Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Engineering with Computers (2003) 19: 142–151

DOI 10.1007/s00366-003-0253-3

ORIGINAL ARTICLE

J. Rudolph · E. Weiß · M. Forster

Modeling of welded joints for design against fatigue

Received: 15 March 2002 / Accepted: 19 December 2002


 Springer-Verlag London Limited 2003

Abstract Historically, design against fatigue of welded approaches with specific applications, validity ranges and
joints is widely based on the nominal stress approach with modeling requirements (Radaj and Sonsino 1998).
categorized structural details and appropriate S-N-curves According to Fig. 1, a basic classification has to be carried
without any need of detailed numerical analysis. Recently, out into the categories of stress/strain-based approaches
local approaches based on structural and notch stresses on the one hand, and crack propagation approaches on the
are of growing importance for the lifetime assessment of other (Radaj and Sonsino 1998). The primary feature of
components without post-weld treatment for dominant stress life approaches is their range of validity presuming
elastic material behavior in high cycle fatigue (HCF), and dominant elastic behavior of the material and the
for design against the endurance limit. Finally, high qual- structure. Stress ranges used for estimating the fatigue life
ity weld seams submitted to post-weld treatment such as of the component are therefore directly and linearly pro-
grinding and TIG or plasma dressing and/or operating portional to the external loading applied
conditions in low cycle fatigue (LCF) with high plastic
⌬␴ ~ ⌬L; ⌬␴ = c · ⌬L (1)
deformations are advantageously considered within a
strain-based approach. A concept-conforming modeling where c is the elastic transformation factor (Seeger 1996),
strategy as a relevant module of lifetime assessment is which can be determined by stress analysis or strain
required to ensure reliable and reproducible analysis measurement (Haibach 1968). The information from the
results for all local approaches of design against fatigue. fatigue curve (S-N-curve), together with possible correc-
A unique modeling strategy based on an adaptive nominal tion factors for mean stress and wall thickness, is used to
weld seam contour is derived covering a wide range of estimate the number of cycles to failure, or to derive a
typical welded pressure vessel joints for stress and strain complete component fatigue curve. Variable amplitude
based approaches. An alternative method for the determi- loading requires the further application of a damage
nation of structural stresses is proposed to avoid the uncer- accumulation hypothesis such as Miner’s rule (Radaj and
tainties of extrapolation and description of structural stiff- Sonsino 1998). The whole procedure for stress-based
ness. approaches is shown in Fig. 2. Fracture of the component
or a defined crack length (e.g. half of the wall thickness)
Keywords BEA · FEA · Local approaches · Low
cycle fatigue (LCF) · Modeling · Strain-based approach
· Structural and notch stresses

Introduction
The design of welded pressure vessel components against
fatigue is currently based on different conceptual

J. Rudolph (쾷) · E. Weiß


Department of Biochemical and Chemical Engineering,
University of Dortmund, D-44221 Dortmund, Germany
E-mail: rudolph@ct.uni-dortmund.de

M. Forster
Fluxus Technology Ltd, Rendsburger Landstrasse 345, D-24111
Kiel, Germany Fig. 1 Fatigue of welded components – conceptual overview
143
Fig. 2 Application of stress based approaches
schematically

is usually the limiting failure criterion in these cases. Gen- stresses is not always evident. Methods of linear or quad-
erally speaking, local yielding effects cause a difficulty ratic extrapolation, as well as direct analysis or measure-
for these approaches (Dowling 1999), and application of ment at a certain distance from the weld toe, are under
the stress-life curves should be limited to high cycle discussion at the moment. Code-based design mainly con-
fatigue (HCF) and endurance limit, as shown in Fig. 2. It tains a linear extrapolation procedure with points of extra-
is important to mention these restrictions as they are even polation at 0.4 * wall thickness and 1.0 * wall thickness,
neglected within the latest editions of pressure vessel respectively (prEN 13445–3 n.d.), as shown in Fig. 3 for
design codes, such as (AD-Merkblatt S2 (1998). a typical nozzle-to-vessel intersection. Typically, the
Among the methods mentioned, the nominal stress maximum structural stress is determined within shell or
approach can be considered as the most classical. It is coarse solid models from analytical expressions as closed
based on experimental or estimated (synthetic) fatigue form solutions or approximation formulae derived from
curves (Dowling 1999) without any need for detailed numerical analyses. The application of thin shell theory
numerical analysis. The elastic transformation factor of causes serious problems in the case of different structural
Eq. (1) reduces to the relation between external load and stiffness (e.g. thin nozzles on thick vessels), and thick
defined nominal stress in this case. The appropriate shells in general. Correction factors have to be introduced
component fatigue curve yields the admissible number of
cycles as a function of the nominal stress for a certain
probability of failure (Radaj and Sonsino 1998; Seeger
1996).
Structural and notch stress approaches take into account
the local features of the weld seam within the model and
analysis. Usually, these stresses – still being a linear func-
tion of external load – have to be determined by finite
or boundary element analysis. In these cases, the elastic
transformation factor c (Seeger 1996) of Eq. (1) represents
the structural or notch stress parameter. Detailed modeling
of weld toe notches is required to analyze highly local
notch stresses. The notch root radius is often considered
to be r = 1 mm within the worst case or mean value/scatter
assumptions (Radaj and Sonsino 1998) in connection with
appropriate fatigue curves such as FAT225, according to
Hobbacher (1996). Modeling requirements will be much
simpler if the design approach is based on structural
stresses at the weld toe. In this case, the model does not
have to reflect the detailed notch geometry of the weld toe.
Nevertheless, the determination of the required structural Fig. 3 Determination of structural and notch stresses at the weld toe
144

to account for size effects (AD-Merkblatt S2 1998; prEN adequate and realistic modeling strategy of the weld seam
13445–3 n.d.; Hobbacher 1996). region is a relevant module of lifetime assessment,
One further important feature of stress life approaches especially within this approach, but is desirable for all
to design against fatigue is their common application for other local approaches (structural and notch stress) as well
weld seam qualities without post-weld treatment, where (Rudolph et al. 2001a). This strategy is outlined in its
inherent crack-like flaws readily precipitate fatigue crack- essential parts within this article. For further reference,
ing (Maddox 1994). Fatigue curves with typical steep Rudolph et al. (2001b) is recommended.
slopes of m = 3 in double logarithmic scale and inclination
points to the endurance limit at NE = 2,000,000–5,000,000
(see Fig. 2) are used for components with negligible Derivation of the adaptive nominal and special weld
initiation phase to an engineering crack size. They follow seam contours
the equation

冉冊
1 Within the developed strategy, the modeling of the weld
NE m seam contour emanates from a so-called ’adaptive nomi-
⌬␴ = ⌬␴E · (2) nal weld seam’, which is taken directly from the technical
N
drawing and which contains the main geometric para-
where ⌬␴ is the characteristic stress range of the fatigue meters like height and width. These global weld seam
model and ⌬␴E is the stress range at endurance limit. parameters already have a significant influence on the
These curves are invalid for high plastic deformations in fatigue behavior of the component, and are supposed to
low cycle fatigue (LCF). The extrapolation of these stress be responsible for size effects, as geometric similarity can
life curves to N = 100 load cycles in AD-Merkblatt S2 be lost with increasing wall thickness (Köttgen et al.
(1998) is inadmissible from a conceptual point of view. 1991). The geometry of the weld seam contour should be
The application of the concept of fatigue-curves with known prior to fatigue analysis.
uniform scatter bands, equal slope and point of inclination This nominal weld seam can be described by a simple
to the endurance limit (Haibach 1968) is recognized, set of geometric parameters (line segments L1 to L4 and
widespread and commonly used in the context of weld Keypoints K1 to K5), as shown in Fig. 4. The contour
seam qualities without post-weld treatment (Hobbacher line segments described by L1 to L4 can be of arbitrary
1996). The value of the endurance limit is the only shape (e.g. circular arcs or splines), and the introduction
remaining unknown to adjust and construct the fatigue of the angle ␥ covers a lot of important practical appli-
curves for the different structural details, as shown in Fig. cations such as oblique nozzles and butt welded joints.
2. A fracture mechanics approach is a common alternative Special weld seam contours for the different design
in these cases (Maddox 1994). approaches based on structural and notch stresses and
Obviously, this concept does not hold true for high strains can be derived from this adaptive nominal weld
quality weld seams submitted to post-welding operations seam. As an example, the contour of an idealized weld
such as disc or burr grinding and TIG or plasma dressing. seam for notch stress approaches is generated by introduc-
Crack initiation to an engineering size will occupy a sig- ing two fillet radii at the weld toes, R11 and R21, according
nificant portion of life, and the welded joint will behave to Fig. 4. For the special case R11 = R12 = 1 mm, this
more like a non-welded component (Maddox 1994), contour is in complete accordance with the modeling
including shallower slope m of the fatigue curve and shift-
ing of the point of inclination to the endurance limit
towards lower numbers of cycles (Seeger 1996). In these
cases, the local strain approach which describes the dam-
age phase of crack initiation to an engineering crack size
(0.25 mm depth and 0.5 mm length) is the appropriate
alternative. This method as a strain-based approach (Fig.
1) is as flexible as being applicable for the low cycle, high
cycle regimes and the endurance limit. The fatigue curves
are derived from non-notched and polished standard
specimens according to a standard testing procedure of the
American Society for the Testing of Materials (ASTM)
(ASTM 1997). The module numerical stress and strain
analysis is definitely based on detailed Finite-Elements or
Boundary Elements models containing all macroscopic
notches. Thus, the standard requirements of the calcu-
lation model are very high. In the context of welded press-
ure vessel components, this design concept is extremely
important for high quality weld seams, as all other
approaches are inappropriate especially for low cycle Fig. 4 Derivation of the adaptive nominal weld seam and special
operating conditions with high plastic deformations. An contours
145

requirements of the currently most advanced and stiffness and thickness problems. Instead of the real weld
experimentally well supported mean value/scatter notch seam contour with sharp notches at the toes, it is proposed
stress approach according to Olivier et al. (1994), and the to introduce the shallow fillet radius as explained before,
fatigue curve FAT225 according to Hobbacher (1996), and shown on the right-hand side of Fig. 5. Within this
respectively. substitute model, the notch effects of the weld toes are
The modeling of weld seam contours within the local completely removed geometrically. The numerical analy-
strain approach is even more sophisticated as the sis of this singularity free model excludes the notch stress
geometry description has to be as realistic as possible. A fraction (approximately), as desired within the structural
minimum set of descriptive geometric parameters has stress approach and avoids the uncertainties of the extra-
been derived for this type of weld seam for ground weld polation procedure. For practical modeling, it is proposed
toes. Weld toe grinding is a generally accepted method of to combine line segments L1/L2 and L3/L4 and determine
post-weld treatment to improve quality and fatigue the fillet radius R11 from
behavior. Guidelines and specifications are for instance
R11 = MIN(length L1; length L2) (3)
part of the European pressure vessel design code (prEN
13445–3 n.d.). Structural stresses at the fictitious locations of the weld
According to Fig. 4 the set of geometric parameters toes can be directly determined at T1 (e.g. weld toe at the
consist of the primary weld toe radii, R11 and R21, the vessel) and T2 (e.g. weld toe at the nozzle) according to
secondary radii, R12 and R22, the notch depths, k1 and k2, Fig. 5. No extrapolation or linearization procedure is
the height of the weld excess material, h, and a factor for required. The structural influence on structural stresses is
its possible excentricity e. The secondary radii, R12 and considered appropriately only eliminating a small part of
R22, are usually very small. They avoid points of singular- the nominal weld seam as shown in Fig. 5 (hatched area).
ities within the numerical analysis, and can alternatively The strategy has been applied successfully to an intersec-
be replaced by a spline between the primary weld toe radii tion of cylindrical nozzle and spherical vessel in Rudolph
and the adjacent line segments. The geometrical descrip- et al. (2001b). Further development of the strategy based
tion of the weld excess metal and its combination with on parametric studies of cylindrical nozzles on cylindrical
the weld toe radii is based on a cubic spline with appropri- shells is reported in Rudolph et al. (2002).
ate slope at the end points. Within the developed modeling
strategy, this special weld contour is completely derived
from and connected to the nominal weld seam described Modeling and analysis of butt welded seams
above, as can be seen in Fig. 4. Combined with guidelines
for FE- and BE-meshing, reproducible analysis results are The contours of butt welded seams can be derived from
ensured within this strategy. Details of the geometry the same adaptive nominal seam for the particular case of
description in general and within special preprocessors are an angle of 90 degrees, as shown in Fig. 6. The modeling
given in Rudolph et al. (2001b). strategy is valid for both local stress and strain
The complete grinding of the weld seam eliminating all approaches. The excess weld metal and its possible
weld toe notches and creating a smooth and shallow notch excentricity (factor e) is included in the modeling strategy.
contour can be considered as an ideal but expensive post- Different contours without and with undercut are equ-
weld treatment. In the absence of further flaws, the ally covered.
component will definitely behave like a non-welded As an example, the modeling strategy mentioned has
component if the basic requirements of weld seam quality been applied to the FE- and BE- analysis of a special butt
are respected. The code-based design using structural and welded joint submitted to tension load. The FE-analysis
notch stresses is not applicable, or is much too conserva- was carried out using the ANSYS program, and an
tive in this case. The derivation of the contour from the element type with quadratic shape functions. The meshing
nominal weld seam is done by simply introducing a fillet strategy in the region of maximum notch stresses is based
radius, or alternatively, a spline between the line segments on a surface layer with elements of nearly quadratic shape
L2 and L4 (see Fig. 5). according to a validated non-adaptive meshing strategy
An alternative strategy for the determination of struc- explained in detail in Weiß et al. (1999). This is extremely
tural stresses within the hot-spot-method is closely con- important for notch stress analysis with significant gradi-
nected with this model, but is based on a completely dif- ents. To reduce the meshing labor cost, the transition grid
ferent conceptual line. The basic idea is the elimination in regions with homogeneous stress distribution or low
of two serious deficiencies of the hot spot method as a stress gradients can be hybrid or triangular without loss
widespread variant of the structural stress approach: insuf- of accuracy Weiß et al. (1999). An experienced user of
ficient description of stiffness and thickness effects within commercial FE- or BE-software will be able to generate
thin shell theory (FE-analysis based on shell elements, the meshes shown in Fig. 7 within about 15 minutes.
closed form analytical solutions) and arbitrary definition For the first principal stress, a stress concentration fac-
of locations for linear or quadratic extrapolation of the tor (or elastic transformation factor c (Seeger 1996)) of
course of stress according to Fig. 3. Therefore, it is rec- c = 2.774 with respect to the tension load is calculated.
ommended to use solid instead of shell models within the The geometrical parameters are: wall thickness 10 mm,
numerical analysis of structural stresses, avoiding the height of excess weld metal 2 mm, width of excess weld
146
Fig. 5 Smoothing of weld toe notches

Fig. 6 Contour of butt welded seams for local


stress and strain approaches

Fig. 7 FE- and BE-analysis of butt


welded joint

metal 10 mm, weld toe notch radius 1 mm, depth of The comparative BE-analysis of the same weld detail
undercut 0.5 mm. The secondary radius of 1 mm elimin- based on quadrilateral elements with quadratic shape
ates all singularities from the model and the transition functions yields almost the same result with respect to
between the weld toe and the weld excess metal is smooth maximum notch stress concentration. The stress concen-
and tangential because of the spline approximation. tration factor is 2.788. The commercial BE-code B3
147

within the CADFIX system was used for this analysis. highly accurate. Comparative 3D-BE-analysis is
As can be seen in Fig. 7, a comparatively coarse surface extremely interesting for this example and shown in Fig.
mesh is sufficient for notch stress results of high accuracy. 9. As can be seen in the meshed model for 3D-BE-analy-
Modeling effort is confined to the surface of the structure. sis (Fig. 9), the grid can be much coarser compared to
As linearly elastic material behavior is assumed, the FE-analysis, and there is no need to model the interior
analyses are carried out within seconds on a usual PC volumes separately. Optionally, stress results for interior
platform (e.g. Intel Pentium 5, 512MB RAM). points can be obtained if required for the calculation of
stress gradient effects or other purposes. The division of
the structure into subregions A and B is required for
Nozzle-to-vessel intersections within local stress and numerical stability of the current version of the commer-
strain approaches cial BE-code B3. The comparison of the maximum notch
stresses of the FE- and BE- analyses (313.3 N/mm2 for
Nozzle-to-vessel intersections are typical structural FE and 309.9 N/mm2 for BE) shows that the results
components in pressure vessel technology. There is an obtained are nearly identical. Once again, the qualification
urgent need for safe and concept-conforming design of the 3D-BEM for accurate notch stress analysis at mini-
against fatigue of these special welded joints. The appli- mum modeling effort is proven impressively. Times of
cation of local approaches is particularly difficult when modeling and execution are reduced to about one quarter
rotational symmetry is lost. Apart from simple shell mod- compared to the FE analysis.
els, detailed three-dimensional solid modeling strategies In a next step, the more complicated 3D structure of a
have to be applied in these cases. nozzle-to-cylindrical vessel according to Fig. 3 has been
As one example, the 3D-solid model of a circular plate analyzed as a truly three-dimensional structure, which
with centred nozzle has been used as a starting point for cannot be generated by simple rotation of plane areas. In
3D-FE- and BE meshing and notch stress analysis. The this case, the nozzle with an outside diameter of 118 mm
weld seam contour of the nozzle-to-plate intersection has is small compared to the vessel with 800 mm. The differ-
been modeled according to the requirements of the local ence in wall thickness (5.6 mm and 10 mm, respectively)
strain approach according to Fig. 4. The solid model of also implies a change of the structural stiffness in the
the weld seam is free of any geometric singularities as region of intersection. A unit pressure load of 1 N/mm2
sources of erroneous results of numerical analysis. is to be applied.
The essential details of the global geometry and the Both weld seam contours derived in Fig. 4 are foreseen
detailed weld seam description are given in Table 1. for local stress and strain approaches. FE-modeling
The modeling and meshing labor cost for the FE-analy- becomes extremely tedious within the standard prepro-
sis (see Fig. 8) is higher than for the BE-analysis (see Fig. cessor, especially for meshes consisting exclusively of
9) due to the fact that interior volumes have to be meshed quadrilateral solid elements, as shown in Fig. 10. Mode-
adequately. As explained before, a hybrid meshing strat- ling times strongly depend upon the skills of the engineer
egy with a notch surface layer of nearly quadratic quadri- and the capabilities of the CAD or FE environment used.
lateral brick elements has been used for the FE-analysis. It may easily take one working day to generate the com-
The detailed model can be generated and meshed within pletely parametric hexahedral mesh with surface layer in
one hour. The FE-analysis will take about two hours. Iso- the notched regions according to Fig. 10. Interfaces to spe-
paramatric elements with quadratic shape functions are cialized CAD-systems and the availability of automated
generally recommendable for notch stress analysis. In this tools are vital for the widespread practical application of
case, a unit pressure load of 1 N/mm2 yields a maximum the local notch stress approach. In this example, FE-mesh-
notch stress concentration or elastic transformation factor ing (nearby 58,000 nodes) was based exclusively on
(Seeger 1996) of 313.3 N/mm2. This result has been con- quadrilateral solid elements. The standard brick elements
firmed by a 2D-FE-analysis using plane solid elements of the ANSYS program with linear and quadratic shape
with rotational symmetry option. The maximum stress functions have been used. A surface layer with four and
result for the notched structure can be considered as six elements, respectively, along the notch radius ensured
reproducible and consistent results for the notch stress
analysis. No further increase in notch stress concentration
Table 1 Geometry data for nozzle to circular plate intersection
is noted, although six elements are used along the weld
Wall thickness of plate t1 = 45 mm toe notches and the total number of nodes and elements
Wall thickness of nozzle t2 = 10 mm has been increased considerably. Computing time is
Outer diameter of nozzle d2 = 48.3 mm increased considerably as well. This would be a decisive
Outer diameter of cylindrical shell d3 = 800 mm
Wall thickness of cylindrical shell t3 = 10 mm factor of influence in a nonlinear stress and strain analysis.
Length of line segments L1 and L3 0.7 t2 Linear finite elements analysis for this example (about five
Primary and secondary radii R = 1.0 mm hours’ execution time) yields a maximum principal stress
Undercut at weld toe towards plate k1 = 0.75 mm value of 129 N/mm2 for the weld toe, as well as for the
Undercut at weld toe towards nozzle k2 = 0.5 mm
Weld excess material h = 1.5 mm crotch corner according to Fig. 10. Both locations could
Factor of excentricity of weld excess material e = 0.25 be the starting point for practical fatigue failure. Crack
initiation and propagation from the weld toe often depends
148
Fig. 8 3D-FE-analysis of a nozzle to circular
plate intersection

Fig. 9 Comparative 3D-BE-analysis

upon the distribution of microscopic defects such as drical vessel intersection as an example. The weld seam
deposits of weld material within the region of macro- region was cut into slices (strategy cut) to model the con-
scopic stress concentration. It is interesting to note that a tour correctly in the circumferential direction. Within this
hybrid meshing strategy considerably diminishes the labor strategy, weld toe radii may even vary in the circumfer-
cost for detailed modeling if applied appropriately. In this ential direction.
case, a factor of 2 compared to the meshing strategy The basic model was generated within a CAD-system
exclusively based on hexahedral elements is estimated. Of and then transferred into CADfix for further processing
course, regions of high stress gradients have to be meshed using the generic IGES format. The detailed weld model
with well shaped quadrilaterals (2D) or hexahedrals (3D). was built from faces. The final valid solid is shown in
That is why the strategy of a notch surface layer has been Fig. 11. It can be used directly for BE-meshing whereas a
applied in the three-dimensional model. The maximum reasonable FE-meshing is not possible, because of missing
principal stress value amounts to 130.8 N/mm2 compared interior volume divisions. As can be seen, the geometry
to 129 N/mm2 for the model introduced in Fig. 10. The of the weld seam may even vary along the circumference
difference is negligible. qualifying the strategy for the realistic description of non-
In a next step a weld seam model for the local strain symmetric three-dimensional contours.
approach was generated using the same nozzle-to-cylin- Of course, the parameters of mesh refinement have to
149
Fig. 10 3D-FE-analysis of a cylinder-cylinder
intersection based on notch stresses

extremely tedious for this complicated three-dimensional


structure. The division of interior volumes for this purpose
is shown in Fig. 13 on the left-hand side. End points of
lines in the cross section are starting points for intersection
curves in the circumferencial direction. Each intersection
curve has to be constructed correctly by proper overlap
operations. A further problem is connected with surface
description by NURBS (Non Uniform Rational B-
Splines), which is sometimes not as accurate as being able
to model the complicated notch geometry although bor-
dering lines are constructed precisely. This problem is due
to the aspect ratio of the bordering lines and the curvature
of the complicated intersection with notch radius. The
intersection curves are extremely long compared to the
circular arc segments describing the weld toe notch.
Sometimes concave curvature of the surface description
Fig. 11 Detailed weld seam contour for a cylinder-cylinder intersection
by NURBS is changing to convex curvature along the cir-
cumference of the weld toe notch and the FE-meshing
be determined for the Boundary-Element-analysis. In a operation may fail. To cure this problem, a partial cut
first step, a coarse mesh with 683 nodes and 202 quadrilat- strategy (subdivions of the intersection curves) has been
eral elements with quadratic shape functions have been applied for the surfaces concerned, as can be seen in Fig.
used according to Fig. 12. Only four element divisions 13 for the weld toe notches. The aspect ratio of the bor-
were set along the circumference of the quarter model. dering lines is significantly improved. Surface description
Numerical stability of the analysis requires the division of the smaller surface regions becomes sufficiently accur-
of the structure into three subregions. The analysis based ate, but labor cost for modeling is increased significantly.
on the coarse mesh yielded a maximum stress value of Automated and parameterized tools as well as push-button
152 N/mm2 for the weld toe at the vessel, as shown in solutions are required for practical application. These can
Fig. 12. This time, the elastic stress concentration is much be based both on Finite-Elements and Boundary-Elements
higher at the weld toe than at the crotch corner. Fatigue analyses (Rudolph et al. 2001a). As shown in the right
failure will emanate from the weld toe. In a comparative part of Fig. 13, the maximum first principle stress of
study, mesh refinement concerned especially the circumf- 149.72 N/mm2 in the weld toe at the vessel for FE-analy-
erencial direction where eight elements have been used. sis (ANSYS) is in excellent accordance with the BE-
The maximum stress concentration is slightly reduced due result (B3). The reduction of the problem dimension
to a smoother geometry description. With a maximum (meshing of surfaces instead of volumes) is a strong argu-
value of 148.7 N/mm2 it does not differ very much from ment for the application of the Boundary-Elements-
the result of 152 N/mm2 for the coarse mesh. Method, especially in the context of complicated three-
Solid modeling in preparation of FE-meshing is dimensional structures such as the cylinder-cylinder-inter-
150
Fig. 12 3D-BE-analysis of a cylinder-cylinder
intersection

Fig. 13 Comparative 3D-FE-analysis

section of Figs 11–13. Boundary elements of the current analysis guidelines for Finite-Elements and Boundary-
generation can obviously be applied for thin walled struc- Elements applications is indispensable. A flexible and
tures with sharp notches, as has been shown by direct comprehensive modeling strategy based on a simple
comparison with FE-results. adaptive nominal weld seam contour has been derived to
Results of both types of numerical stress and strain ensure reproducible analysis results for all kinds of local
analysis can directly be applied within the design against approaches of design against fatigue. The uncertainties of
fatigue according to local approaches (Dowling 1999), as extrapolation procedures for determining structural
one important module of lifetime estimation. stresses at the weld toe can be avoided by the application
of solid models with bluntly notched fillet radii (Rudolph
et al. 1999). From a practical point of view, there is an
Summary and Conclusions urgent need for efficient and applicable tools eliminating
the costs of labor for detailed 3D solid modeling and FE-
Concept-conforming modeling of welded pressure vessel or BE-meshing. In this context, the qualification of the
joints has been identified as a significant factor for Boundary-Elements-Method (BEM) as an efficient
improved lifetime assessment within local approaches of numerical method for 3D notch stress analysis and a use-
design against fatigue. The strategy has been derived with ful complementation of standard Finite-Elements analysis
respect to typical welded pressure vessel components. has been shown.
Welded joints have to be classified into categories with
and without post-weld treatment as the fatigue behavior,
and conceptual consequences are substantially different. References
Concerning the numerical instruments, the definition of 1 AD-Merkblatt S2 (1998) Berechnung auf Wechselbeanspruchung. AD-
clear qualitative and quantitative modeling, meshing and Regelwerk, Carl Heymanns Verlag KG, Köln
151
2 ASTM (1997) Constant-Amplitude Low-Cycle Fatigue Testing. ASTM. 9 prEN 13445–3 (n.d.) Unfired pressure vessels. Part 3: Design
Annual Book of ASTM Standards, Vol. 03.01, American Society for 10 Radaj D, Sonsino CM (1998) Fatigue assessment of welded joints by
Testing and Materials, West Conshohocken, PA., No. E606 local approaches.
3 Dowling NE (1999) Mechanical behavior of materials. Engineering 11 Seeger T (1996) Grundlagen für Betriebsfestigkeitsnachweise. Stahlbau-
methods for deformation, fracture, and fatigue. Prentice Hall, Upper Handbuch, Band1, Teil B. Stahlbau-Verlagsgesellschaft, Köln
Saddle River, NJ 12 Rudolph J, Weiß E, Forster M (2001a) Efficient tools for detailed mode-
4 Haibach E (1968) Die Schwingfestigkeit von Schweißverbindungen aus ling and analysis of welded pressure vessel joints according to local
der Sicht einer örtlichen Beanspruchungsmessung. Laboratorium für fatigue approaches. ECITTT Euro Conference on CAE Integration –
Betriebsfestigkeit (LBF), Bericht Nr. FB-77, Darmstadt Tools, Trends and Technologies. Cambridge, UK
5 Hobbacher A (1996) Fatigue design of welded joints and components. 13 Rudolph J, Weiß E, Forster M (2001b) Konzeptkonforme Modellierung
Document XIII-1539–96/XV-845–96, The International Institute of ermüdungsgefähr- deter Druckbehälterschweißverbindungen als Modul
Welding, Abington Publishing, Abington, Cambridge einer modernen rechnerischen Lebensdauerabschätzung. SHAKER Ver-
6 Köttgen VB, Olivier R, Seeger T (1991) Schwingfestigkeitsanalyse für lag, Aachen
Schweißverbin- dungen auf der Grundlage örtlicher Beanspruchungen. 14 Rudolph J, Rauth M, Weiß E (2002) Modellgebundener Strukturspan-
DVS-Bericht 133 “Berechnung, Gestaltung und Fertigung”, Düsseldorf, nungsnachweis für zyklisch druckbeanspruchte Zylinder-Stutzen-Ver-
pp. 75–84 bindungen ohne Schweißnahtnach-bearbeitung. TÜ-Zeitschrift
7 Maddox SJ (1994) Fatigue Strength of Welded Structures. Abbington 43(6):19–26
Publishing, Woodhead Publishing, Cambridge 15 Weiß E, Rudolph J, Hoffmann J (1999) Vernetzungskriterien für
8 Olivier R, Köttgen VB, Seeger T (1994) Untersuchung zur Einbindung kerbbeanspruchungs-orientierte FE-Analysen. Technische Mechanik
eines neuartigen Zeit- und Dauerfestigkeitsnachweises von Schweißver- 19(2):103–114
bindungen aus Stahl in Regel-Werke – Schweißverbindungen II.
Forschungskuratorium Maschinenbau (FKM), Frankfurt/Main, For-
schungsheft 180

You might also like