 15.2
32.2 4.8 0.15 21 5000 12000 3.1 47.5 40 2864 741 3605 18.53
32.2 4.8 0.15 21 5000 17000 3.8 51.3 40 2920 939 3859 23.46
40 4 0.1 0.9 100 17100 1.1 52.4 40 2934 1019 3953 25.74
E
53.6 45.5 2994 2568 5562

43.5 3.5 0.08 30 0.22 53.82 43.5 4426 160 4586
54.65 41.75 4558 2400 6958
35 5 0.14 18 0.4 55.05 39 5806 128 5934
F
24.9 5.3 0.21 20 4200 4200 9.35 64.4 10 5878 688 6566 68.8
G
0.7 65.1 18.75 5966 2760 8726

18.5 2 0.11 30 0.15 65.25 18 7454 152 7606
0.2 65.45 17.25 7482 2416 9898 74.3
H

11.25 1.25 0.11 20 5010 5010 4.05 69.5

8.7 0.85 0.11 20 2100 7110 7.3 76.8
L. Milovi
1
, S. Bulatovi, Z. Radakovi, V. Aleksi, S. Sedmak

A-static tension for determination of compliance; B-fatigue-crack propagation
a
0I
with determination of; C-J
R
-curve; D-fatigue-crack propagation a
0II
with
determination of compliance; E-J
RII
-curve with LCF phases; F-HCF; G- J
RIII
-curve
with LCF phase; H-HCF
4.2. TESTING OF SPECIMENS FROM WELD METAL
The plan of testing of CT specimen (B=11.8 mm; W=86 mm; a
0
=32 mm)
with a notch in WM differs from the previous one, as in this testing LCF has
had the most important role. The data from this testing are presented in Tab.
6, but it should be noted that monitoring of the data on testing was stopped
after the second LCF (R0.5).

5. DISCUSSION
5.1. TESTING OF SPECIMENS FROM PM
Starting compliance of the specimen with a notch in PM was determined
by static loading (test phase A in Tab. 5).
Table 6 Data on complex tests for determination of J
R
curves with phases of
LCF for CT specimens taken from WM

FATIGUE
COMPLIANCE
M
a
x
.

s
t
a
t
i
c

f
o
r
c
e
,

k
N

CRACK OPENING
DISPLACEMENT (COD)
C
o
m
p
l
i
a
n
c
e

C
,

m
/
k
N

T
e
s
t

p
h
a
s
e

R
e
g
i
s
t
e
r
e
d

d
a
t
a

p
o
i
n
t

U
p
p
e
r

f
o
r
c
e

F
m
a
x
,

k
N

L
o
w
e
r

f
o
r
c
e

F
m
i
n
,
,

k
N

R
=
F
m
i
n
/
F
m
a
x

F
r
e
q
u
e
n
c
y
,


f
,

H
z

I
n
c
r
e
a
s
e

N
o
.

o
f

c
y
c
l
e
s
,

N
,

c
y
c
l
e

T
o
t
a
l

N
o
.

o
f

c
y
c
l
e
s
,

N

c
y
c
l
e

I
n
c
r
e
a
s
e
,

c
r
a
c
k

l
e
n
g
t
h
,

a
,

m
m

T
o
t
a
l

c
r
a
c
k

l
e
n
g
t
h

,

a
,

m
m

R
e
s
i
d
u
a
l
,

m

C
u
r
r
e
n
t
,

m

T
o
t
a
l
,

m

A 1
32 35 242 242 6.93
B
2
36 3.6 0.1 25 200 200 0.5 32.5 35 252 252 7.2
3
33 3.3 0.1 25 27600 27800 1.5 34 35 261 261 7.46
4
33 3.3 0.1 25 14600 42400 2.4 36.4 35 280 280 8
C 5
4.1 40.5 88 944 915 1859 10.4
D
17
71 7.1 0.1 0.5 50 50 70 840 12
18
70.5 7.05 0.1
0.5
50 100 69 862 12.5
19
69 6.9 0.1
0.5
50 150 67 911 13.6
20
66 6.6 0.1
0.5
50 200 62 992 16
21
62 6.2 0.1
0.5
50 250 57 1049 18.4
107

22
57 5.7 0.1
0.5
50 300 54 1064 19.7
23
53 5.3 0.1
0.5
50 350 48 1113 23.2
24
46 4.6 0.1
0.5
50 400 41 1148 28
25
38 3.8 0.1
0.5
50 450 32 1145 35.8
26
27 2.7 0.1
0.5
50 500 21 61.5 19 1500 906 2406 47.7
E
27
19.84 9.9
0.5 0.5
50 50 22 1084 49.3
28
21.92 10.96
0.5 0.5
50 100 21.6 1080 50
29
21.52 10.76
0.5 0.5
50 150 213 1075 50.5
30
21.36 10.68
0.5 0.5
50 200 21.2 1087 51.3
31
20.96 10.48
0.5 0.5
50 250 20.8 1096 52.7
32
20.64 10.32
0.5 0.5
50 300 20.6 1092 53
33
20.16 10.08
0.5 0.5
50 350 20 1094 54.7
34
19.68 9.84
0.5 0.5
50 400 5.3 66.8 19.4 1800 1100 2900 56.7
F
8.2 75
A- static tension; B-HCF with determination of compliance; C-J
R
curve; D-LCF
immediately after determine of J
R
curve; E-LCF R0.5; F-final HCF.
Residual strain at the point at which the data were registered, 3, resulted
from exceeded yield stress limit and small extension of the root notch. In
phase B, fatigue crack 37.92 mm-long was initiated, with initial and final
HCF between which 300 cycles of LCF, designated in Fig. 2, were realized.
Fatigue was interrupted from time to time, for measurement of compliance.
After that, standard determination of J
IC
in phase C using J
RI
curve followed.
Crack propagation in phase C was 4.18 mm and the value obtained, J
IC
=
212.78 kJ/m
2
, was in accordance with the results of previous testing. Upon
completion of phase C, measurement of COD was resumed from zero (knife
edges that support the gage arms were moved, as the full working range of
the COD-measuring device of 4 mm was exceeded). It was then that testing
of the specimen was interrupted for the first time. In the next phase, D, the
specimen was again subjected to fatigue, with determination of compliance
at intervals. After 4 cycles of HCF, 100 cycles of LCF followed. The
specimen prepared in that way, with initial fatigue-crack length a
02
=52.4
mm, was subjected to complex testing, phase E, by applying monotonous
loading and successive unloading.
5.2. TESTING OF SPECIMENS FROM WM
The testing started with three successive HCF, during which an increase
of physical crack length of 4.4 mm was attained, so that static loading F-
COD was conducted with initial fatigue crack of 36.4 mm. After sufficient
number of unloading points for determination of the value of plane-strain
fracture toughness was obtained, LCF was tested with minimum force
maximum force ratio R0.1. When the value of amplitude of upper force
dropped below 20 kN, testing of LCF was resumed at R0.5. To mark
L. Milovi
1
, S. Bulatovi, Z. Radakovi, V. Aleksi, S. Sedmak

attained crack length, we resumed with HCF until specimen fracture. In Fig.
3, the appearance of fracture surface is given, showing the differences in
behaviour of WM induced by heterogeneous structure and existence of the
defects. Face of the fatigue crack is blocked in development on one side,
probably because of the existence of some defect in the structure (surface A
in Fig. 3). That is why the crack face under static loading had the shape of an
irregular triangle (surface B in Fig. 3), with clear flat fracture in the crack
plane, that only after crack growth of 1.4 mm deflects at an angle of 45 in
the direction of maximum tangential stresses (plane stress state). In initial
phase of LCF, surface C in Fig. 3, the crack first propagates along the edges
of the specimen, face line becomes more regular in shape and fatigue crack
propagates, but still under conditions of plane stress state. The effect of the
fusion line becomes apparent. Namely, at the side surfaces fracture develops
at the boundary between heat-affected zone (HAZ) and WM. One should
also observe the surface D, on which fracture is partially brittle due to
structural defects in WM. HCF characteristic for the crack length shows
similar fracture surface, and it is only in the final stage of fracture that shear
lips forms again.
6. CONCLUSION
In present paper, experimental procedure for analysis of material crack
resistance using the parameters of fracture mechanics for three types of
effective loadings (HCF, monotonously increasing loading and LCF), as well
as at combined loading (monotonously increasing loading with LCF phases),
has been presented. By applying the above-specified loadings to PM of
HSLA steel NIONIKRAL 70B and its WM obtained by SAW welding
procedure, the properties relating to resistance of these two constituents of a
welded joint were determined and compared. The values obtained, J
IC
210
kJ/m
2
for PM and J
IC
90 kJ/m
2
for WM, indicate degradation of the
properties of the joint in WM. More inferior properties of the joint in WM
are also indicated by decrease of the value of J-integral in the phases of LCF.
Acknowledgment. This experiment has been performed within the
project TR35011 funded by the Republic of Serbia, Ministry of Education
and Science, whose help is gratefully acknowledged.
109

Fig. 3 Fracture surface of CT specimen cut from WM

References
[1] Application of Fracture Mechanics Parameters for the Assessment of Welded
Joint Behaviour in Conditions of Fatigue Loading, Doctoral Thesis Mohamed
Ahmed
[2] ASTM E1823-11: Standard terminology relating to fatigue and fracture
testing
[3] ASTM E606-04e1: Standard Practice for strain-controlled fatigue testing
[4] ASTM E647-11e1: Standard test method for measurement of fatigue crack
growth rates
[5] ASTM E813-81: Standard test method for J
IC
, a measure of fracture
toughness
[6] Chow, C.L., Lu, T.J., Cyclic J-integral in relation to fatigue crack initiation
and propagation, Engineering Fracture Mechanics, 39(1), 1991, 1-20.




111


THE APPLICABILITY OF STRATEGY FOR RISK
BASED MAINTENANCE TO A PENSTOCK
Tamara Sedmak
1
, Stojan Sedmak
2
, Ljubica Milovi
3


1
University of Belgrade, Faculty of Mechanical Engineering, Serbia
2
Society for Structural Integrity and Life (DIVK), Belgrade, Serbia
3
University of Belgrade, Faculty of Technology and Metallurgy, Serbia
Email adress: tamarasedmak87@hotmail.com

Abstract: The risk of brittle fracture of pumping up penstock was recognized in an
early stage of design. To reduce the cost of penstock construction one tunnel was
accepted. For that most stressed penstock section HSLA steel, 700 MPa yield stress,
47 mm thick, was selected. Two problems were induced regarding brittle fracture.
Cold cracks in welding could be avoided applying undermatching approach. This
required to specify welding procedure and to qualify it for used thickness. Designed
thickness might produce plain strain condition. For detailed analysis of cracks effect
and stress state it was accepted to produce two full-scale prototypes of penstock
most stressed part. Obtained results by fracture mechanics methods have shown
sufficient crack resistance of PM, WM and HAZ in welded joints. It was necessary
to assure structural integrity of welded joints and penstock involving corresponding
approach for maintenance and repair. Maintenance procedures are in continuous
development, and most promising is probably risk based approach, developed in last
years. In the paper the base requirements of this approach are considered, and the
possibility and convenience to apply them for penstock is evaluated.

Key words: brittle fracture; welded joint; undermatching; penstock; prototype
model; risk based maintenance


T. Sedmak, S. Sedmak, L. Milovi
1. INTRODUCTION
Hydro-electrical power plant (HEPP) systems might require large amount
of water in the Storage Lake and high fluid flow rate for the operation. For
such a system the consequences of unexpected failure can be catastrophic,
producing a great risk in service. In order to avoid the failure, preventing
measures have to be applied /1/.
One of very important component in HEPP is a penstock, which can be
exposed to high stresses, and for that be susceptible to failure. To reduce the
risk level, operational safety of individual components in HEPP system,
including penstocks, must be at very high level, This requirement comprises
the assurance of high quality of produced components.
The failures of penstocks and pipelines are not frequent. Mechanical
damages observed before and during service, fatigue, corrosion defects,
welding imperfections and environment effect are referred as most important
causes. Plastic collapse and brittle fracture are not cited often as failure
cause, because applied steels for pipelines are ductile, and pipelines are used
in the region above nil-ductility transition temperature.
However, cracks, brittle fracture and leakage had been experienced in
penstocks produced of weldable high strength low alloy (HSLA) steels,
developed and applied for reducing manufacturing costs by lower wall
thickness /2/. Proof pressure test is required before acceptance for service, in
some cases also testing of model is necessary.

2. EXAMPLES OF BRITTLE FRACTURE AND LEAKAGE
Two examples of failures /2/ are selected to illustrate the significance of
risk analysis for penstocks. They also have shown the significance of
welding procedure specification (WPS) for maintenance system and possible
risk in service of a penstock.
Typical example of fast brittle fracture is catastrophic failure of penstock
(length 2640 m, hydrostatic water pressure 864 m), occurred in 1973 year
during pressure proof test in one knee close to machine house in HEPP
Santa Isabel in Bolivia. Failure occurred at pressure 735 m, e.g. at 84% of
design pressure. Water jet passed through hole 1 m long and 0.7 m wide, and
destroyed tropical vegetation along 130 m and 10 m in width. About 6000
m
3
of water leaked for one hour, before the closing the valve in surge tank.
The penstock was repaired by new segments for the knee. The tubes were
produced of quenched and tempered steel Aldur 50/65D. This steel is
designed for penstocks. In damaged knee pipe diameter was 1.15 m, plate
thickness 22 mm. Mechanical properties, tested after failure, corresponded to
specification. Metallographic examination revealed that failure cause is
brittle fracture, initiated in the heat-affected-zone (HAZ) of longitudinal
113

welded joint, performed by sub-merged arc welding (SAW) procedure.
Crack developed from the initiation point on both sides in the directions
parallel to weld, and arrested at transversal manual arc welded joint, where
continued to grow in direction normal to the weldment. It is evaluated that
the possible failure cause is weld repair, since the preheating had not been
applied according to welding procedure specification (WPS). Since involved
steel is ductile, tough and crack resistant, the presence of high level of
welding residual stress due to improper welding technology reduced
penstock strength. After repair, in repeated test, penstock passed pressure
30% above design pressure, corresponding to 1170 m.
The next example, failure of penstock of Peruica, due to partial
leakage, showed the significance of quality assurance in welding. Neither
brittle fracture nor complete leakage occurred, but the occurrence of cracks
in welded joints required measures for preventing of the break of power
plant operation. Cracks occurred in welded joint of a ring, 100 mm wide,
consisting of 6 circumferential segments (pos. 104 in Fig. 1). The diameter
of penstock (pos. 103 left) is reduced from 4000 mm to 3400 mm in inlet
line for turbine (pos. 103 right). Two pipelines were produced by different
manufacturers, and significant misalignment and end distance occurred,
what was solved by the collar (pos. 105), welded before the ring and
accepting the loading by fillet welds. The ring served to fill the gap. After
welding, the tunnel was filled with the concrete outside penstock, and
penstock was protected inside by lacquer. Steel of 450 MPa yield stress,
microalloyed by vanadium, was selected for collar and ring. After 10 years of
service numerous cracks were revealed, several hundreds millimeters long, in
both ring welded joints A and B (Fig. 1), some of them passing through
the weld and reaching fillet weld of pos. 105, i.e. the space between collar and
ring. During the examination of empted penstock the water was found in
cracks, confirming that crack depth reached the thickness of the ring, i.e. tube.
The monitoring of cracks showed that they did not grow, or grow slowly.
T. Sedmak, S. Sedmak, L. Milovi

Figure 1. Design of collar (pos. 105) and ring (pos. 104) for joining the penstock

Examination had shown that cold cracks are in question, occurred due to
improper preheating and the rigidity of penstock at ring welding. Once
initiated, they developed due to corrosion. Significant for their occurrence
was an overloading of penstock in early stage of service, followed by
vibrations. Cracks did not endanger directly the integrity of structure, but
affected the contact of inner welded joints of collar with the water, what, in
addition to acting stress concentration, could initiate cracks also in loaded
welded joint of collar. For this reason the repair is performed by change of
all segments of a ring.

3. DEVELOPMENT OF MAINTENANCE SYSTEM
Maintenance of technical systems has been developed and improved
during the years. Corrective maintenance, which implies to eliminate the
effects of experienced failures, is the first generation of maintenance
strategies, simple, bur no more attractive. Second generation was the
scheduled maintenance, which considers higher plant availability, longer
equipment life and lower costs. Past thirty years many complex strategies
have been developed as third and fourth generation. Those include TPM
(total productive maintenance), LCC (life-cycle costing), RCM (reliability
centered maintenance), RBI (risk based inspection), and RBM (risk based
maintenance) /3/. Nowadays most attractive are risk based maintenance and
inspection, because they can assure best practical results
115

Risk is defined as the combination of the probability of an event and its
consequences. Risk analysis can provide information for different type of
consequences that can arrive from failures of equipment, like environmental,
health, safety and business consequences. This is very important for large
and complex industrial systems, like oil refineries, petrochemical and
chemical plants, steel production and power plants.
However, current practice of inspection and maintenance planning in
power plants is still mostly time oriented and based on prescriptive empirical
rules and experience rather than being an optimized process where risk
measures for safety and economy are integrated /4/.
The major challenge for a maintenance engineer is to implement a
maintenance strategy, which maximizes availability and efficiency of the
equipment, controls the rate of equipment deterioration, ensures a safe and
environmentally friendly operation, and minimizes the total cost of the
operation. This can be achieved only by adopting a structured approach to
the study of equipment failure and the design of an optimum strategy for
inspection and maintenance /5/.
For the selection of a maintenance strategy using a risk-based approach is
essential to develop cost effective maintenance polices for mechanized and
automated systems because in this approach the technical features (such as
reliability and maintainability) are analyzed considering economic and safety
consequences, according to Kumar, /5/.
Further more, the use of risk-based methods in inspection and
maintenance of piping systems in power plants gives transparency to the
decision making process and gives an optimized maintenance policy based
on current state of the components.
Lack of unique standard for RBM results in various methods and
techniques for analyzing risk and making inspection decisions based on
those analyses. According to Ref. /6/ there is no unique way to perform risk
analysis and involve RBM. Different approaches are reported, ranged from
only qualitative to complete quantitative /5/.
Only available applicable risk based standard is American Petroleum
Institute standard (API 581, Risk Based Inspection-Base Resource
Document). However this is standard for American industry and applicable
only for process plants.
Extensive European project RIMAP /7/, started in 2001 and finished in
2004, had been induced to offer a European standard for RBM. It has
produced four industry specific workbooks for the petrochemical, chemical,
steel and power generation industries, aimed to provide more specific
guidance on how to apply the RIMAP approach in these sectors. However,
this approach is to complex, and will not be considered in this paper.
In this paper it is accepted that the level of failure consequences is very
high /1, 2/, and for that the probability of failure should be very low.
Relevant parameters for failure occurrence are required, and they can be
T. Sedmak, S. Sedmak, L. Milovi
established by theoretical and experimental analysis, and, when possible,
supported by a numerical analysis /8/.

3.1. Risk based maintenance optimization
A qualitative risk assessment ranks system and components relative to
each other. When a qualitative risk assessment should be perform, relative
failure probability and consequence severity can be classified in broad
groups, assigned as high, medium and low. Although any number of
groups could be applied, probably maximum five failure probability and
consequence severity groups can be accepted with sufficient confidence.
Qualitative analysis uses words to describe the magnitude of potential
consequences and the likelihood that those consequences will occur. These
scales can be adapted or adjusted to suit the circumstances, and different
descriptions may be used for different risks /9/.
Quantitative analysis includes acquisition and elaboration of the data,
related to the equipment history and failure modes and consequences. It is
necessary to quantify probability of failure occurrence and consequences,
which product represents risk value.
According to API, as well as to RIMAP, this can be performed on three
different levels, depending on detail of analysis. In API approach the levels
are categorized as qualitative, semi-quantitative and quantitative analysis,
categories in RIMAP are at screening, intermediate and detail levels.
It is well known that there are also other different scales for consequences
and likelihood, and corresponding risk matrix. Scales and matrixes can be
defined in respect to specific analyzed problem, with no strict rules. For that
it is difficult to select a proper matrix.
First, i.e. qualitative level is based on one general matrix, presented in
Table 1, which is applied in the Project Risk NIS /10/. In this matrix,
consequences are categorized, based on several parameters (health, safety,
environment, business, security) as A to E, A indicating low, almost
negligible consequences, and E fatal and serious consequences. Probability
categories are graduated by 1 to 5, category 1 representing very unlikely
detrimental event, once in more than 100 years (1x10
-4
), and 5 representing
very probable event occurring at least once in a year (1x10
-1
).








117

Table 1. Scheme for risk based qualitative evaluation of maintenance
A B C D E
5
Very high
risk
4
High risk
3
Medium
risk
2
Low risk
1
(Very low,
negligible risk)
Consequence category
P
r
o
b
a
b
i
l
i
t
y

c
a
t
e
g
o
r
y


As it has been mentioned, the consequences of penstock failure can be
extremely serious /2/, indicating category E. For safe and reliable use of
penstock it is of utmost importance to assure extremely low frequency of
detrimental event, measured as 1x10
-4
, or once in more than 100 years. This
can be achieved by special measures in all steps of construction and
operation. On the other hand, very strict requirements have to be posed for
such a structure, complex regarding design and manufacturing, to prove that
detrimental events are unlikely to occur. This might be possible only with
sufficient and confident data, experimentally verified.

3.2. Peculiarities for risk-based approach to penstock
Dominant failures of pressure equipment are fast fracture, leakage and
corrosion. Fast fracture could be brittle fracture under plane strain condition
or ductile fracture due to overloading in plane stress condition. Leakage is a
consequence of through wall crack, achieved as time dependent stable crack
growth. Corrosion can be developed in specific environment condition, and
stress corrosion is supported by applied stress. Common feature of these
three failure modes is the existence of crack. Welded joints are prone to
cracking, and they are most critical regions of welded structure. For that the
quality level of performed welded joint has to be assuring following strict
requirements during manufacturing, according ISO 9000 series standards.
Structural integrity depends of crack behavior. For the control of a crack
two aspects are important. It is necessary first to detect crack and to identify
its location and size by different non-destructive testing (NDT) methods.
Then crack significance has to be assessed applying convenient parameter
and method based on fracture mechanics. To assure safe operation according
to European directives for pressure equipment (PED) these two aspects
should be considered before the equipment is accepted for European market,
T. Sedmak, S. Sedmak, L. Milovi
since the welding quality can not be verified on final product but has to be
induced during manufacturing /11/. Welded joint quality in service has also
to be controlled by inspections in maintenance system.
However, penstock is in service not available for inspection in short time
distances or by continuous monitoring, but only when it is emptied.
Accepted inspection interval for HEPP is ten years, and NDT of penstock
should be performed then on selected critical welded joints. For that, from
risk point of view it is necessary to assess the risk level for all welded joints
before inspection, and perform the inspection only on joints of high risk.

4. PROTOTYPE MODEL OF A PENSTOCK
Pumped-up storage HEPP Bajina Batain design period, about 40 years
ago, represented the solution of worlds highest head and water speed.
Design is performed by Toshiba, Japan, based on single-stage two units
Francis type reversible pump-turbine of maximum output 315 MW at
maximum head and revolving speed of 428.6 rpm. Pumping rate is
maximum 621.3 m, discharge 50.8 m
3
/s, maximum pump input 310 MW.
Rotor runaway speed is 650 rpm. Maximum pressure in penstock
corresponds to the water height level of 900 m.
The failure in HEPP Santa Isabelin Bolivia /2/ was clear warning to
investor regarding the risk of brittle fracture when penstock is produced of
HSLA steel. Accordingly, it was decided to perform experimental testing of
two full scale prototype models of the penstock. Prototype, produced of
quenched and tempered HSLA steel SUMITEN 80P (SM 80P), 47 mm thick,
"Sumitomo", Japan. (Yield strength 700 MPa, ultimate tensile strength above
800 MPa), presented the most stressed part of the penstock (Fig. 2).
Penstock welding had to be partly performed in the field. i.e. in tunnel,
and requirements could be satisfied only by educated, skilled and approved
welders. Due to limited experience with this class of steel, certification of
welders, welding procedure specification (WPS) and its qualification were
required by user, similar to nowadays standards EN287 and EN288. The
second problem was adopted safety margin of only 1.7 regarding steel yield
strength according to German specification /12/ for HSLA steel. In other
specifications, higher safety margin have to be accepted, in Japan for this
steel grade recommended safety margin is 2.07. Full scale test is the most
informative and serves as final proof for the quality and safety of a welded
structure, giving realistic response of loaded welded joints behavior. For
that, fitness for service (FFS) assessment was required, in order to
understand better crack significance, resistance to brittle fracture and stable
crack growth /13, 14, 15/.
119

It is to emphasize that the main motivation for full scale model testing
was to gain sufficient data for risk significance assessment, based on
experience and available knowledge in that time. These data are helpful also
in the actual analysis based on risk. For that, most important information
about the design and manufacturing of the prototypes, including welding and
welded joints properties, are presented here sufficiently to explain possible
application of experimental results for determination of parameters relevant
for risk, at least on qualitative level. Extended experimental investigation
according to specified programs regarding brittle fracture, stable crack
growth and welded joints properties enabled to assess confident data about
material and welded joints behavior under operational loading /8/.
Two prototypes manufacturing (Fig. 2), was also used to gain necessary
experience in welding.

Figure 2. Full-scale model of penstock most-stressed segment:
1-mantle; 2-lid; 3-stiffener; 4- supports of leg L-longitudinal, C-circular;
MAW - manual arc welding (M); SAW-submerged arc welding (S)

Matching between strengths of parent metal (PM) and weld metal (WM)
has an important role in the service of welded structures, since high stress
should cause local plastic strain due to applied undermatching and stress
concentration. Welded joints are generally designed as overmatched, with t
T. Sedmak, S. Sedmak, L. Milovi
WM superior in strength compared to PM, with plastic deformation localized
in PM. The situation turns to be complex in high strength steel which should
be welded as undermatched to prevent cold cracking. In undermatched
welded joint plastic deformation will start in WM, and PM can start to yield
when WM strain hardening capacity is partly exhausted /16/. Anyhow, in
welded joints microstructure of HAZ is heterogeneous. In non-uniform
microstructure the straining can be constrained and plane strain can prevail,
critical for brittle fracture. Stress concentration and residual stress, inevitably
present in welded joints, can contribute to plain strain condition.
The overall behavior of a welded penstock under load was analyzed based on
results of three approaches (crack initiation, crack propagation and arrest,
stable crack growth), allowing an evaluation of crack significance. Obtained
data in experimental investigation of prototype, together with the data about
material properties and heterogeneity effect /15/, can also serve to select crack
parameters and to calculate and specify their referent values for risk based
inspection and maintenance strategies. To achieve this, additional analysis is
necessary, generally followed by numerical modeling,.

4.1. Manufacturing of penstock prototype
Cylindrical mantle of full-scale model, consisting of 3 segments, 973 mm,
970 mm and 1943 mm long, designed with 5 knee corresponding to the
penstock transition segment, was covered with two shaped lids (Fig. 2). The
steel plates, two for each segment, were rolled and welded.
Welded joints, designed for penstock longitudinal (L) and circular (C)
weldments with preparation given in Fig. 2, were MAW and SAW welded,
using basic low hydrogen electrode LB118 for MAW and core wire
U8013+M38F flux for SAW welding, produced by "Cobe Steel", Japan. To
minimize the influence of the lids on the stress state in the mantle, two ring
stiffeners were welded near circular weldments. Certified welders welded
the prototype models and later the penstock. Specified and qualified manual
arc welding (MAW) and submerged arc welding (SAW) procedures were
used, also applied in subsequent penstock fabrication.

4.2. Burst test - resistance to fast crack propagation
The burst test was performed on the first model, produced by "Metalna"
company. Here will be presented only some details of testing, main results
and conclusion. More details about this test can be found in Ref. /17/.
Three additional longitudinal SAW welds were made on cylindrical
mantle bottom shell (Fig. 2), providing three crossings with circular MAW
weld metal. Two artificial surface cracks, sized 1804.3 mm and 506 mm,
were produced in longitudinal welded joints with the tip positioned in HAZ,
121

and one crack (403 mm), was embedded in third longitudinal weld. The
instrumentation consisted of two crack opening displacement (COD) gages
over surface cracks, 22 strain gages, 9 Moir grids, 3 acoustic emission
sensors, pressure transducer and a measuring system of outer periphery.
The full-scale model was first statically pressurized by uniformly raising
water pressure in two steps, achieving maximum level of 117 bar, which
corresponded to the hoop stress
t
= 518 MPa in an ideal cylindrical vessel.
Next test was dynamic, to simulate water hammer, was performed by two
successive explosions of controlled rate, after the static pre-loading of 60 bar
(
t
= 265 MPa). The plastic deformation followed the explosion loading rate,
indicating stable crack growth.
The non-destructive examination showed that longer surface crack
extended for 28 mm in length, 10 mm and 18 mm on sides, and the second
surface crack did not propagate in length. Two small new cracks initiated in
testing. Detailed analysis revealed that fracture developed through brittle
region in HAZ, close to fusion line. No one crack passed the wall thickness.
The conclusion was that full-scale model and penstock can safely with-
stand the working pressure (90 bar) and effect of water hammer pressure in
the presence of large flaw (6 mm deep, 50 mm long) and imperfection,
caused by misalignment of weldments.
Burst test has shown that penstock is resistant to fast fracture. In that test
an approximate answer is obtained for critical crack size and crack behavior
during loading, similar to possible working loading.

4.3. Hydro-pressure test - resistance to stable crack growth
The hydro-pressure test of the second model with no crack enabled the
post-yield analysis of weldments. Trial samples for additional testing were
welded simultaneously with the model, see Ref. /18/.
The analysis of the behavior of crack free penstock under design loading
and expected overloading, and the effect of initial plastic deformation on
welded joint toughness, was the aim of the experiment, performed by GOA
company. This was achieved by comparing the results of mechanical and
fast fracture tests of corresponding specimens (Charpy V impact, drop weight
DW, and explosion bulge tests of MAW and SAW welded joints), taken
from the prototype after hydro-pressure test and local plastic deformation in
weldments, and from non-deformed trial samples.
Hydro-pressure test was performed at an ambient temperature (between
+6C and -3C). Strains were monitored by strain gages, and controlled by
moir grids. Acoustic emission sensors in the critical regions enabled the
control of large plastic strains or crack initiation, to prevent a catastrophic
fracture during pressurizing. In the first testing step the pressure reached
90.2 bar (
t
=399 MPa), corresponding to operating pressure. In the second
step pressure was 120.6 bar (
t
=533 MPa), that is close to the total operating
T. Sedmak, S. Sedmak, L. Milovi
and water hammer load. The measured values of strains for selected location
and strain gages (SG) (2; 34; 53; 59) enabled to quantify residual plastic
strains,
pl
, (Fig. 3). The level b corresponds to maximum strain achieved in
first step, level a indicates residual strain; level d is maximum strain in the
second step, and level c indicates total residual strains. Strain developed
uniformly in PM (53) and circular SAW WM SC (59). Total plastic strain of
0.1% after pressurizing is found on circular CS WM (59). Plastic strain
0.24% was found in WM LS1 (2, 34). The loop in Fig. 3 is attributed to the
combined effect of undermatching, WM strain hardening and strain redistri-
bution in unloading and reloading due to stored elastic energy release in PM.
Differences in tensile properties (yield strength; ultimate tensile strength;
elongation of weld metal) for welded joints and WM were remarkable,
regardless that the results fulfilled specification. The results of hardness
measurement were acceptable. There are two basic reasons for differences in
mechanical properties in tested prototype. Stress concentration is the first,
global, due to 5 knee, and local, due to weldment shape that affected strain
distribution. The second reason is level of undermatching, which is different
for SAW and MAW welded joints. Longitudinal welded joints are more
stressed then circular, and this produces more pronounced difference in
stress and strain distribution. It is also necessary to take into account the
tendency of pressure vessel to take under pressure simplest form of sphere or
cylinder to achieve more uniform stress distribution. With all this in mind it
is possible to conclude that the yielding will start first in position of lower
material strength, in undermatched weld metal, and at the location of highest
stress concentration in the vicinity of knee and where it is caused by
imperfection of welded joint shape.

Figure 3. Typical relationships between pressure and strain
123

Important conclusion from performed test is non-uniform behavior and
different local plastic deformations (Fig. 3), indicating complex material
response to loading. But it is not critical, since the behavior of plastically
deformed material in next loading will behave elastically.

4.4. Prediction of stable crack growth resistance
Next step was the assessment of residual strength of model cracked in
PM and in WM, and resistance to stable crack growth, by J integral,
comparing crack driving force (CDF), obtained by numerical Ratwani-Erdo-
gan-Irwin model (REI), and material J-R curve /19/.
Set of CDF, calculated for the prototype (Fig. 2) is given in Fig. 4. for
different pressures, defined by the ratio pR/WR
p0.2
. (p is pressure, R radius, W
wall thickness). CDF depends on material yield strength R
p0.2
. Resistance
curves for PM (steel SM80P) and SAW WM had been obtained by three
point bend specimens (SENB) 22.5x45x180 mm (ASTM E1152). They are
transferred to the CDF plot in Fig. 4. For assumed ratio a/W = 0.25 (crack
depth a = 11.75 mm) crack will grow in a stable manner for 3.75 mm in PM
and 4.25 mm in WM, and critical pressure reached 155 bar and 144 bar,
respectively. For a/W = 0.5 (a=23.5 mm) corresponding values are 8.5 mm
and 140 bar for PM, and 6.1 mm and 104 bar for WM. The results have
shown high crack resistance of SM80P and its SAW weld metal.


Figure 4. Procedure for residual strength assessment of cracked pressure vessel, with
crack driving force and J-R curves, applied for preliminary prediction

T. Sedmak, S. Sedmak, L. Milovi
5. DISCUSSION
The risk of failure by brittle fracture or leakage, which could endanger in-
service safety, had been well recognized in the case of HEPP Bajina Bata
penstock. In order to assess the level and significance of possible risk,
experimental investigation of fast fracture and stable crack growth were
performed with two full scale prototypes according to specified program.
The motivation for this investigation was to gain sufficient data for risk level
assessment, based on experience and available knowledge in that time, about
thirty years ago. Most important results, presented here, allowed to made
qualitative assessment regarding brittle fracture, strain distribution in loaded
penstock (Fig. 3), and crack resistance (Fig. 4). Obtained results have shown
that applying proper measures in design and manufacturing, including
welding; it is possible to assure safe operation of the penstock, based on
qualitative assessment of the risk. However, much valuable information
gained in this investigation has not been used and still are interesting for
additional consideration on how to quantify risk levels of probability and
consequence in meanwhile developed approach for inspection and
maintenance /7, 9,10/. In that sense this paper, aimed to analyze applicability
of new developed principle based on risk, can be considered as a
continuation of performed investigation.
Large amount of stored water in HEPP, which might be released in the
case of failure, indicates possible severe consequences, of level E (Table 1).
This level can be accompanied by probability category 2 for medium risk
and by category 1 for low risk, according to accepted approach. In both cases
confident data are required, experimentally checked for determination
quantitative level for the assessment of critical values for selected parameter.
As it is already mentioned, detrimental events which can cause the failure
occurrence in penstock are fast fracture, leakage after stable crack growth
and corrosion. All three failure modes are related to crack, its existence or
initiation. Fracture mechanics introduced convenient parameters for each of
them. For fast fracture it is plane strain fracture toughness K
Ic
, as critical
value of applied stress intensity factor K
I
. For stable crack growth most
convenient is J integral in the forms presented in Fig. 4 for curves of crack
driving force CDF and J crack resistance: the tangent point of corresponding
curves indicates the end of stable crack growth and beginning of fast fracture
after plastic deformation. Time dependent rate of crack advancement is
relevant for corrosion, and in the case of stress corrosion it is sub-critical
value of the stress intensity, designated as K
Iscc
. In previous chapters only
general procedure is presented, but in fact several other aspects had been
also considered in investigation, like heterogeneous microstructure of
welded joint /15/, especially in HAZ /20/, and undermatching effect /16/,
which could help for quantification of selected parameters. Based on
125

experience and available data from literature, corrosion and stress corrosion
are not critical in this penstock. Consequently, they were not requested by
HEPP investor for the analysis. Requirements for penstock quality and safe
service have been considered and proved in this investigation, at least at an
empirical and indirect level. In the time of performed experiments it was
sufficient to evaluate the risk just qualitatively. Namely, actual risk based
approach has been introduced many years after the construction of the
penstock. The examples like pop-in occurrence in HAZ, J integral direct
measurement, the effect of undermatching are explained in Ref. /13, 15/ and
can be used for quantification of critical relevant parameters.
From risk aspect of view following four steps can cover the activities in
inspection and maintenance:
1. Non-destructive testing (NDT) for cracks detection, if they are present in a
structure, including their location and size.
2. Assessment of crack significance using specified parameters and their
determined measurable values.
3. Decision about repair actions for detected cracks.
4. Repair performance.
Let us consider actual situation first. Penstock is accepted for use thirty
years ago after strict testing according standards and executed proof pressure
test. So, this is sufficient to assume that penstock was defect free initially.
Having in mind penstock construction and operation data it is accepted that
the regular inspection should take place when HEPP operation would be
stopped for general inspection of the system, after every ten years in service.
According to available data this was performed twice till now. Only step 1
had been performed. It is to mention that, according to valid rules, educated
and skilled personnel, certified for applied NDT methods, has to be engaged
by independent testing institution. The results have shown that there is no
need for next step.
Meanwhile an intensive development in inspection methods and devices
took place, followed by extension of modeling and numerical programming.
As a result, continuous monitoring could be applied for different objects, like
that presented in Ref. /21, 22/. It might be expected that continuous
monitoring will be developed also for penstock. This will be beneficial for
risk based inspection (RBI).
For proper maintenance and safe operation, step 2, assessment of relevant
parameters, is important. For penstock they can be defined and quantified
based on presented results of performed experimental investigation of
prototypes, after an additional theoretical and numerical analysis. The
relevant data for risk based approach of considered penstock are:
- Characteristics of used material;
- Characteristics of welded joints and their constituent (PM, WM, HAZ);
- Effect of matching and heterogeneity of microstructure;
T. Sedmak, S. Sedmak, L. Milovi
- Crack resistance and material behavior in the region ahead the crack
tip;.
This is a complex task. It requires a series of standard testing, but this is
not sufficient for undermatched welded joint, applied in considered case.

6. CONCLUSSION
Risk significance regarding brittle fracture and stable crack advancement
preceding the leakage had been evaluated thirty years ago, after the
experimental analysis at the level sufficient to accept the penstock of HEEP
Bajina Bata. Development regarding inspection and maintenance, which
took place from that time, opened the possibility to assess risk consequence
and probability at more accurate level. Considering this new development
and performed experimental analysis it is found that new principles are
applicable to the penstock. The experimentally obtained results also contain
many valuable data, which can be additionally used for risk evaluation. For
that further investigation in this field is necessary and welcome.

7. REFERENCES
[1]T. Sedmak, E. Veg, Failure prevention of rotating equipment by
vibrodiagnostics, to be published in Structural integrity and life (IVK) 2/2012.
[2]J. Rastresen, Failures of penstocks, in Penstocks in hydroelectrical power
plants, (in Serbian), Yugoslav Electricity, Belgrade, pp. 229-236, 1979.
[3]T. Sedmak, Appliance of vibrodiagnostics in a tero technological risk
management master thesis, (in Serbian), Faculty of Mechanical Engineering,
Belgrade, 2011.
[4]J. Bareia, P. Buckb, B. Matscheckob, A. Jovanovicc, D. Balosc, M. Perunicic,
RIMAP demonstration project. Risk-based life management of piping system in
power plant Heilbronn; International Journal of Pressure Vessels and Piping 81
(2004) 807813
[5]L. Krishnasamy, F. Khan, M. Haddara, Development of a risk-based maintenance
(RBM) strategy for a power-generating plant; Journal of Loss Prevention in the
Process Industries 18 (2005) 6981
[6]N.S. Arunraj, J. Maiti; Risk-based maintenanceTechniques and applications;
Journal of Hazardous Materials 142 (2007) 653661
[7]R. Kauer, A. Jovanovic, S. Angelsen - Gjermund Vage ; Plant asset management
RIMAP (risk-based inspection and maintenance for European industries) the
European approach; ASME PVP-Vol. 488, Risk and Reliability and Evaluation
of Components and Machinery July 25-29, 2004, San Diego, California, US
PVP2004-3020
[8]Fracture mechanics of weldments, US-Yu joint project, Principal investigator S.
Sedmak, Annual reports, University of Belgrade, Faculty of Technology and
Metallurgy (TMF), University of Belgrade, 1982-1991.
127

[9]RBI-PETROL: RBI Risk Baed Inspection Petrol; ESPRIT Course #4a; Stuttgart,
june 2009
[10]Project: RiskNIS: Risk management and use of risk-based approaches in inspection,
maintenance and HSE analyses of NIS a. d. plants; TECHNICAL REPORT RBI
STUDY Gas Refinery Elemir (RGE) 2009
[11]A. Brean, J. Kurai, European Pressure Equipment Directive (PED) and
structure (integrity, (in Serbian), Structural Integrity and Life (IVK), 3(1), 31-
42, 2003.
[12]Dast Richilinie 011 The application of high strength weldable fine-grained
structural steels StE47 and StE70 (in German), Deutscher Ausschuss fur StahIbau,
1974.
[13]B. Boi, S. Sedmak, B. Petrovski, A. Sedmak, Crack gowth resistance of
weldment constituents in a real structure, Bulletin T. Cl de I'Academie serbe
des Sciences at des Arts, Classe des Sciences techniques, No 25, Belgrade, 21-
42, 1989.
[14]S. Sedmak, A. Sedmak, The experimental analysis of crack behavior in welded
pressure vessels, IIW Doc. X-1447, 1998.
[15] S. Sedmak, A. Sedmak, Welded penstock produced of high-strength steel and
application of fracture mechanics parameters to structural integrity assessment,
in Safety, Reliability and Risks Associated with Water, Oil and Gas Pipelines,
Eds. G. Pluvinage, M. H. Elwany, Springer, 271-286, 2008.
[16]N. Gubeljak, J. Predan, I. Rak, D. Kozak, Integrity assessment of HSLA steel
welded joint with mis-matched strength, Structural Integrity and Life (IVK),
9(3), 157-164, 2009.
[17]I. Rak, V. Gliha, J. Kuder, The testing of penstock model, welded of quenched
and tempered structural steel for pumping-ap hydroelectric power plant Bajina
Bata (in Slovenian), Metalna Strokovni bil. 84(1), 15-23, 1984.
[18]S. Sedmak, A. Radovi, Lj. Nedeljkovi, The strength of welds in HSLA steel after
initial plastic deformation, in Mechanical Behavior of Materials Vol. III, Eds. K. J.
Miller, R. F. Smith, Perganion Press, Oxford and New York, 1979, p. 435-446.
[19]M. M. Ratwani, S. Sedmak, B. Petrovski, The assessment of residual strength of
pressure vessel with surface crack by resistance curve, in Fracture Mechanics
of Weldments (in Serbian), Ed. S. Sedmak, GOA, TMF, Belgrade, 1985, p.
131-162.
[20] A. Sedmak, S. Sedmak, Lj. Milovi, Monograph Pressure equipment integrity
assessment by elastic-plastic fracture mechanics methods, Society for
Structural Integrity and Life (DIVK), Belgrade, 2011,
[21]N. Gubeljak, Application of stereometric measurement on structural integrity,
Structural Integrity and Life (IVK), 6(1-2), 65-74, 2006.
[22] N. Gubeljak, J. Predan, T. Valh, D. Valh, Continuous monitoring and structure
integrity assessment, in Fundamentals of Fracture Mechanics and Structural
Integrity Assessment Methods, Ed. S. Sedmak, MF, TMF, DIVK, IMS,
Belgrade, 2009, p. 261-284.

129



CONSTRAINT FACTORS USED IN LIMIT
ANALYSIS OF POLYETYLENE PIPES
SUBMITTED TO INTERNAL PRESSURE

M. Mouwakeh*, G. Pluvinage**, S. Masri*

Department of Applied Mechanics, Faculty of Mechanical Engineering,
University of Aleppo*
E.N.I.M., Metz, France**
Corresponding author : pluvinage@cegetel.net

Abstract : By using codes and finite element (FE) analysis, limit load
solutions of pipes containing surface cracks are determined. The study
covers cracks with constant crack length and depth in pipes with
different diameters D: 75, 90, 110, 125 and 160 mm . The crack
configurations consist of semi-elliptical surface cracks with a/D =0.067 ,
0.056 , 0.045 ,0.040 and 0.031.
The cracked pipes are subjected to internal critical pressure values
fromby codes ASME B31G and Chois formulas. Due to ductile
behavior of polyethylene pipe, failure occurs plastic collapse i.e when
the critical net stress reaches ultimate strength multiply by constraint
factor. In this paper the constraint factor is evaluated and it=ts evolution
with pipe diameter is analyzed. Three different definitions of the
constraint factor based on global or local approaches are also compared

Key words: Limit Analysis, Constraint Factor,Crack, Ultimate Pressure,
Finite Element.








M. Mouwakeh, G. Pluvinage, S. Masri


1. INTRODUCTION

The theory of limit analysis has appeared in the late 30s of last century, it
constitutes a branch of the theory of plasticity related to an elastic
perfectly plastic behavior.
A lot of works has been conducted to obtain limit load solutions in pipes
containing surface cracks subjected to internal pressure only or
combined load (internal pressure and bending) [1,2,3,4,5,6,7,8,9].
Generally, pipes fail in a ductile manner due to the behaviour of the
constitutive material. For these situations, failure prediction tools are
based on limit analysis. This failure criterion assumes that failure occurs
when critical net stress

reach ultimate strength Rm. One notes that


ductile failure is sensitive to net stress
N
(load divided by the ligament
cross section) whatever brittle fracture is sensitive to gross stress
g
(load
dividedby the entire section). The above mentioned criterion need to be
modified to take into account, constraint, geometry and loading mode
effects in the following manner:

= . (1)

where L is the so called constraint factor.

Design codes for pipes such as code ASME B31G [10] and Chois
formulas [11]are based on limit analysis to calculate a critical internal
pressure. They are based on limit analysis and incorporate safety factor
through a lower bound of a plot of experimental results. The basic
question is the values of the constraint factor which is incorporate in
these codes and if this value is close to values generally obtained. In
addition, it is interesting to know the evolution of constraint factor with
ligament size, defect and pipe geometries.

|
.
|

\
|
=
t
a
D
a
L L , (2)

where D is pipe diameter, t thickness and a defect depth.
Determination of critical pressure by codes ASME B31G and Chois
formulas has been made for 5 pipe diameters D, 5 thickness t and
constant defect depth a. These critical pressures p
c
lead to values of
critical net stress

and then to constraint factor L :



=

(3)

Calculation of stress distribution along the ligament has done using
Finite Element method depending on pressure values obtained from
ASME B31G and Chois formulas and then average value of maximal
131

principal stress
m
is calculated. Determination of constraint factor
deduced from finite element method can be expressed as following:

L

=
m
R
m
(4)

Stress distribution ahead of the crack tip led us to determine failure
process zone using the Volumetric Method [12] which is is a local
fracture criterion.Inside the fracture process, the effective stress
ef

which is the average value of the stress distribution acting as a local
fracture stress. Another definition of the constraint factor can be made:

L

=
eq
R
m
(5)

These three different constraint factors are compared in a discussion.


2. MATERIAL AND DEFECT GEOMETRY
The studied pipes are made of high density polyethylene which has the
following characteristics:
Table 1 : Mechanical properties of polyethylene

(MPa) (kg/m
3
) (MPa)
23.00 960.00 400.00 0.45

where
y
, , E and are respectively yield stress, volume mass, Youngs
modulus and Poisons ratio.

3. CRITICAL PRESSURE

Critical Pressure is calculated for different pipe diameters by using
codeASME B31G and Choiformulas as following:

ASMEB31G

For defects Parabolic ,the critical failure pressure is given by the
following expression:
4
t
D
D
L
0.8 for ,
t
D
D
L
0.8 1 M where
,
(a/t)/M (2/3) 1
(a/t) (2/3) 1
D
t ) 2(1.1
P
2 2
Y
f

|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
|
.
|

\
|
+ =
(



=
(6)
M. Mouwakeh, G. Pluvinage, S. Masri


where,
f
P , D, a
,
t, M,
Y
andL are the failure pressure, outer diameter,
maximum corrosion depth, wall thickness, bulging factor, yield stress
and longitudinal corrosion defect length, respectively.

Choisformulas
(7)
where
f
P ,
U
,
i
D , a, t and R are the failure pressure or maximum
pressure, ultimate tensile strength, inside diameter, defect depth, wall
thickness and average pipe radius, respectively.
In this study five pipe diameters with different wall thicknesses are
chosen. The dimensions of pipes are as the following:
Table 2 : pipe diameter and thickness.
D (mm) t (mm)
75.0 6.8
90.0 8.2
110.0 10.0
125.0 11.4
160.0 14.6

where D is the Diameter and t is the wall thickness.The crack geometry
is assumed to be semi-elliptical where 2c is the crack length and a the
crack depth. The crack configuration is identical for all pipes. The crack
dimensions are as the following: 2c =100 mm , a=5mm.
Evolution of computed critical pressure versus a/D ratio is shown in
figure (1). One notes that code ASME B31G and Chois formulas codes
have the same attendance, i.e. the critical pressure decreases in linear
manner with increasing of crack depth or a/D ratio. By comparison of the
two methods, we notice that the maximum relative difference is about
30%.Curve fitting procedure in Matlab gives the following relationships:

ASME B31 GP
f
=-37.77(a D )+5.1869

(Choi)P
f
=-63.244(a D )+6.0994

|
.
|

\
|
(
(

|
.
|

\
|
|
.
|

\
|
+

<
+ +

= ,
6
Rt
L
,
Rt
L
4
C
3
C
i
D
t )
U
2(
1
6
Rt
L
,
2
Rt
L
2
C
Rt
L
1
C
0
C
i
D
t )
U
2(
0.9
f
P
133


Figure 1: variation of critical pressure versus a/D ratio

4. CRITICAL NET STRESS
Critical net stress is calculated from critical pressure values for different
diameters by using thefollowing formula:

= p
f
. D/2t

(8)

where

is the critical netstress, p


f
is the critical pressure and t* is the
length of the ligament :

t

= t a (9)

wheret is the ligament thickness and a is the crack depth.From figure2
we notice that the critical stress increases with polynomial manner with
the increasing of a/D ratio or with the decreasing of diameter.By
comparison of the two methods, we notice that the maximum relative
difference is about 30%.By using curve fitting procedure , we obtain :

(ASME B31G)

=1E+07(a D )
4
- 2E+06(a D )
3
+135403(a D )
2
-3626.2(a D )+66.558
(Choi)

=1E+07(a D )
4
- 2E+06(a D )
3
+134817(a D )
2
- 3188.4(a D )+57.355


Figure 2:variation of critical net stress versus a/D ratio


Choi
ASME
0
1
2
3
4
5
6
0 0,02 0,04 0,06 0,08
P
f

(
M
p
a
)
a/D
ASME
Choi
10
20
30
40
50
60
0 0,02 0,04 0,06 0,08

(
M
p
a
)
a/D
M. Mouwakeh, G. Pluvinage, S. Masri



5. CONSTRAINT FACTOR L
L is calculated for each different pipe diameters by using the formula :

L =

R
m
(10)

Where

is critical net stress obtained from codes ,

ultimate tensile
strength.

Figure 3: variation of constraint factor L versus a/D.

One note from figure 3 that the constraint factor L increases with the
increasing of a/D ratio or decreasing diameter.The maximum relative
difference between the two methods is about 30% according to :

(ASMEB31G)
L=531608(a D )
4
- 87133(a D )
3
+5586.4(a D )
2
- 149.61(a D )+2.746

(Choi)
L =537902(a D )
4
- 92389(a D )
3
+5562.2(a D )
2
- 131.55(a D )+2.3663


6. AVERAGE CRITICAL STRESS OVER LIGAMENT
; CONSTRAINT FACTOR L*
The Finite Element Method (FEM) program ABAQUS [13] was used for
the computations of the stress distribution over ligament. The material is
assumed to be completely elastic perfectly plastic obeying the Von-
Mieses flow criterion. Critical pressure values obtained from codes are
applied on the internal surface of the pipe. FE calculations are done for
different diameters toobtain the maximal principal stress and

stresswhich is the average of the distributed stress over the


ligament.For this purpose 3D FE model is used for the analysis as shown
in the figure 4.

ASME
Choi
0
0,5
1
1,5
2
2,5
0 0,01 0,02 0,03 0,04 0,05 0,06 0,07
L
a/D
135


Figure 4: FE model of a surface crack in pipe

Mean stress increasesin polynomial manner with the increasing of a/D
ratio until a/D=0.06 and then starts decreasing .We obtained polynomial
function by using curve fitting as following :
(ASMEB31G)

m
= -2E+08(a D )
4
+3E+07(a D )
3
- 2E+06(a D )
2
+56990(a D )- 580.32
(Choi)

m
=-1E+08(a D )
4
+3E+07(a D )
3
- 2E+06(a D )
2
+51002(a D )- 520.79

L* is obtained from the formula :
L

=
m
R
m
(11)

where
m
is the average of distributed stress in the ligament obtained from
FE method and Rm is the ultimate tensile strength.Figure 5 shows the
variation of constraint factor L* versus a/D ratio for both codes. We
notice that constraint factor L* increase in polynomial manner with
increasing of a/D ratio until a/D=0.06 and then starts decreasing.

7. DETERMINATION OF CONSTRAINT FACTOR
THROUGH VOLUMETRIC METHOD
The volumetric method is a local fracture criterion, which assumed that
the fracture process requires a certain volume. This volume is generally
assumed as cylindrical with a diameter called effective distance X
ef
. The
physical meaning of this effective distance corresponds to the size of the
high stressed region at defect tip. This effective distance is considered as
the distance of the inflexion point on the stress distribution at defect tip.
A graphical method based on the relative stress gradient associates the
effective distance to the minimum of .The relative stress gradient is
given by:

M. Mouwakeh, G. Pluvinage, S. Masri


r
(r)
(r)
1
(r)
yy
yy

=
(12)

(r)
yy
is the opening stress. The effective stress is defined as the average
value of the stress distribution over the effective distance and weighted
by the stress gradient.
dr (r)) r (1 (r)
X
1

ef
X
0
yy
ef
ef
=

(13)

r is distance. Figure 6 gives an example of stress distribution along
ligament at defect tip for the pipe of 125 mm diameter submitted to
critical pressure calculated by ASME B31 G code.


Figure 6:stress distribution along ligament at defect tip for the pipe of 125 mm
diameter submitted to critical pressure calculated by ASME code.

Figure 6 gives also the relative values of these stresses compare to the
ultimate strength Rm=24.38MPa.

ef
>
N,c
>
m
>
Rm
>
g,c
(14)


8. DISCUSSION
Constraint factor gives an idea of the stress elevation due to constraint
introduces by geometry, ligament size, thickness and gradient effect.
However several definition of the critical stress can be used as the
critical net stress, the effective stress, the mean stress or the critical gross
137

stress. They leads to different definition of constraint factor, L, L*,
L**and L***:

=

=
m
R
m
L

=
eq
R
m
L

=
g,c
R
m


L** cannot be considered as a constraint factor because plastic collapse
is sensitive to net stress and not gross stress. It is given as indicative
values.


Figure 7: Chart of constraint factor values according to diameter and critical
pressure method (A: ASME code ; C Chois method).

Values of L*** L

=
g,c
R
m
are always less than unit. This point
confirms that plastic collapse is not sensitive to gross stress and then
L*** cannot be considers as a constraint factor.
L

=
eq
R
m
is a priory the most realistic value of the constraint factor
if we assumes that ductile failure needs a fracture process zone and that
in this zone fracture occurs when then effective stress reach a critical
value. Maximum value of L** is L** =2.45 which is less than the
theoretical value of L**=3 for pure plane strain conditions and for a
Poissons ratio of =0.3. Except for small and large diameters L** has a
value greater than 2 which seems acceptable.
=

values given by codes are less than 1.5 and conservative.



Values of L* L

=
m
R
m
are less than values of L. Both definitions
refer to an average value of the net stress over the ligament.
m
is an
average value over the ligament the longitudinal stress distribution and

N,c
is the gross stress of a pipe of a reduced thickness. This explains
certainly this difference.

M. Mouwakeh, G. Pluvinage, S. Masri




Figure 8:Variation of constraint factor L* versus a/D


(ASMEB31G)
L

=-7E+06(a D )
4
+1E+06(a D )
3
- 81130(a D )
2
+2351.3(a D )- 23.942
(Choi)
L

=6E+06(a D )
4
+1E+06(a D )
3
- 71868(a D )
2
+2104.2(a D ) - 21.487

9.CONCLUSION
Polyethylene pipe fail in ductile manner and failure criterion is plastic
collapse predicted by Limit Analysis. In this criterion, failure occurs
when net stress reaches the ultimate strength multiplied by the constraint
factor L.
This constraint factor quantifies the increases of flow stress due to the
plasticity preventing due to geometrical effect, scale and gradient effects.
In this study critical internal pressure has been obtained from codes
ASME B31 and Chois method.
Three definitions of the constraint factor has been proposed L; L*and
L**. The last one L** is based on a local failure criterion called the
Volumetric method and is certainly the more realistic if we assumes that
ductile failure needs a fracture process zone and that in this zone fracture
occurs when then effective stress reach a critical value . This definition
leads to L** close to 2 but sensitive to pipe geometry which seems
realistic. Value L obtained directly from critical pressure given by
ASME code or Chois method are more conservative.
Values of constraint factor are helpful for the choice of pipes material
through the value of the ultimate strength.

REFERENCES

[1] DNV-RP-F101.Corroded Pipelines,Det Norske Veritas,(1999).
[2] Dowling, A.R., Towley, C.H.A., The effect of defect on structural
failures: a two-criteria approach, International Journal for Pressure
Vessels and Piping,Vol. 3, 2 : 77-107,(1975).
[3] Pluvinage,G.General approaches of pipeline defect assessment.
Laboratoire de FiabilitMcanique ENIM-UPV Metz (France),(2008).
ASME
Choi
0,000
0,200
0,400
0,600
0,800
1,000
1,200
1,400
1,600
1,800
0,000 0,010 0,020 0,030 0,040 0,050 0,060 0,070
L
*
a/D
139

[4] Staat, M.Local and global collapse pressure of longitudinally flawed
pipes and cylindrical vessels, International Journal for Pressure Vessels
and Piping ,82, (3) : 217225,(2005).
[5] Staat, M. Vu Duc Khoi. Limit loads of circumferentially flawed pipes
and cylindrical vessels under internal pressure. International J ournal for
Pressure Vessels and Piping,83, (3) :188196,2006, .
[6] Sattari-Far,I.Dillstrom,P.Local limit load solutions for surface cracks
in plates and cylinders using finite element analysis. International
Journal for Pressure Vessels and Piping,81,(1) :5766,(2004).
[7] Lei,Y.J -integral and limit load analysis of semi-elliptical surface
cracks in plates under combined tension and bending. International
Journal for Pressure Vessels and Piping, (2003).
[8] Yagawa,G.Kitajima,Y.Ueda,H.Three-dimensional fully plastic
solutions for semi-elliptical surface cracks. International Journal for
Pressure Vessels and Piping, 74 :7780, (1997).
[9] Miller, A.G.Review of limit loads of structures containing defects.
International Journal for Pressure Vessels and Piping, 32:197
327,(1988).
[10] ASME B31G, Manual for determining the remaining strength of
corroded pipelines, The American Society of Mechanical Engineers,
New York, USA1991,.
[11] Choi, J . B. Goo, B. K. J . Kim, C. Kim, Y. J . Kim, W. S.
Development of limit load solutions for corroded gas pipelines,
International Journal of Pressure Vessel and Piping, 80, (2) :121-
128,(2003).
[12] Pluvinage, G. Fracture and Fatigue emanating from stress
concentrators, Editeur Kluwer, (2003).
[13] ABAQUS, RI: Hibbitt, Karlsson and Sorensen,Inc V5.8,
Providence, New York ,USA,(1998).









141


SERVICE ORIENTATION IN DISTRIBUTED
AUTOMATION AND CONTROL SERVICE

Cristiana Voican

University Politechnic of Bucharest,

Email address: voicancristiana@yahoo.com


Abstract
An experimental study shows the feasibility ofservice-oriented architectures
for industrial automation and control systems even with respect to lower,
real-time dependent control functions. For that purpose, general SOA-
guidelines were refined in order to cover the distribution of control functions
between services and the lay-out and management of devicebased sensor,
actor and control services. Particular emphasis was placed on the dynamic
lease-based binding of services which on the one hand provides flexible and
loose coupling of system components but on the other hand has to ensure
reliable communication and cooperation. The guidelines were applied to the
experimental implementation of a manufacturing cell control system using a
real-time version of the Java Runtime Environment.
The Device Profile for Web Services (DPWS) was used as basic
infrastructure technology. Test and evaluation were performed under
distributed simulation of technical processes and devices.








C. Voican
1.INTRODUCTION

Today, many modern business applications adhere to the paradigms
of service orientation and service oriented architectures in order to
create loosely coupled, modular software systems, easy to maintain
and to extend. In the field of automation and control systems, SOA-
based flexibility is of even more interest, because it contributes to
substantial reductions of installation and setup costs These costs are of
particular importance since manufacturing plants again and again have
to be adapted to new products resulting in changes of the technical
equipment and the process flows performed. Additional
reconfigurations are applied occasionally in the course of repair
measures in order to bypass defect equipment and to avoid expensive
production downtimes.
Despite the desired flexibility, however, there is a needs for stable
and reliable operation phases since the efficiency of the production
equipment usually depends on steady operational conditions. For a
certain manufacturing operation usually an ensemble of suitable
devices, machines and transport equipment is necessary. The members
of the ensemble must initially be configured in harmony with each
other and thereafter be available for a certain minimal period of
operation time, which may only be aborted due to exceptional
circumstances. The members of the ensemble have to be allocated
before configuration, some of them because they can only be used
exclusively, others may be sharable but have to allow for the
additional load.
In the service-oriented setting this means, that a client which may
be either a control application or a compound service must be able
to search, find and allocate a suitable ensemble of used services. Since
a used service may already have other obligations, it may not be
disposable and deny a current allocation request.
Then, one member of the planned ensemble fails, and the ensemble
as a whole is currently not useful.
Therefore, the client shall be able to withdraw the other allocation
requests and look for alternative ensembles.
In order to fulfill these functional requirements of temporary and
atomic ensemble allocation we extended the approach of lease-based
allocation by introducing an explicit reservation phase in a way that
reservation and allocation perform a two-phasem commitment.
143

Moreover we transposed the architecture of hierarchical control
systems to the field of service systems using the platform the Device
Profile for Web Services (DPWS) as basic infrastructure technology
supporting the communication between devices via service interfaces
as well as the exploration and binding of services. The application of
the resulting architecture guidelines and the usage of the lease-based
allocation were exemplified by means of a production cell scenario
using a real-time Java Runtime Environment.
In the sequel, we outline DPWS and its application to service-
oriented industrial applications.


2.SERVICE ORIENTED ARCHIECTURES

In SOA, interoperability of different platforms is established
through the definition of common communication protocol and
message exchange standards.
But not only in enterprise domain software service-orientation is a
feasible way of creating flexible software systems, as through the
growth of computing power of embedded devices these paradigms are
also applicable to embedded software solutions.
Universal PlugnPlay (UpnP was the first specification of a
service oriented infrastructure to be used in embedded application
scenarios, using SOAP and HTTP as a basic communication layer and
providing mechanisms for service discovery, action invocation and
event based communication schemes.
Its successor, the Devices Profile for Web Services (DPWS) , is
completely based on standardized Web service specifications and
defines a profile (a subset) for the use of Web service technology in
the embedded domain.


2.1.DEVICES PROFILE FOR WEB SERVICES

The Devices Profile for Web Services defines a common subset of
web service based communication patterns for use in embedded
devices. The protocol stack utilizes standardized internet protocols,
namely TCP/IP and UDP (Single- and Multicast). For basic messaging
HTTP and SOAP respectively SOAP-over- UDP are employed.
C. Voican
On top Web service protocols are arranged that deal with service
and device description, discovery, eventing and security. A DPWS
device may host several services, which can be discovered and used
by DPWS clients. The DPWS protocol stack is depicted in Figure 1.

2.2 SOA in Industrial Automation

The emergence of powerful but less power consuming, affordable,
and embedded computing components facilitates the employment of
SOA paradigms even in the world of industrial automation. Currently
a lot of proprietary standards in device control and communication
protocols often prevent the vendors.
Thus upgrades or extensions of the manufacturing automation
system tend to be costly and time consuming .
The usage of SOA in industrial automation provides a common
ground for interoperability of all devices in a device network.
Moreover an integration of low-level devices and highlevel
enterprise applications (e.g. an ERP system) is possible.
In the European ITEA SIRENA project the applicability of DPWS
in an industrial automation scenario was demonstrated for the first
time.



Figure 1. DPWS protocol stack

145

3. AUTOMATION AND CONTROL

An industrial control system commonly has a structure as depicted
in Figure 2. This architecture could be divided into three main layers:
sensors and actors, control and management.
The actual technical process is located at the bottom of the control
hierarchy and subsumes all technical lowlevel components involved in
the production process like motors, pushers or drilling machines. The
process is monitored by sensors, collecting data from the involved
resources including e.g. temperature, rpm or the position of work
pieces (indicated by a light-barrier state change).
This information is send via a specialized communication
infrastructure to the process control level and is repeatedly evaluated
by the control algorithm. Based on the sensor information the control
algorithm computes control signals which are in turn send to the
actuators connected to the technical process. Moreover status
information from the process control level is sent to the process
management level.
This may include forwarded sensor values, proinformation and
fault messages. At process management level a human operator
monitors the overall process behavior, adjusts particular parameters
and sends configuration commands to the process control system.
Besides the remote high-level controlling and monitoring of the
technical process, in some occasions (e.g. a severe fault that requires
local intervention and repair) the operator may be forced to directly
intervene with the low-level hardware components via the attached
control pan


4. SERVICE CONTROL ARCHITECTURE

The process control architecture shown in the last paragraph is the
structural basis for the service-oriented architecture presented in this
paper.
The serviceorientation of the devices involved in the technical
process and the attached sensors suggests the use of service-
orientation also on the control and management levels.
The sensors and actuators export their functionality through
defined interfaces which can be used by higher level control services.
Control services may also be layered and arranged in a service
C. Voican
hierarchy. Figure 3 illustrates this architecture: the application process
interacts with the technical process using the supplied control services.
The control services themselves are acting both as a service
consumer (client role) and service provider (server role) and thus
enable control service layering.
For example, a rotary disk consists of a rotation motor and a motor
for moving the conveyer belt on top of the disk.
Additionally the disk is supplied with sensors, detecting the
location of the work piece currently transported on the conveyer belt
and a sensor to measure the position of the rotary disk itself.
Both, the rotary part and the transportation part are each controlled
by their own control service.
For the control of the overall process of moving a work piece on
the disk, stopping the conveyer, turning the disk to its new position
and finally transporting the work piece away from the rotary table, an
additional control service is provided that uses the control services of
the particular parts of the rotary disk.
' Therefore the control services themselves offer service
functionality to higher level control or management services.
However, the stacking of control services is constrained by the
real-time requirements of the process, as each new layer of control
implies additional, time consuming
communication between the services.
.


Figure 2 Control system Figure 3. Service hierarchy

147

5. CONCLUSIONS

The services (e.g. sensor or controller) offer different interfaces
which can be categorized using the follow three classes:
functional purpose
discovery and description
service binding
The functional interface offers the functionality of the service, e.g.
a getVariable method for sensor or a setVariable method for actuator
services. The functional service interface of control services offers
high-level methods like drillHole.
The control services comply with the notion of so called function
building blocks (IEC 61499).
Each building block comprises input and output variables plus
local status variables. The functionality of a particular function block
is defined by the algorithm that is used to compute the outputs by
using the inputs and the local variables.
The discovery interface contains the necessary methods for services
to be able to answer to search requests and to provide data concerning
device type, location and binding address.
Finally, the binding interface subsumes the features for lease based
service binding and reservation.

6. REFERENCES

[1] H. Smit, F. Jammes, Service-Oriented Paradigms in Industrial
Automation, IEEE Transactions on Industrial Informatics, Vol. 1,
No. 1, pp. 62-70, 2005.
[2] C. Gray, D. Cheriton, Leases: an efficient fault-tolerant
mechanism for distributed file cache consistency, ACM SIGOPS
Operating Systems Review, Vol. 23, Issue 5, pp. 202-210, Dec. 1989.
[3] Universal Plug and Play (UPnP), http://www.upnp.org, 1999.
[4] Devices Profile for Web Services (DPWS),
http://schemas.xmlsoap.org/ws/2006/02/devprof/, 2006.
[5] Service Infrastructure for Real-time Embedded Networked
Applications (SIRENA), http://www.sirena-itea.org, 2006.
[6] Sun Microsystems, Jini, Network Technology,
http://www.sun.com/software/jini, 1999.
[7] Kapsers, Kfner, Messen Steuern Regeln: Elemente der
Automatisierungstechnik, Vieweg Verlag, 6th Edition, p. 253, 2006.
C. Voican
[8] WS4D.org Java Multi Edition DPWS Stack,
http://www.ws4d.org, 2007.
[9] Sun Java Real-time System 2.0 (Java RTS),
http://java.sun.com/javase/technologies/realtime, 2007.
[10] PROFINET, http://www.profibus.com/pn/, 2007.

























149


FLEXIBLE SERVICE BINDING IN DISTRIBUTED
AUTOMATION AND CONTROL SYSTEM

Cristiana Voican

University Politechnic of Bucharest,

Email address: voicancristiana@yahoo.com



Abstract. Particular emphasis was placed on the dynamic lease-based
binding of services which on the one hand provides flexible and loose
coupling of system components but on the other hand has to ensure reliable
communication and cooperation. The guidelines were applied to the
experimental implementation of a manufacturing cell control system using a
real-time version of the Java Runtime Environment.
The Device Profile for Web Services (DPWS) was used as basic
infrastructure technology. Test and evaluation were performed under
distributed simulation of technical processes and devices



1. INTRODUCTION

One of the key features of service-orientation is the use of loosely
coupled components. As all devices, sensors and actuators provide a
service interface the coupling of components can correspond to the
flexible binding of services.
This flexible binding of services demands for service description,
discovery and selection, and service association and linking
mechanisms.
The service description subsumes three basic parts:
C. Voican
Type and interface definition,
Binding and communication information,
Functional properties.
The type and interface definition of a service specifies the methods
and parameters associated with a specific service type. All services
that comply with a specific service type offer the same interface.
The binding and communication information contains information
about the actual communication endpoints and the basic
communication mechanisms, such as IP addresses and ports, and
application protocol regulations. At last, the functional properties
complete the information on devices in the automation system. They
e.g. include, which sensor is attached to which conveyer and what is
the exact position.
The service description is the basis for the discovery and selection
of matching services by the automation process and control services.
In our system, the discovery and description phase are based on
DPWS technology and thus adhere to the WS-Discovery and WS-
Transfer (for metadata exchange) standards.


2.. FLEXIBLE SERVICE BINDING

The association and linking of matching services with a particular
client is handled by our lease based binding approach to meet the
requirements of a flexible but also stable way for dealing with loosely
coupled services in the domain of industrial automation.
The notion of a lease was first introduced by and was used to
provide an efficient, fault tolerant way for using file caches in
distributed environments. Further on leases were used in Jini to grant
clients access to network services.
In the case a client wants to use a particular service, it issues a
lease-request which contains a duration for which the client wants the
lease to be valid.
The service responds with a denial or a grant. A granted lease is
valid only for the duration.
Thus the client has to request another lease for service use after the
current lease has expired or may prolong it before its valid duration
has passed.
151

In automation systems a client usually uses a set of services
(sensors, actuators, and controllers) and has to allocate a suitable
ensemble.
Therefore we extend the lease model by adding support for the
atomic allocation of service ensembles.
The atomicity property guarantees that a client either is granted
the leases for all requested services or it gets no lease at all. This
atomicity is achieved by a 2-phase algorithm, which is similar to the
2-phase-commit protocol. It is a lease granting algorithm with explicit
reservations (cf. Figure 1).
During the coupling phase the client asks the suitable services for
reservations. Reservations are binding for a short duration. If all
services positively respond, the client submits lease-requests that
yield to valid usage leases.
If at least one service cannot satisfy the reservation request, the
client cancels all other reservations.
After the coupling phase is completed, the interaction of coupled
components starts. The client process configures and initializes the
services and finally starts production (cf. Figure 2).
When the leases are about to expire, the client either issues a
prolongation request to extend the production phase or stops the
services and performs cleanup operations.
The prolongation of existing leases uses the same 2-phase
algorithm as used at initial lease creation. In the decoupling phase the
expired leases are fairly released and deleted.


3. APPLICATION EXAMPLE

The service-oriented control software presented so far was
experientially evaluated for an example industrial automation setup.
The example system and the tested applications scenarios are
presented in this section.


3.1. EXAMPLE STRUCTURE

The structure of our evaluation example is depicted in Figure 6.
The work pieces enter the system through conveyer conv1 and conv2.
C. Voican
Both conveyers are located next to a rotary disk, which is able to
collect work pieces from either conv1 or conv2 by rotating the disk
and using the conveyer element conv3 on top of the disk.
This conveyer transports the work pieces to conveyer conv4 which
in turn moves them through the lacquer machine.
After being painted by the lacquer machine, the work pieces are
checked by a laser sensor.

Inaccurate pieces are pushed into a disposal box by a pusher.
Proper items are moved out of the system to the next work station.
The devices and sensors (not depicted) are exporting services as
described in section.
The logical control of the conveyers is implemented using a PID
controlling algorithm which could be differently parameterized for
evaluation purposes.
Figure 1. Lease lifecycle

Figure 2. Leases and production


3.2. APPLICATION SCENARIOS

The example system was evaluated using different application
scenarios. The scenarios use different service hierarchies and thus
model different levels of control in the application process.
The first scenario comprises the following process:
1. Work pieces are picked up from conv1 or conv2.
153

2. The rotary disk and conv3 transport the work pieces to conv4.
3. The lacquer machine paints the work pieces.
4. The inaccurate work pieces are detected and pushed into the
disposal box.
5. The acceptable work pieces are moved out of the system.
The service hierarchy for this application process is depicted in
Figure 4. The application process uses six different control services
(light gray), each responsible for a specific part of the example
system. The control services themselves are using a set of sensor and
actuator service interfaces to interact with the hardware at technical
process level (dark grey). In contrast, the rotdisk control service for
controlling the rotary disk and conv3 on top of the disk as a whole
uses the control services of the single components. It implements an
algorithm for the balanced use of the two attached input conveyers.
The second application scenario uses only the input conveyer
conv1, thus the usage of the rotdisk control service is not necessary:
0. Statically move the rotary disk in conv1-conv4 position using
the disk service.
1. Conv1 transports the work pieces to conv3.
2. Conv3 forwards the work pieces to conv4.
3. Conveyer conv4 moves the pieces through the lacquer machine.
4. The pusher sorts out erroneous pieces.
5. Acceptable items leave the system.
The service hierarchy used for the second scenario is depicted in
Figure 5.
The application process of the second scenario uses seven control
services. The sub-component services of the rotary disk now are
directly used to initially set up the right direction of the disk and to
control the conv3 at runtime. This change in the process outline does
not infer changes in the service implementations of the devices used.
Further scenarios were used to evaluate the applicability of
multiple application processes, each controlling a part of the overall
process.


4. EVALUATION

The evaluation environment comprises three major components:
the DPWS stack, the Java Real-time VM and the simulation system.
C. Voican
The WS4D.org DPWS stack, developed by Dortmund University
and Materna, is a Java based implementation of the DPWS protocol
stack and provides a service oriented communication infrastructure.
It was developed with modularity and extensibility in mind and
thus can be adapted to varying application scenarios, ranging from
small client-only implementations for mobile phones to multimedia
or file-sharing services for embedded settop boxes.
The Java Real-time System comprises technologies and concepts
for correct reasoning about the timing of Java real-time applications.
It contains new types of real-time threads, memory handling schemes
preventing the garbage collector from influencing the runtime
behavior in a nondeterministic way), high precision timers with
nanosecond resolution and direct memory access for implementing
device drivers purely in Java. Nevertheless, the Java RTS depends on
the real-time capabilities of the underlying operating system.
Figure 3. Example system


For evaluation purposes we developed a testing environment, split
into two blocks: a simulation system and the sensor, actuator and
control service implementations.
The time discrete simulation system is composed of four major
components.
The simulation model component manages a grid model for
locating devices, sensors and work pieces in the system and a
155

component model for preserving the state of the simulated
components.
The simulation control component periodically updates the model
information.
Figure 4. Scenario 1 service hierarchy


Changes in the internal state of sensors and actuators are sent to
and received from the distributed components via an UDP based
communication protocol. It was especially designed to consume few
network bandwidth.
A graphical user interface is used to track and control the
simulation.
The simulated system comprises sensor, actuator and control
service implementations. The sensor and actuator implementations
are connected to the simulation system via the UDP based
communication protocol (s.a.) to receive and publish state
information.
The simulations were run on an Athlon64 X2-3800 machine with
two GB of memory and an OpenSolaris installation as basis for the
Java RTS.
C. Voican


Figure 5. Scenario 2 service hierarchy


4.1. EXPERIMENTAL RESULTS

A series of experiments focused on the evaluation of the functional
behavior of the control system. Particular test sequences checked the
feasibility and stability of the lease-based allocation. Atomic
allocation and setup of service ensembles were as well tested as
atomic lease prolongation and occasional aborts followed by the
searching and switching to alternative ensembles.
In the course of additional experiments the service call roundtrip
times (using simple input and output parameters) were measured in
order to check the current real-time limits of Java VM and DPWS
based control system implementations.
Table 1 presents the values obtained for local VM-internal (on the
OpenSolaris host) and for remote DPWS-based service calls (between
the OpenSolaris and the PC host).
The configuration was able to support low to medium realtime
requirements (e.g. cycle times >50ms).


5. CONCLUSIONS

We have presented a service-oriented control architecture for
automation systems. The architecture forms a service hierarchy
ranging from low-level sensor and actuator services, over a number of
control service levels up to application processes. Instead of
statically associating services for the different client operations, a
flexible lease based binding approach is used.
157

Table 1. Action call roundtrip times


This approach follows the loosely coupled nature of components
in service-oriented architectures. The algorithm used for the flexible
binding approach was tested in different application scenarios.
The evaluation results regarding action call roundtrip time exhibit
that the Java-based service-oriented approach may not yet be a
feasible solution for all applications. However, the applicability can
be extended by using e.g. hardwarebased message processing and
real-time capable network infrastructures .


6. REFERENCES

[1] H. Smit, F. Jammes, Service-Oriented Paradigms in Industrial
Automation, IEEE Transactions on Industrial Informatics, Vol. 1,
No. 1, pp. 62-70, 2005.
[2] C. Gray, D. Cheriton, Leases: an efficient fault-tolerant
mechanism for distributed file cache consistency, ACM SIGOPS
Operating Systems Review, Vol. 23, Issue 5, pp. 202-210, Dec. 1989.
[3] Universal Plug and Play (UPnP), http://www.upnp.org, 1999.
[4] Devices Profile for Web Services (DPWS),
http://schemas.xmlsoap.org/ws/2006/02/devprof/, 2006.
[5] Service Infrastructure for Real-time Embedded Networked
Applications (SIRENA), http://www.sirena-itea.org, 2006.
[6] Sun Microsystems, Jini, Network Technology,
http://www.sun.com/software/jini, 1999.
[7] Kapsers, Kfner, Messen Steuern Regeln: Elemente der
Automatisierungstechnik, Vieweg Verlag, 6th Edition, p. 253, 2006.
[5] Service Infrastructure for Real-time Embedded Networked
Applications (SIRENA), http://www.sirena-itea.org, 2006.
C. Voican
[8] WS4D.org Java Multi Edition DPWS Stack,
http://www.ws4d.org, 2007.
[9] Sun Java Real-time System 2.0 (Java RTS),
http://java.sun.com/javase/technologies/realtime, 2007.
[10] PROFINET, http://www.profibus.com/pn/, 2007.

159

NUCLEAR MAGNETIC RESONANCE
PROFILING OF HIGH-STRENGTH CONCRETE
SAMPLES

Marius Simina
1
, Ioan Ardelean
1


1
Technical University of Cluj-Napoca, Physics Department, 400114, Cluj-Napoca, Romania
Email adress: Marius.Simina@phys.utcluj.ro


Abstract: The present work proposes an indirect way of identifying and
monitoring the evolution of inter C-S-H and capillary pores from the surface layer of
some high-strength concrete samples with different water to cement ratios as a
function of depth. The method is based on measuring the nuclear magnetic
resonance spin-spin relaxation of the water confined inside the porous matrix of the
cementitious material. The concrete samples were prepared using a mixture of
cement (CEM I 52.5 R), quartz sand, sillica fume (Elkem Microsillica), water and
superplasticizer (Glenium ACE 30 - BASF). All experiments were performed in low
fields using a single-sided NMR-MOUSE (MObile Universal Surface Explorer)
operating at the protons resonance frequency of 11.7 MHz. The NMR profiles were
acquired for each sample with a maximum depth of 2.4 mm, and the evolution of the
moisture content versus depth was extracted. The relaxation time distributions were
obtained from the echo decays using a regularized numerical Laplace inversion
algorithm. The results show a decrease of the relaxation time with the increase of the
measuring depth, proving that the moisture content decays along with the measuring
depth. The method can easily be applied for nondestructive in situ monitoring of the
curing, setting and hardening processes of concrete structures.
Key words: NMR relaxometry, high-strength concrete, single sided NMR
M. Simina, I. Ardelean

1. INTRODUCTION
Owing to its completely non-destructive character, nuclear magnetic
resonance (NMR) is a powerful tool for investigating high-strength concrete
both with respect to its structure and the dynamics of molecules (especially
water) under confinement conditions. NMR measurements of diffusion
coefficients [1] and relaxation times [2] render quantitative data on the
dynamics of confined molecules and the restrictions of the confinement
imposes on their translational and rotational mobility. NMR relaxation
experiments are suitable in studying liquids (water) in porous media mainly
due to an observed enhanced relaxation rate under confinement conditions
[3, 4]. This behavior is generated by the interactions of the probed molecule
with the surface depending on the porous sample composition, morphology,
the nature of the filling fluid and its distribution inside pores [3, 4].
Consequently, relaxation time measurements may provide information about
molecules location inside porous media and were comprehensively discussed
in the NMR logging applications [5].
Due to its porous structure and the relative high water content, concrete
makes an excellent candidate for low-field NMR spin-spin relaxometry
investigations. Many NMR studies have been performed on concrete
samples including diffusion and relaxation during hydration [6],
microstructure determination [7] or micro-pore size analysis [8, 9] of the
hydrated cement paste.
High-strength concrete has usually a very compact structure and low
porosity and permeability due to the lack of coarse aggregates, the addition
of silica fume or fly ash for enhancing the structural density and due to the
low water-to-cement ratio generally obtained with the use of a super-
plasticizer (water reducing agent). Although the structure gains high-
compressive strength values after following all these steps, the composite
material becomes less ductile and more prone to shrinkage. Often the
defective curing treatment and exposure conditions may lead to the
appearance of micro-cracks in the internal structure interconnecting the
pores and leaving the material exposed to many moisture related deleterious
processes.
Concrete structures are generally expected to display a high durability and
resistance to chemical, physical and mechanical deleterious phenomena. One
of the governing factors of concrete durability is permeability which also
finds itself controlled by the pore structure and connectivity [10]. The pore
types of concrete can be identified by NMR methods using the water
reservoirs present in the structure of the material [11]. Once one has a
pertinent image of the pore structure of a concrete material it will be easier
to predict and to avoid the damage induced by the moisture transport. That is
why this study investigates the pore-size distribution of some high-strength
161

concrete samples in order to understand and characterize this type of
composite cementitious materials.
2. METHODS
Three high-strength concrete samples with different water-to-cement
ratios (0.2, 0.3 and 0.4) were prepared using a mixture of cement CEM I
52.5 R (Carpatcement Heidelberg Cement Group), quartz sand (particle
diameter < 0.31 mm), silica fume (Elkem Microsilica), water and super-
plasticizer (Glenium ACE 30 BASF) following the exact mass-ratios
presented in Table 1. The concrete mixture was produced following the
structural model of non-fibered RPC 200 [12], but using materials available
on the local market.

Table 1. Typical high-strength concrete composition (by weight)
Materials W/C 0.2 W/C 03 W/C 0.4
Cement CEM I
52.5 R
1 1 1
Silica fume

0.25 0.25 0.25
Quartz sand

1.1 1.1 1.1
Superplasticizer

0.026 0.026 0.026
Water

0.2 0.3 0.4

The samples were poured-in-place in polystyrene Petri dishes of 65 mm
diameter and kept at a relative temperature of 25
0
C. After one day the
shrinkage of the concrete allowed the easily striking of the samples from
the Petri dishes and they were immersed in distilled water until they were
investigated with the NMR profiler after 28 days of hydration. After 28 days
the surface of the concrete samples was dried-up and the individual samples
were placed back into the Petri dish in order to diminish the water
evaporation and laid-down on a single-sided NMR scanner (NMR Mouse
[13]) with the sensitive sensor underneath the centre of the Petri dish. The
device has an accessible vertical range of 2.5 mm and is operated at 11.7
MHz proton resonance frequency.
The Carr Purcell Meiboom Gill (CPMG) [14] pulse sequence was
applied to acquire the depth profiles and the echo decay trains for each
sample. The placement of the sensitive volume along the surface of the
sample in order to measure the profiles is done owing to a powered high-
precision elevation/dropping system. For all the samples the position of the
M. Simina, I. Ardelean

sensor was moved from the surface of the concrete sample to a maximum
depth of 2.4 mm in twelve steps of 200 m with a slice thickness of 200 m.













Figure 1. The NMR profiles of the high-strength concrete with different water-to-
cement ratios versus the depth inside the sample


Sixty-four echoes were accumulated for each CPMG echo-train, with an
echo-time of 56.5 s and a repetition time of 50 ms, 4096 scans were used to
obtain a better signal-to-noise ratio. Each point of the profile (Figure 1)
represents the sum of the area of the second to fifth echoes in the CPMG
train in order to reduce the influence of the transversal magnetization decay
constant T
2
in the determination of the NMR signal intensity [15]. For the
sample with a water-to-cement (w/c) ratio of 0.4, we observed an
enhancement of the signal. Consequently, the CPMG echo decays were
measured using a longer echo-time of 87.5 s and a smaller number of
echoes (only 42). The contribution of the diffusion in the overall signal
intensity was estimated as being negligible.
One can observe multi-exponential decays of the CPMG echo train
envelopes for each sample indicating the heterogeneity of the sample under
study. In order to extract the relaxation time distributions, the experimental
data in Figure 2 have been analyzed numerically using the inverse Laplace
transform based on the regularized CONTIN algorithm [16]. The results are
depicted in Figure 3 for all three water-to-cement ratios. Note however that
the numerical Laplace inversion is ill-conditioned. That is, for a given
CPMG decay envelope, there are a multitude of possible combinations of
characteristic times which fit well the solution. This problem appears
especially in the case of measurements which contain noise in the measured
data and may lead to large perturbations in the resulting relaxation time
distributions. Therefore the numerical inverse Laplace transforms must be
used with precaution when interpreting the experimental results.
163

Figure 2. The echo amplitude versus echo time in a CPMG relaxation experiment on
high-strength concrete samples with different water-to-cement ratios: w/c=0.2 (a),
w/c=0.3 (b), w/c=0.4 (c)
3. RESULTS AND DISCUSSION
The NMR moisture profiles (Figure 1) reveal a decrease in the signal
amplitude with the penetration depth. For the samples with a w/c ratio of 0.2
or 0.3 respectively, the decrease is more severe until the depth of 1mm. After
this range the slope is diminishing until 2 mm depth wherefrom the
amplitude of the signal is almost constant. The sample with a w/c ratio of 0.4
follows the same behavior as the other two for depths bigger than 1 mm, but
in the range of 0.2 1 mm the NMR signal amplitude profile exhibits a peak
indicating a higher moisture content at a depth of 0.6 mm.
M. Simina, I. Ardelean


Figure 3. The relaxation time distributions of water molecules confined inside the
high-strength concrete samples with different water-to-cement ratios: w/c=0.2 (a),
w/c=0.3 (b), w/c=0.4 (c)


The CPMG echo decays depicted in Figure 2 show an enhancement of the
signal-to-noise ratio of the samples with the increase of the water-to-cement
ratio. This enhancement starts from the more scattered data in Fig. 2a
(w/c=0.2) where the structure of the material is more dense and the porosity
is lower and evolves to the much more accurate echo trains corresponding to
the w/c=0.4 in Fig. 2c for whom the porosity and implicitly the NMR signal
intensities are higher. Also analyzing the CPMG echo decays one can notice
that all three w/c ratios exhibit a double exponential behavior indicating the
contribution of two different relaxation components. This behavior is
obvious when the data are analyzed using the regularized Laplace inversion
algorithm CONTIN [16]. Hence on the relaxation time distributions in
Figure 3 the presence of two peaks is noticed, each corresponding to a
distinct water configuration inside the sample. The shorter component, with
T
2
values ranging between 0.1 0.35 ms is assigned to the gel pore water
[11], the longer component with values ranging between 0.3 3 ms is
attributed to the capillary pore region [11]. Any longer components can
derive from water in cracks or adsorbed on the surface of the sample or, it
can be an artifact induced by the inverse Laplace transform. Therefore in this
study the analysis has focused only on the evolution and the behavior of the
shortest components, corresponding to the gel and capillary pores and the
165

long relaxation times representing the surface adsorbed water contributions
were neglected.
By exploring the relaxation time distributions in Figure 3 one can observe
that the position of the first peak has a more dynamic evolution for the
sample with a 0.2 w/c ratio than for the other two. For this sample (w/c=0.2)
the peak position (Fig. 3a) is not just decreasing in amplitude but it has a
slight and clear shift to the right. This indicates that the pore dimension
decreases with the penetration depth. For the other two samples the
evolution is similar but less obvious.
Figure 4 shows the evolution of the transverse relaxation times (T
2
) of the
two components capillary (Fig.4a) and gel (Fig.4b) versus the measuring
depth. The values were extracted from the relaxation time distributions in
Figure 3. As can be observed the behavior of the gel T
2
values (Fig. 4b) are
similar to that of the NMR profiles indicated in Figure 1 offering a clear
distinction between the different water to cement ratios. The slowly decrease
of the T
2
value with the depth indicates that the gel porosity decays with the
penetration depth. On the other hand, the capillary water evolution presented
in (Fig. 4a) has a much unpredictable behavior and one can easily notice that
a clear separation between samples is hard to be done. The scattered
evolution of the capillary water can be explained by the appearance of
micro-cracks in the structure of the concrete sample which will perturb the
pore system and give rise to the peaks with values larger than 3 ms one can
notice in (Fig. 3 a). Those cracks may appear as a result of defective curing
or can be induced by the endogenous shrinkage and the self-desiccation
which are common phenomena in the high-strength concrete [17].


Figure 4. The evolution of the capillary (a) and gel (b) pores transverse relaxation
time versus depth as extracted from the Laplace inversions in Fig.3.

M. Simina, I. Ardelean

4. CONCLUSIONS
In the present work it is proved that the low-field NMR relaxometry
profiling technique can be successfully used as a tool in diagnostic of high-
strength concrete samples or other cement based materials. Using a
numerical Laplace inversion it was possible to extract information on liquid
distribution inside partially saturated high-strength concrete samples at
different depths. Two peaks were found in the relaxation time distribution at
each depth independently of the water-to-cement ratio. These two peaks
could be attributed to different locations of the water molecules: the gel pore
water and the capillary water. The monitoring of the transverse relaxation
time versus penetration depth for three water-to-cement ratios (0.2, 0.3 and
0.4) indicates a decrease of the gel pores with the penetration depth for the
samples with 0.2 and 0.3 W/C ratio. The results presented here may
contribute to a better understanding of the deleterious processes taking place
in concrete under the influence of the water transport inside the porous
structure of the composite cementitious materials.
It is also important to note that the NMR MOUSE profiling technique
presented here can be used as a completely nondestructive tool for
investigating the cement based materials in situ. Thus it can provide a quick
and precise moisture content monitoring or even the crack formation
analysis at different depths inside concrete walls.


ACKNOWLEDGMENTS
We acknowledge Prof. S. Stapf for allowing us to use the NMR surface
explorer from TU Ilmenau, Germany. Financial support by the Romanian
National Authority for Scientific Research, CNCS UEFISCDI, project
number PN-II-ID-PCE-2011-3-0238 and by the European Social Fund
(POSDRU/88/1.5/S/60078) is also gratefully acknowledged.
REFERENCES
[1] I. Ardelean, R. Kimmich. Annual Reports on NMR Spectroscopy, 49,
43 115, 2003.
[2] R. Kimmich, E. Anoardo. Progress in NMR Spectroscopy, 44, 257
320, 2004.
[3] S. Bhattacharja, F.DOrazio, J.C. Tarczon, W.P. Halperin, R.
Gerhardt. J. Am. Ceram. Soc. 72, 2126-2130, 1989.
[4] F. DOrazio, S. Bhattacharja, W. P. Halperin. K. Eguchi, T. Mizusaki,
Phys. Rev. B, 42, 9810 9818, 1990.
167

[5] G. R. Coates, L. Xiao, M. G. Prammer. NMR logging: principles and
applications, Haliburton Energy Services, Houston, 1999.
[6] K.Friedemann, F. Stallmach, J. Krger. Cement and Concrete
Research 36(5), 817-826, 2006
[7] S. Bhattacharja, M. Moukwa, F. Dorazio, J. Y. Jheng, W.P.
Halperin, Adv. Cem. Bas. Mat 1:67-76, 1993.
[8] A. Plassais, M. P. Pomis, N. Lequeux, P. Boch, J. P. Korb. Magn.
Reson. Imag. 19, 493 495, 2001,
[9] A. Plassais, M. P. Pomis, M.-P. N. Lequeux, P. Boch, J.-P. Korb, D.
Petit, F. Barberon, Magn. Reson. Imag. 23, 369-371, 2003.
[10] I.O. Yaman, N. Hearn, H.M. Aktan. Mater. And Struct. 35,102-109,
2002.
[11] P.J. McDonald, V. Rodin, A. Valori. Cement and Concrete Research
40, 1656 1663, 2010.
[12] P. Richard, M. Cheyrezy. Composition of Reactive Powder
Concretes, Cem.Concr. Res., Vol.25, 1501 1511, 1995.
[13] B. Blmich, J. Perlo, F. Casanova. Mobile single-sided NMR,
Progress in Nuclear Magnetic Resonance Spectroscopy 52, 197 269, 2008.
[14] S. Meiboom, D. Gill. Rev. Sci. Instr. 29, 688 691, 1958.
[15] S. Ghoshal, C. Mattea, P. Denner, S. Stapf. Heterogeneities in gelatin
film formation using single-sided NMR, J. Phys. Chem. B. 114, 16356
16363, 2010.
[16] S. W. Provencher, Comp. Phys. Comm. 27, 229 242, 1982.
[17] F. H. Wittmann. Crack formation and fracture energy of normal and
high strength concrete, Sdhan 27, 413 423, 2002.

169

Residual stresses and fatigue behavior of hybrid butt
welded joints

C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi
Politecnico di Bari, Dipartimento di Ingegneria Meccanica e Gestionale,Viale Japigia, 182
70126 Bari
casavola@poliba.it

Abstract: Ti6Al4V alloy combines mechanical strength, deformability, excellent
fatigue and corrosion resistance and high strength to weight ratio. Furthermore, the
mechanical behavior remains excellent at high temperature. Such characteristics
make this material attractive for numerous applications (structural, aerospace and
naval) because of recent improvements in welding techniques (laser, hybrid
laser/MIG) that allow realizing high quality titanium welded joints. However some
problems related to the welding, as deteriorated material properties, residual stresses
and distortions, need further investigations. Residual stresses can have a significant
influence on the fatigue life of structural engineering components. For the accurate
assessment of fatigue life a detailed knowledge of the residual stress profile is
required. This paper presents a study on residual stresses of Ti6Al4V butt plates
welded by hybrid laser/MIG process. Residual stresses were measured by hole
drilling method using electrical strain gage rosettes bonded at different position, in
order to evaluate magnitude and distribution of residual stresses along the cord.
Residual stresses curves obtained are presented and discussed in terms of transversal
and longitudinal residual stresses. Residual stresses were also measured at surface
by means of X-ray diffractometer. Besides residual stress measurements, this paper
presents experimental fatigue results on titanium grade 5 butt welded joints. The
fatigue curves expressed in terms of nominal amplitude of stress (according to the
traditional global method) and in terms of local amplitude of strain (according to the
local Wel.Fa.Re. method) have been plotted after the experimental tests. Electrical
strain gauges bonded close to the weld toe have been used to calculate the local
strain amplitude. The aim of this work is to present experimental data related to
welding processes in order to confirm the validity of currently procedures or
improve them.
C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi

Key words: Residual stress, hybrid welding, fatigue performance, Ti6Al4V, butt joints
1. INTRODUCTION
Titanium and its alloys are widely applied because of their favorable
strength to weight ratio and their resistance to corrosion in oxidizing
environments [1, 2]. The high corrosion resistance is caused by passivation
due to a thin oxide layer that naturally cover the surface. Titanium alloys
welded joints fabricated for several applications such as automotive,
chemical implants, aerospace, military vehicles, are often subjected to
fluctuating loads. This kind of loading causes small cracks to grow during
life of the component and leads to fatigue failure. A detailed study of fatigue
behaviour of the material could prevent unexpected failures. Welding of
titanium alloy is complicated due to the high reactivity of this material with
atmospheric gases such as oxygen, nitrogen and hydrogen causing severe
embrittlement of the joint [3]. Because of its high level of specific power and
the limited area involved, laser technology appears to be a promising
solution. The laser beam sources of the new generation are capable of
producing deep and narrow seams. However, the laser welding process leads
to a non-uniform temperature distribution and associated thermal strains,
which generate residual stresses due to melting and subsequent cooling of
the weld and surrounding material [4]. Eboo [5] showed that an addition of
electric arc to the laser beam used for welding and cutting could decrease the
needed power of laser with respect to the case when only laser is used for
welding. This technique combines the best characteristics of both laser and
arc welding processes, acting simultaneously in the same process zone [6].
The advantages of the hybrid welding technique compared to pure laser
welding or arc-welding are well known and include an increase in (a) the
welding speed, (b) the weldable material thickness, (c) the gap bridging
ability, and (d) the welding process stability and efficiency [7]. In particular,
laser-arc hybrid welding offers many advantages for heavy industrial
applications involving thick-walled materials as it enables full penetration
weld of thick plates without the need of multiple passes and at that reduces
welding after works such as cutting for adjustment and fairing at the
assembly stage. Hybrid welding improves productivity by two to four times
as compared to the conventional arc welding [6]. In addition, in the fusion
zone of hybrid laser-arc welding, equiaxed grains exist, whose sizes are
smaller than that obtained from arc welding but larger than those from laser
beam welding. Welding process unavoidably involves a stage of residual
stress in the welded structure that could lead to some problems in terms of
dimensional stability and structural integrity. Because of the variety of
factors involved in welded parts, it is very difficult to predict and control
171

residual stresses caused by welding. Residual stresses can have a significant
influence on the fatigue lives of engineering components [8]. In particular,
near surface tensile residual stresses tend to accelerate the initiation and
growth phases of the fatigue process while compressive residual stresses
close to a surface may prolong fatigue life. Significant advances have been
made in recent years for obtaining accurate and reliable determinations of
residual stress distributions.
Fatigue strength of welded joints is lower than base material because of
microstructure modifications, misalignment, geometrical discontinuity of the
cord, residual stresses and weld defects. In order to take into account all the
variables affecting the fatigue strength, the fatigue design of welded
components is done on the basis of experimental results, using fatigue curves
available in codes [9].
The fatigue behavior of a ! + " Ti6Al4V alloy butt welded by hybrid
technique has been investigated in this paper. The experimental results have
been compared with a previous study on similar welded joints, but obtained
with different process parameters [10-11]. Fatigue fracture surface have been
observed by scanning electron microscope. Besides in this paper
experimental hole drilling method and X-ray diffraction measurements have
been reported.
2. MATERIAL AND METHODS
The material used in the tests was titanium grade 5 (Ti6Al4V), an !#"
alloy with aluminium that is an !-stabilizing element and vanadium that is a
"-stabilizing element. Ti6Al4V has good mechanical properties both at room
and high temperature. Its density is half steel density (4.4 kg/dm
3
[2]), the
ultimate strength is 980 MPa and the yield strength is 760 MPa. Titanium
grade 5 mechanical properties and chemical composition are indicated in
tables 1 and 2.

Table 1: Mechanical properties of grade 5 titanium alloy [12]
Yield strength
(MPa)
Ultimate strength
(MPa)
Youngs Modulus
(MPa)
Elongation at
fracture (%)
760 980 110200 14

Table 2: Chemical composition of titanium grade 5 according to ASTM B265 [2]
Element C Al V Fe H N O
% < 0.08 5,5 6.75 3,5 4.5 < 0.30 < 0.015 < 0.03 < 0.25

At room temperature, unalloyed titanium has a hexagonal close-packed
(hcp) crystal structure called !-phase. At 883 C this transforms to a body-
C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi

centered cubic (bcc) structure called "-phase. The manipulation of these
crystallographic forms with alloy elements and thermo-mechanical processes
is the base for the developing of a wide variety of alloys with different
properties.
The weld cord is smooth and regular, but it has a golden coloration which
indicates a partial oxidation of the cord surface. A filler material with the
same chemical composition of the plates has been used during the joining
process.
In welded structures, the residual stress values vary from one point to
another and the measurement needs to be carried out in different positions.
The measurement is usually executed on panels designed specifically for this
purpose. The most common specimen, if not explicitly specified otherwise,
is a rectangular, butt-welded panel. Since the panels are being designed for
testing purposes only, even a destructive test can be considered if this
facilitates the measurement considerably [13].
The HDM is the most widely used technique for residual stress
measurement. The principle involves introduction of a small hole at the
location where residual stress is measured. Due to drilling of the hole,
residual stresses are relieved and the corresponding strains on the surface are
measured using strain gauges bonded around the hole. From these measured
strains, residual stresses are calculated using appropriate calibration
constants derived for the particular type of strain gauge rosette used. For
HDM measurements type B strain gage rosettes are bonded close to the weld
cord, with the hole at 2 mm from the weld line. The drilling was carried out
in a series of small steps (20 steps in 1 mm). A high-speed air turbine and
carbide cutters were used to drill the hole without introducing any further
machining stresses and thereby modifying the existing stress system. The
strain data at pre-determined depths were precisely acquired. Different stress
calculation methods are used to arrive at the residual stress system from the
measured strains. In this case, the integral method with Tikhonov
regularization of H-DRILL software has been used to measure the variation
of the residual stresses through the thickness.
The XRD method is the most common non-destructive method to
determine the surface residual stress in crystalline materials. XRD is
accurate, but it is limited by the fact that only information is obtained about a
relatively thin surface layer. The XRD-sin2! technique was developed from
the theories of crystallography and solid mechanics. Given the limited
penetration of X-rays in solid surfaces, what the XRD-sin2! technique
measures is the surface residual stress in a depth of up to a few micrometers.
For XRD measurements, XSTRESS 3000 equipment [14] and the sin2!
method have been utilized; TiK" radiation (# = 2.2909 ) is used as X-ray
source. XRD residual stress measurement has been performed on the
transversal direction to the weld cord of butt welded plates. The collimator
with 1 mm illuminated spot size has been used and the exposure time has
173

been set at 90 sec to ensure the adequate intensity. The operation voltage and
current were 30 kV and 6.7 mA. The 110-diffraction of titanium with the
diffraction angle of 137.4 has been analyzed. During the stress
measurement, ! (the angle between the diffraction vector and the normal to
the surface) of 0, 19.3, 27.9 and 35 have been selected to plot d
(interplanar spacing) sin2 ! curves on both positive and negative ! range.
Considering the texture effect on the stress measurement, a ! oscillation
with 3 has been used during the measurement. Indeed, to give reliable
values of residual stress, measurements have been replicated on the same
points.
Fatigue specimens have rectangular geometry, with the cord positioned in
the middle. Experimental tests have been carried out on a RUMUL Vibro-
Forte resonant machine 500 kN capacity. The load applied has a constant
amplitude and load ratio equal to 0.1 and the work frequency is 60 Hz. The
tensile fatigue load has been applied along the direction normal to the weld
cord. Before the execution of the test two electrical strain gauges have been
bonded on each side of the cord in order to measure strain values and to
calculate the local amplitude of strain. 1.5 mm grid length strain gauges have
been used in this study and they have been bonded with their transversal
axes at 1 mm from the weld toe.
The Wel.Fa.Re. method utilises the local strain amplitude $
a
measured at
the weld toe [15-21] because it presumes that this parameter can include the
effects of all variable influencing the fatigue life of welded joints. The local
strain amplitude is measured before the fatigue test applying a static nominal
load amplitude. Consequently, with the execution of the fatigue tests it is
possible to obtain the fatigue life curve expressed in terms of local strain
amplitude $
a
:
!
!
!
!
!"#
! !
!"#
!


and in terms of nominal stress amplitude %
a
:

!
!
!
!
!"#
! !
!"#
!


Four electrical strain gages have been bonded on the specimens according
to the Wel.Fa.Re. guidelines. The value of local strain has been measured at
the weld toe. System 5000 by Micro Measurements Inc. (USA) has
registered strain gages measured values.
There are several approaches to study the fatigue behaviour of welded
joints. The global methods and FEM analysis can characterize the overall
strength of the joints, but their models are often too simple and do not
consider important aspects such as material inhomogeneity due to changes in
microstructure, residual stresses due to temperature gradients and local stress
C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi

concentrations. Recently, Eurocode Standard [22, 23] has introduced the Hot
Spot method, but only for steel and aluminium alloys. The correct evaluation
of the fatigue strength for complex geometries remains a difficult task and in
some cases only the local methods based on experimental measurements can
correctly interpret all the factors that affect the strength of welded structures.
3. TEST PLAN
Geometry of welded plates and locations of residual stress measurements
are shown in Figure 1 and Figure 2. Residual stress measurements have been
carried out by means of HDM and XRD on the welded plate before the
cutting of specimens for fatigue tests. The thickness of the welded plate was
3 mm; six specimens 40 mm width and 400 mm long have been cut from the
butt welded plate using the milling cutting procedure. Outer sides of plates
have been removed because of defects and welding irregularities.


Figure 1. Geometry of welded plates and location of drilled holes in HDM


Figure 2 Geometry of welded plates and location of XRD measurements

Table 4 shows the plan of fatigue tests.
Before the execution of the tests the ! angle (figure 3) has been measured
in order to evaluate the secondary bending effect on fatigue tests. The
distortion of the plates is caused by the welding thermal cycle. The heating
175

and subsequent cooling of a narrow area and the gradient of temperature
imposed to the plate are the causes of distortions and residual stresses in the
welded joints.

Table 4: Fatigue test plan
Joint
geometry
Plate
thickness
[mm]
Number of
specimens
Welding technique
Butt 3 6 Hybrid (Laser + MIG)



Figure 3.Distortion angle on the specimen after the weld process.

Table 5 reports the values of the distortion ! angle measured on the
specimens obtained from the four welded plates. The ! angle has a very low
value with a mean value of 0.23.

Table 5: Distortion angle ! measured on the test specimens.
Specimen Distortion angle ! []
A1 0.12
A2 0.06
A3 0.20
A4 0.32
A5 0.19
A6 0.49
mean value 0.23
standard deviation 0.15
4. RESULTS AND DISCUSSION
Figure 4 and Figure 5 show HDM calculated residual stress respectively
in the transversal and longitudinal directions with respect to the weld cord of
the hybrid butt welded plate.

C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi


Figure 4. Calculated residual stresses transversal to the cord by HDM


Figure 5. Calculated residual stresses longitudinal to the cord by HDM

Stresses are plotted against the hole depth h. Plotted residual stresses are
calculated by Integral method. Tensile stresses are produced in either
direction and it was found that the longitudinal residual stresses are higher
than transversal residual stresses. These characteristics agree well with the
results reported in the literature (Figure 6). A welding process generates a
significant residual stress in the longitudinal (welding) direction, $
y
, that
generally has a peak tensile stress in the center of the weld bead and a peak
compressive stress in lateral zones (Figure 6). At the same time, stress in the
transversal direction is present, $
x
; this reaches lower values with respect to
the stresses acting in the longitudinal direction (Figure 6). If the lateral
contraction of the joint is restrained by an external constraint such as a series
177

of springs, tensile stresses approximately uniform along the weld are added
as the reaction stress. An external constraint, however, has little influence on
the distribution of longitudinal residual stresses. Regarding possible
symmetries, a panel designed conveniently for measuring residual stress will
have a double symmetric geometry (xz and yz symmetry planes in Figure 6).
This, however, does not imply a complete symmetry of temperature field
during the welding process. The joint in fact is not realized simultaneously
along its length and differences can be found between the areas at the
beginning and at the end of the joint. Therefore some strange values,
measured longitudinal to the cord at the hole 1 and hole 2 (Figure 5)
probably are due to this phenomenon.


Figure 6. Typical distributions of residual stresses in butt welded plate [13]

Figure 7 shows values of residual stress normal to the weld cord of the
hybrid butt-welded plate in the case of XRD measurements. In this case,
locations of measurements have been chosen in the center of the plate, at
progressive distance from the weld cord. XRD values are commonly referred
to the surface of specimen. Figure 7 also shows residual stress level
measured by HDM in the direction normal to the cord at intermediate
location between the first two points of XRD measurements near the surface
(0.033 mm depth) in case of hybrid butt plates. Residual stress measured by
means of HDM gives the average stress related to the area (the hole diameter
range between 1.8 and 2.0 mm) where the material is drilled (the area
corresponding to the hole drilled by HDM is marked in Figure 7). It can be
observed that the XRD measurements no. 1 and 2 are localized at 0.5 and 3
mm from the weld line, while HDM hole cover the area between 1 and 3 mm
from the cord, due to the inherent limitation laying in the size of strain gage
C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi

rosette. Residual stress levels given by HDM (about 60 MPa corresponding
to 0.033 mm depth) seem to be an intermediate value.


Figure 7. Calculated residual stresses normal to the cord on hybrid butt welded plate
by XRD

The results of the experimental fatigue tests are reported in table 6.

Table 6: Experimental results.
Specimen
Nominal Stress
Amplitude %
a

[N/mm
2
]
Local Strain
Amplitude $
a

[$]
Cycles
to Failure
N
Final Crack
Location
A1 80.6 700 151390 Weld cord
A2 85.6 800 142497 Weld cord
A3 60.5 512 589342 Weld cord
A4 50.4 487 2181159 Weld defect
A5 55.0 519 1533205 Weld defect
A6 52.5 539 1313911 Weld defect

It should be noted that the fatigue failure occurs within the weld cord, or
near weld defects such as drops of fused material when they are presents.
Similar results have been obtained in previous works [10, 11, 24] and they
confirm the susceptibility of titanium to this kind of defects.
The fatigue strength of titanium grade 5 welded joints is also strictly
dependent to the joining process parameters. Figures 8 and 9 show the
fatigue curves, for a load ratio of 0.1, expressed in terms of nominal
amplitude of stress %
a
and local amplitude of strain $
a
, according to the
Wel.Fa.Re. method, of two different series of welded joints, welded both by
hybrid technique, but with different parameters. A series has been studied
in this work, B and C series are reported in a previous work [10].
179

Experimental data are fitted with a power law of Basquins type where the
number of cycles to failure increases as a negative power function of stress
range %
a
and of the local amplitude of strain $
a
.


Figure 8.Fatigue curve %
a
N.


Figure 9.Fatigue curve $
a
- N.

It can be noted that A series has lower fatigue strength compared to B
and C series, and this result is due to different set of parameter in the
joining process. The fatigue resistance for A series butt joints is lower of
about 28% in terms of nominal amplitude of stress and of about 35% in
terms of local amplitude of strain.
Figure 10 shows some the fatigue final fracture paths that are localized
within the weld cord (butt A2, butt A3) and near drops of fused metal (butt
A4). Crack initiation is highly affected by these defects and fatigues cracks
initiate always from these sites if they are present.

C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi


Butt A2 Butt A3 Butt A4
Figure 10. Different fatigue fracture locations for hybrid butt titanium grade 5
welded joints.

In spite of the presence of these welding defects, fatigue curves expressed
in terms of both nominal amplitude of stress and in terms of local amplitude
of strain are well correlated. It should be noted that for the A series the
correlation factor is higher for the nominal amplitude of stress fatigue curve
and this result is different for B and C series. The different grid length of
strain gages used could cause it. In fact, in this work, 1.5 mm grid length has
been used in order to measure the strain near the weld toe instead of 3 mm
grid length used for B and C series. The strains are measured in a
smaller area more closely to the weld cord. However the fatigue failures
didnt occur always within the weld cord or at the weld toe and a shorter grid
length integrates the strains in smaller area in order to capture the local effect
near the weld toe. For this reason, in this case, the local amplitude of strain
can describe the fatigue behaviour of titanium grade 5 welded joints only in
a partial way.
The analysis on the cross section of the fatigue failure surfaces by using a
scanning electron microscope (SEM) has been carried out using a ZEISS evo
microscope. Fractographic observations have been performed using the SEM
operating in the secondary electron mode. In particular, the spacing of the
fatigue striations on the fracture surfaces have been observed on surfaces
perpendicular to the incoming electron beam. These observations were
intended to give an indication of the local fatigue crack initiations and
propagation during the test. Specimens have been cut and prepared in order
to obtain the sample suitable for this kind of analysis. The effect of drops of
fused metal distant from the weld cord has been investigated observing the
cross section of the final fatigue fracture initiate from this site.
Figure 11 shows some fractographic observations on the fatigue fracture
surface for the specimen Butt A2.

181


Figure 11. SEM observations of fatigue fracture surface for Butt A2 specimen.

In this case, the fatigue failure occurs within the weld cord, and it can be
noted the presence of porosity within the fused zone (Figure 11c). However
the main crack responsible for fatigue failure is not localized near the
porosities, but in the lower part of the specimen (Figure 11d). Figure 11a
shows the fatigue striations and Figure 11b is the observation of zone with
the static failure. The initiation of the main fatigue crack, can be noticed in
Figure 11d. This result indicates that fatigue cracks can initiate in
correspondence of weld defects, but the final failure can depend on other
causes such as the local geometry of the weld cord and sharp notch present
on the cord surface can influence more the fatigue strength of titanium grade
5 butt welded joints.
Figure 12 shows some fractographic observations on the fatigue fracture
surface for the specimen Butt A4.

a)
b)
c) d)
C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi


Figure 12. SEM observations of fatigue fracture surface for Butt A4 specimen.

In this case, the fatigue failure occurs near the drop of fused metal (Figure
12a), from this site it can be noticed the main fatigue crack. Figures 12b and
12c show the fatigue striations in different zones of the cross section, and
Figure 11d shows a global observation obtained by the SEM microscope of
the fracture surface. The fatigue crack growth rate is lower in Figure 12b and
12c, compared to Figure 12a. The higher crack propagation rate noticed in
Figure 12a can be attributed to the presence of the fused metal drop that
increases the stress concentration.
5. CONCLUSION
Hybrid (laser/MIG) butt welded joints have been considered in this work.
Residual stress measurements have been executed by means of X-ray
diffractometer, which gives a concentrated localized value of stress.
Measured points are located along a direction transversal to the cord.
Analysis of residual stresses measured values suggests the following
considerations:
HDM measurements on hybrid butt welded plate show that tensile
stresses are produced in the transversal and longitudinal directions
with respect to the weld cord and that longitudinal residual stress are
higher than transversal residual stresses. These characteristics agree
well with the results reported in the literature.
Regarding possible symmetries, a panel designed conveniently for
measuring residual stress will have a double symmetric geometry.
This, however, does not imply a complete symmetry of temperature
a)
b)
c) d)
183

field during the welding process. The joint in fact is not realized
simultaneously along its length and differences residual stress can be
found between the areas at the beginning and at the end of the joint.
Tensile stresses in the transversal directions with respect to the
weld cord near the cord become compressive distant from the weld
cord.
Stress level measured near the surface in the direction normal to
the cord by XRD on hybrid butt plate are in good agreement with the
stress level relived by HDM measurements.

The fatigue behaviour of butt hybrid welded joints in titanium grade 5 has
been studied. %
a
N and $
a
N fatigue curves have been obtained. $
a
N
fatigue curve has been plotted according to the Wel.Fa.Re. method
recommendations: two electric strain gauges have been bonded with their
principal axes at 1 mm from the weld toe for each weld cord, then the local
amplitude of strain is calculated applying the maximum and the minimum
load of the fatigue test.
The local strain amplitude can account for the effects of many factors
influencing fatigue life of welded joints and that the nominal stress concept
could not appreciate. However in this case, the fatigue failures didnt occur
always within the weld cord or at the weld toe and the local amplitude of
strain can describe the fatigue behaviour of titanium grade 5 welded joints
only in a partial way.
Finally a comparison with the fatigue resistance of Ti6Al4V butt hybrid
welded joints from a previous work has been presented. The fatigue
resistance of titanium hybrid welded joints is strictly dependent to joining
process parameters.
REFERENCES
1. ASM International. ASM Handbook Properties and Selection:
Nonferrous Alloys and Special Purpose Materials, vol. 2. 1992.
2. ASTM B265, Standard Specification for Titanium and Titanium Alloy
Strip, Sheet and Plate. 2006.
3. Noolua NJ, Kerra HW, Zhoua Y and Xieb J. Laser weldability of Pt
and Ti alloys. Materials Science and Engineering A, 397:8-15, 2005.
4. M. Ya, P. Marquette, F. Belahcene, J. Lu, Residual stresses in laser
welded aluminium plate by use of ultrasonic and optical methods,
Materials Science and Engineering A Elsevier, 382:257264, 2004.
5. M. Eboo, W. M. Steen,J. Clarke. Arc-augmented laser welding. Proc.
of 4th International conference on advances in welding processes,
Harrogate, UK, 1978.
C. Casavola, C. Pappalettere, F. Tattoli, F. Tursi

6. Bang Hee Seon, Bang Han Sur, Kim You Chul, Oh Ik Hyun. A study
on mechanical and microstructure characteristics of the STS304L butt
joints using hybrid CO2 laser-gas metal arc welding. Materials and
Design - Elsevier, 32:23282333, 2011.
7. F. Kong, R. Kovacevic. 3D finite element modeling of the thermally
induced residual stress in the hybrid laser/arc welding of lap joint.
Journal of Materials Processing Technology - Elsevier, 210:941
950, 2010.
8. G. A. Webster, A. N. Ezeilo. Residual stress distributions and their
influence on fatigue lifetimes. International Journal of Fatigue -
Elsevier, 23:S375S383, 2001.
9. Structural Welding Code Titanium. AWS D 1.9/D 1.9/M, 2007.
10. Casavola C, Pappalettere C and Tattoli F, Fatigue behaviour of hybrid
Ti6Al4V cruciform welded joints. Proceedings of New Trends in
Fatigue and Fracture, Polignano a mare Bari (Italy), July 2011.
11. Casavola C, Pappalettere C and Tattoli F, Confronto sulla resistenza a
fatica di diversi tipi di giunti saldati in lega di titanio. Proceedings of
IGF Gruppo Italiano Frattura, Forni di Sopra Udine (Italy), 1-3
March 2012.
12. C. Casavola, C. Pappalettere and F. Tattoli. Experimental and
numerical study of static and fatigue properties of titanium alloy
welded joints. Mechanics of Materials, 41:231-243, 2009.
13. G. Ivetic, A. Lanciotti, C. Polese. Electric strain gauge measurement
of residual stress in welded panels. Journal of Strain Analysis,
44:117126, 2009.
14. Stresstech group, X-ray Diffraction. The method for residual stress
testing, 2003.
15. C. Casavola, R. Nobile and C. Pappalettere. Fatigue strength by the
wel.fa.re. Local Strain Method: application to 3-5 mm cruciform and
butt welded joints. SEM Annual Conference and Exposition on
Experimental and Applied Mechanics, June 10-12 2002, Milwaukee
(USA). 2002.
16. C. Casavola, R. Nobile and C. Pappalettere. Fatigue life predictions
by the wel.fa.re. method: influence of residual stresses. SEM Annual
Conference and Exposition on Experimental and Applied Mechanics,
2-4 June 2003, Charlotte (USA). 2003.
17. C. Casavola, R. Nobile and C. Pappalettere. A local strain method for
the evaluation of welded joints fatigue resistance: the case of thin
main plates thickness. Fatigue & Fracture of Engineering
Materials & Structures, 28:759-767, 2005.
18. C. Casavola and C. Pappalettere. Industrial application of a new local
strain method for fatigue strength evaluation of welded structures.
ICEM 12 International Conference on Experimental Mechanics,
August 2004, Bari (Italy). 2004.
185

residual stresses caused by welding. Residual stresses can have a significant
influence on the fatigue lives of engineering components [8]. In particular,
near surface tensile residual stresses tend to accelerate the initiation and
growth phases of the fatigue process while compressive residual stresses
close to a surface may prolong fatigue life. Significant advances have been
made in recent years for obtaining accurate and reliable determinations of
residual stress distributions.
Fatigue strength of welded joints is lower than base material because of
microstructure modifications, misalignment, geometrical discontinuity of the
cord, residual stresses and weld defects. In order to take into account all the
variables affecting the fatigue strength, the fatigue design of welded
components is done on the basis of experimental results, using fatigue curves
available in codes [9].
The fatigue behavior of a ! + " Ti6Al4V alloy butt welded by hybrid
technique has been investigated in this paper. The experimental results have
been compared with a previous study on similar welded joints, but obtained
with different process parameters [10-11]. Fatigue fracture surface have been
observed by scanning electron microscope. Besides in this paper
experimental hole drilling method and X-ray diffraction measurements have
been reported.
2. MATERIAL AND METHODS
The material used in the tests was titanium grade 5 (Ti6Al4V), an !#"
alloy with aluminium that is an !-stabilizing element and vanadium that is a
"-stabilizing element. Ti6Al4V has good mechanical properties both at room
and high temperature. Its density is half steel density (4.4 kg/dm
3
[2]), the
ultimate strength is 980 MPa and the yield strength is 760 MPa. Titanium
grade 5 mechanical properties and chemical composition are indicated in
tables 1 and 2.

Table 1: Mechanical properties of grade 5 titanium alloy [12]
Yield strength
(MPa)
Ultimate strength
(MPa)
Youngs Modulus
(MPa)
Elongation at
fracture (%)
760 980 110200 14

Table 2: Chemical composition of titanium grade 5 according to ASTM B265 [2]
Element C Al V Fe H N O
% < 0.08 5,5 6.75 3,5 4.5 < 0.30 < 0.015 < 0.03 < 0.25

At room temperature, unalloyed titanium has a hexagonal close-packed
(hcp) crystal structure called !-phase. At 883 C this transforms to a body-

1

INFLUENCE OF GEOMETRIC SHAPE OF
SPECIMEN IN FATIGUE LIFE
CHARACTERIZATION ON WELDED JOINT IN
TITANIUM ALLOY

C. Casavola
1
, V. Giordano
1
, C. Pappalettere
1
, C. I. Pruncu
1*

1
Dipartimento di Ingegneria Meccanica e Gestionale, Politecnico di Bari,
Viale Japigia, 182 - 70126 Bari, Italy

Email address: catalin.pruncu@poliba.it


Abstract: From literature we can find many theories on description of behaviour of
different kind of materials. But just a few of them take into account an exact
correlation between theoretical and experimental applications. In many instances the
behaviour of these materials could be represented in function of some characteristics
as density, Young modulus and Poisson ratio. One material with good characteristics
of these listed above, that has been subjected to examination in this study, could be
represented by titanium alloy and in particular titanium alloy grade 5. This material
is suitable for the majority of applications in aeronautical, aerospace and naval
industries and biomedical area. To be used in the field mentioned above, often
titanium must be incorporated in different structures that can be obtained with
welding process assembly. This new component could be more sensitive to crack
initiation near the welded area due to the influence of the welding process. In this
paper we try to elucidate some problems related to the fatigue life resistance of
components simulating the activity during the real life of an assembly. The results
have been confirmed by fractography analysis of broken surfaces, with a Scanning
Electron Microscopy (SEM) and an optical microscope and have been correlated
through numerical analysis.
Key words: FATIGUE LIFE, CRACK INITIATION, WELDED JOINTS
C. Casavola, C. Pappalettere, V. Giordano, C. I. Pruncu

1. INTRODUCTION
Material inhomogeneities are, in general, defects in the translational
invariance of material properties. In a specimen different material regions
can be seen, with rapid changes of microstructure and properties, which may
be more or less smooth or even abrupt (1). Obviously it is virtually
impossible to manufacture particles with perfectly smooth geometries (2).
The presence of micro defects in material due to certain variation of
temperature or void merge during the process of fabrication is inevitable.
The study of the physical process about the behavior of material can be
dedicated to the nucleation phase of crack or void described with the
inequality (a < a
m
) (3), where a
m
is the smallest crack length detectable by
current technology, which is about 0.1!m. Heterogeneities in order of few
micrometer are present in stainless steel as sulfide inclusions with size of 10
!m (4). In effect damage tolerance analysis which describes fatigue crack
growth should be one connection to solve the fatigue nucleation stage.
In a theoretical vision, defects correspond to mathematical singularities in
a continuous description that can be easily detected in the elastic field. An
important subject of contemporary theoretical and applied mechanics is the
fracture. The most relevant notions developed by engineers in that context
are those of driving force acting on the tip of a crack and of energy-release
rate.
2. OVERVIEW OF THE STATE OF THE ART
Description of fatigue initiation life in different medium, represented by
solid mechanics and fluid mechanics fields, is yet open for all research
society. Moreover, the technical society until today have provided in
literature particular descriptions, lemmas and corollaries in goal to explain
that topic. In next few lines, we will point some of the most important issue.

2.1 DEFINITION OF FATIGUE LIFE INITIATION

In literature, fatigue initiation life of solid materials, is assumed to be
represented by the phase of crack initiation activity. In this way, for the
description of this phase of crack initiation different definitions have been
proposed. In absence of standard agreement, it is admitted to define the
fatigue life initiation by a threshold crack length. The definition of the
micro-crack initiation length is supposed to be quite equal to 0.1 mm for the
majority of steels. This dimension is generally easy to detect and
corresponds to a defect comparable to the steel grain size average. Fatigue
initiation life is a consequence of cyclic slip in slip bands which implies


cyclic plastic deformation as result of moving dislocations. This micro-
plasticity can occur more easily in the grains at the material surface.
Accordingly the initiation fatigue life is supposed to be completed when
micro-crack growth is no longer depending on the free surface conditions.

2.2 FATIGUE LIFE WITHOUT INITIATION

Some authors like Murakami (5) considered that each material contains a
defect as a crack. Where the size is characterized by the root square of the
area parameter (area of the defect projected on the plane perpendicular to
the direction of the maximum principal stress) and defined that crack
initiation stage is negligible (Ni ! 0). Fatigue life is therefore assumed to be
controlled by the crack propagation law.

2.3 FATIGUE INITIATION PARAMETER

In any case, fatigue resistance to initiation could be represented by a
power relationship of the form Pi= f (Nib) where Pi is an initiation parameter
and b exponent of Basquins type. Several approaches have been used to
define the initiation parameter Pi: effective stress, notch stress intensity
factor, damage parameter issued from SWT parameter (6), energy parameter.

It is acceptable to use stress range as a parameter to express fatigue
resistance to initiation. In this case it is necessary to take into account that
fatigue generally initiates from stress concentration area which induce stress
gradient at notch tip. Stress distribution at notch tip was assumed to derive
from Irwins elastic stress distribution at crack tip by Creager and Paris (7).
To avoid stress singularity at notch tip the Irwin stress distribution at notch
tip is shifted by introduction of a new origin located at distance "/2 to the
previous one. The stress distribution is governed by the (crack) stress
intensity factor K and represented by :


3
cos 1 sin sin
2 2 2 2 2 2 2
I I
yy
K K
r r r
! ! ! " !
#
$ $
% &
= + +
' (
) *
(1)

where r and ! are polar coordinates and " the notch radius

This method assumed that the effective stress is the maximum stress
located at notch tip which is given by:

( )
2
2, 0
max
K
x
yy
! "
# #
$!
= = = =
(2)

C. Casavola, C. Pappalettere, V. Giordano, C. I. Pruncu

Barsom and McNicol (8), Jack and Price (9), Clark (10) and Truchon (11)
have used the parameter
K ! "
to express fatigue resistance on notched
specimens.
This fatigue resistance can be expressed by the concept of effective stress
range (12) #$
ef
which represents the average stress over the fatigue process
volume often reduced to the average stress over the effective distance Xef
due to the assumption that this volume is cylindrical.


0
1
( )
ef
ef yy
ef
X
r dr
X
! !
=
" " #
(3)

Lukas et al. (13) approximated this stress by an equivalent stress in
proximity of a notch tip prolongated with fictitious microcracks by the
following relationship:


,
0
1 4.5
t g
eq Lukas
k
a
!
"
"
#
# =
+
(4)

k
t
is the elastic stress concentration factor $
g
the applied gross stress
range. The fictitious microcrack length used by Lukas et al (14) is here
assumed equal to ao:


( )
2
1
0
K
th
a
D
!
"
#
=
(5)

where $
D
is the fatigue limit and #Kth the fatigue threshold.

Due to stress concentration, cyclic plasticity occurs generally at notch tip
and elasto-plastic stress distribution is generally more complex than the
elastic one and it can be obtained by Finite Element computing. By applying
effective stress range definition given in formula (3), Capelle and al (15)
have presented fatigue resistance to initiation of an X52 pipe steel.
The stress distribution at notch tip at a distance greater than the effective
distance can be described by the so-called Notch Stress Intensity Factor
(NSIF). Lazzarin et al (16) use the elastic NSIF K* as parameter for fatigue
initiation. Boukharouba et al (17) have compared fatigue initiation criteria
with a criterion based on the elasto-plastic notch stress intensity factor K" on
welded specimens made in low strength steel; The weld toe was considered
as a notch with a notch angle of 135. The NSIF has then unit MPa *m
%

191
where % is the power dependence of the stress distribution (% = 0.385).

Smith et al (18) have shown experimentally that the result of equation (6)
is constant whatever the stress ratio is:


.
max
C SWT
E
a
! "
= =
(6)

where $
max
is the maximum stress, E the Youngs modulus and &
a
the
strain amplitude. This result is called the SWT parameter and it is generally
used to convert low cycle fatigue resistance at zero stress ratio to low cycle
fatigue resistance at any stress ratio. Lazzarin et al (16) have derived a so-
called damage parameter D from the SWT parameter and expressed it as a
stress (MPa*units); They have presented that for thin plates notched by
symmetric lateral U-, V- shaped notches with notch root radii ranging from
0.1 to 10 mm and made in steel. All the results merge in a unique curve
independently of the notch radius.

Fatigue crack initiation of engineering components exhibiting
significantly small dimensions imperfections short flaws is expressed as :

0
th
c
K
F a
!
"
#
# =
(7)

where F is a geometrical factor.

This fatigue initiation stress range for both short and long crack is
expressed by Kitagawa and Takahashi diagramme (19) where the crack
growth threshold is plotted against the crack size.Using empirical law
relationship (5) we can write:


( )
0
0
th
c
K
F a a
!
"
#
# =
+
(8)

Another formula is presented by Jelaska and Podrug (20) in cumulative
HCF/LCF loading:


0
2
i
i
i
m
gr
i m
m
U a m
U a m HCF
N
N
n
!
! ! !
! ! !
=
" #
+
$ %
+ &
' (
(9)

where mi ,$
u
, $
a
, $
m
, $
0
, N
gr
, n
HCF
are material constant, ultimate strength,
C. Casavola, C. Pappalettere, V. Giordano, C. I. Pruncu

stress amplitude, maximum stress, endurance limit, sufficiently long fatigue
life, number of HCF cycles.
3. MATERIALS AND WELDED BUTT JOINTS

In last period materials as titanium and titanium alloy are finding
widespread applications due their good characteristics. Consequently they
show a good balance of mechanical and chemical properties and hence cover
many areas of application in diverse industries. The most widely used
titanium alloy is Ti-6Al-4V ( %' alloys ). These alloys contain larger
amounts of beta stabilizers (4 to 6%) (21). Compared with steel, titanium or
titanium alloys have low density and shows a similar fatigue resistance and
presents a very attractive strength weight ratio.
The most influential microstructural parameter on the mechanical
properties of titanium alloys can be influenced by either a alloys (hexagonal
crystal structure) or ' alloys (body centered cubic structure) and ab alloys.
Properties of %' alloys (titanium grade 5) can be controlled through heat
treatments (22), (23). Chemical compositions of tested materials are reported
in Table 1 and Table 2.

Table 1. Chemical composition of titanium alloy grade 5.

Titanium
grade 5
C Al V Fe H N O
ASTM
B265
<0.08 5.5-.75 3.5-4.5 <0.30 <0.015 <0.03 <0.25

Table 2. Mechanical properties of titanium alloy grade 5.

Titanium
grade 5
Modulus of
elasticity E [MPa]
Ultimate tensile strength
R [MPa]
Yield tensile
strength R
02

[MPa]
ASTM
B265
1.138 956.4 760
The alloy is recommended mainly for design disks, blades, spacer rings,
engine body and other aircraft-engine, components operating at temperatures
up to 400450C (24). In the case where these materials are integrated in
structure mentioned above, different procedures for joining are used. One
possibility is to use mechanical joining by welding process as butt welded
joint. Cited by Casavola et al (22), specimens obtained by process of
193

welding and more precisely butt welded joint show sensitivity to fatigue
life initiation at conjunction of weld cord and parent material. Of course, in
this region, which corresponds to the heat affected zone (hosting
microstructural changes) the stress/strain "eld is usually ampli"ed with
respect to the nominal stress values because of the effects of local geometry
of the seam, localized plasticization and secondary bending produced by
misalignments.
In the other hand, one great problem is represented by the slags which
are formed and deposited, during the welding process, in proximity of the
weld cord. Due to the difficult task to eliminate these entities caused by
solidification, life of butt welded joint could decrease. For example using
metal removal in steels and low alloy steels, can cause un-tempered
martensitic to form, which is brittle and causes significant shortening of
fatigue life. In this case the welding processes must be improved, redefining
welding time and speed (25).
Figure 1 shows the approximate pro"les recognized by a tactile device,
acquired namely by a Coordinate Measuring Machine (CMM) for the right
and reverse sides of the butt welded specimen.


Figure 1. Butt welded joint specimen: right (upper) (a) and reverse (lower) (b) side
seam pro!le.
4. RESULTS AND DISCUSSION

After fatigue tests, on specimens of butt welded joint of titanium alloy,
a summary of macrographies about details of microstructure has been
C. Casavola, C. Pappalettere, V. Giordano, C. I. Pruncu

illustrated. Thus such macrography, characterizes areas where initiation of
crack occurs. In this way are taken samples all over the material surface,
both from the welding area connection and also from the area where the
slags irregularities are deposited. With an optical microscope we show the
arrangement of grains, in which we can see the distribution of lamellar
grains. In figure 2 we can see the distribution of microcracks between grains
which occur concerning intergranular life initiation.


Figure 2. Microcrack initiation in intergranular mode, titanium alloy
Ti-6Al-4V (!"#)

On the other hand, in the adjacent area of slag we found some
inclusions which can lead to trigger the process of fatigue. Difference in
color between material layer under the slag and titanium alloy displayed in
figure 3 (a), leads to the idea that, in this area could be occurred a chemical
process in different way that also influences the fatigue behavior. At the
interface of these layers in figure 3 (b), is evident the distribution of line of
fatigue.

Site of life initiation
fatigue
195
Figure 3. Site of fatigue initiation characterized through a) inclusion on the slag
area, b) line of life initiation fatigue

Under some influences of temperature variations, correlated with the
fatigue damage test effect involved, the critical strain in ductile mode
fracture is locally showed. Exceeded this critical strain shear bands and
small voids are formed. These voids are known as strain induced porosity
(SIP). An example of strain induced porosity in Ti-6Al-4V is shown in
figure 4. Once this porosity is created in forging billet, it is not always
healed during subsequent forging operations. Consequently, it can be carried
over into a finished component as a defect that can act as an early fatigue
crack initiation site (25).


Figure 4: Representation of face of crack initiation life
In order to correlate the experimental results, we apply the numerical
analysis model. Models with Extended Finite Element Method (XFEM) have
been realized and the analysis has been executed through Abaqus Software.
The numerical model considered had the same shape of butt welded joint
acquired by the Coordinate Measuring Machines. For fatigue tests a value of
charge loading of 24.5 kN has been imposed. According with the
experimental results, the model has incorporated one irregularity. This
irregularity is exemplified through one crack of about 0.1 mm in the weld
cord area. Shape of specimen is represented in figure 4 a). Although, this
crack may produce implications, expecting an increase of stress in the
welded cord area, an obvious increase in stress in the slag area, has been
registered. These trends of stress distribution are showed in figure 5 b). This
complex behaviour can be explained by the appearance of inclusions under
the slag area as displayed in figure 6.
C. Casavola, C. Pappalettere, V. Giordano, C. I. Pruncu



Figure 5: a) Shape of specimen analyzed with crack in weld area, b) Distribution of
maximal stress versus displacement in the sample with and without inclusion


Figure 6: Model of butt welded joint with inclusion in the slag area

5. CONCLUSION
Experimental determination and numerical analysis of fatigue life
initiation are illustrated in this paper. Thus, factors which could influence the
behavior of butt welded joint specimen are determined. Some specific
parameters which could be mentioned are : distribution of boundary grains
on material which show lamellar form and early phase in microcrack
intergranular behavior, varieties of some inclusion unexpected in base
a)
b)
197
material, difference of temperature in the bonding area represented through
interface slag and parent material. The same could be cited about factors
regarding the growing in size of voids, which involve the strain behavior and
initiation life.
On the other hand, by XFEM is presented the distribution of stress
along the shape of butt welded joint and the behavior of one irregularity in
the body of material alloy, through which we can deduce that, the true life
initiation of fatigue, in titanium alloy, is represented by a complex of
parameters from shape of material to manufacturing process.

REFERENCES
1.Arkadi Berezovski. Numerical Simulation Of Waves And
Fronts In Inhomogeneous Solids. World Scientific Publishing
Co. Pte. Ltd., London,ISBN-13: 978-981-283-267-2, 2008.
2.T.I. Zohdi. Introduction to Computational Micromechanics,.
Society for Industrial and Applied Mathematics, ISBN 978-0-
898716-27-6, 2007.
3.J.K. Paik. Conditionassessment assessment of aged structures.
s.l. : Woodhead Publishing Limited, ISBN 978-1-84569-334-3,
2008.
4.E. McCafferty. Introduction to Corrosion Science. s.l. : Springer
Science+Business Media, LLC., New York, ISBN 978-1-4419-
0454-6, 2010.
5.Y. Murakami, S. Kodamab and S. Konumac. Quantitative
evaluation of effects of non-metallic inclusions on fatigue next
term strength of high strength steels. International Journal of
Fatigue, Volume 11, Issue 5, Pages 291-298, September 1989.
6.K. N. Smith, P.Watson and T. H.Topper. A stress-strain function
for the fatigue of metals,. Journal of Materials, JMLSA, 5, 767-
778, 1970.
7.M. Creager and P. C. Paris. Elastic field equations for blunt
cracks with reference to stress corrosion cracking,. Int.J.
Fracture Mech. 3, 247-252, 1967.
8.M. Barsom and R. C. McNicol. Effect of stress concentration on
fatigue-crack initiation in HY-130 steel,. ASTM STP 559, 183-
204 , 1974.
9.A. R. Jack and A. T. Price. The initiation of fatigue cracks from
notches in mild steel plates,. Int. J. Fracture Mech. 6 (4) ,1970.
C. Casavola, C. Pappalettere, V. Giordano, C. I. Pruncu

10.W. G. Clark, Jr. Evolution of the crack initiation properties of
type 403 stainless steel in air and steam environment,. ASTM
STP 559, 205-224,1974.
11.M. Truchon. Amorage de fissures en fatigue partir d'entaille,
application aux joints souds,. Bulletin Technique du Bureau
Veritas 67 (5),1985.
12.G. Pluvinage. Fatigue and Fracture emanating from stress
concentrators,. Kluwer,Dordrecht, ISBN: 1-4020-2612-9, 2003.
13.P. Lukas, L. Kunz, B. Weiss, and R.Stickler. Non-damaging
notches in fatigue. Fatigue & Fracture of Engineering Materials
& Structures, 9, 195-204, 1986.
14.P. Lukas. Stress intensity factor for small notch-emanating
cracks,. Eng. Fract Mech., 26:4713, 1987.
15.J. Capelle, J. Gilgert, G. Pluvinage. A fatigue initiation
parameter for gas pipe steel submitted to hydrogen absorption,.
International Journal of Hydrogen Energy, Volume 35, Issue 2,
January, Pages 833-843, 2010.
16.P. Lazzarin, R. Tovo and G. Meneghetti. Fatigue crack
initiation and propagation phases near notches in metals with
low notch sensitivity,. Int. J. Fatigue Vol. 19, Nos 8-9, pp. 647-
657, 1997.
17.T. Boukharouba, T. Tamine,L. Niu, C. Chehimi and G.
Pluvinage. The use of notch stress intensity factor as a fatigue
crack initiation parameter,. Engineering Fracture Mechanics,
Vol. 52, No. 3, pp. 503-512, 1995.
18.R.A. Smith, K.J. Miller. Prediction of fatigue regimes in
notched components,. Int J Mech Sci., 20:2016, 1978.
19.H. Kitagawa, S. Takahashi. Applicability of fracture mechanics
to very small cracks or cracks in the early stage. In: Proceedings
of second international conference on mechanical behaviour of
materials,. ASM., p. 62731,1976.
20.D. Jelaska, & S. Podrug. Crack initiation limits prediction at
combined HCF/LCF loading,. Advanced Engineering, ISSN
1846-5900, 2, 2008.
21.Vydehi Arun Joshi. Titanium Alloys An Atlas of Structures and
Fracture Features. Taylor & Francis Group, ISBN 0-8493-5010-
7 , LLC, Boca Raton, 2006.
22.C. Casavola, C. Pappalettere, F. Tattoli. Experimental and
numerical study of static and fatigue properties of titanium alloy
welded joints. Mechanics of Materials 41, 231243 , 2009.
199

23.ASM International. ASM Handbook Properties and Selection:
Nonferrous Alloys and Special Purpose Materials. vol. 2,1992.
24.Valentin N.Moiseyev. Titanium Alloys Russian Aircraft and
Aerospace Applications,. Taylor & Francis Group 6000 Broken
Sound Parkway NW, Suite 300, ISBN 0-8493-3273-7,2006.
25.Gerd Ltjering, James C. Williams. Titannium,. Springer-
Verlag Berlin Heidelberg, ISSN 1619-0181, 2007.

You might also like