Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 74 (2014) 119 – 128

XVII International Colloquium on Mechanical Fatigue of Metals (ICMFM17)

Damage assessment of threaded connections based on an

advanced material model and local concepts
D. Panic*, Th. H. Beier, M.Vormwald
Materials Mechanics Group, Technsiche Universität Darmstadt, Franziska-Braun-Straße 3, D-64287 Darmstadt, Germany

Abstract

In the actual guidelines, for example the German VDI guideline 2230 [1] or the EUROCODE 3 [2], the fatigue assessment of bolted
connections and bolts is based on the nominal stress approach. The FKM guideline for analytical strength assessment [3] does not
include concepts for a threaded connection, it refers to the existing guidelines as [1]. At present there does not exist any guideline,
which provides a damage assessment for general threaded connection by applying local concepts. In the present paper the basics
of such a concept are outlined. The investigated metal is 34CrNiMo6, which is generally used for screw threads. The experimental
determination of the local stress-strain behaviour in mechanically threaded connections is practically not feasible and needs a
numerical approach. This numerical attempt consists of two main tasks, namely the description of the material properties and finite
element (FE) analysis. The first one is concerned with the identification of material parameters, which are necessary for the FE
analysis, especially for the advanced material model. The experimental work is given by determining the initial monotonic stress-
strain curve, the initial unloading curve, the cyclic stress relaxation and cyclic creep. The parameter identification is accomplished
using the software MATLAB, which includes the essential nonlinear optimization algorithms. The obtained parameters are entered
in the FE software, which is used for the further FE analysis. The distinctive features of the FE analysis are the nonlinearity,
especially nonlinear material behaviour, nonlinear contact algorithms and the consideration of large deformations, the high number
of load steps and the various load ratio values (R = 0.06 and R = 0.71). The results of the FE analysis are used for the assessment
of threaded connections. The damage assessment is realized with the local strain approach based on the damage parameters PSWT
and PJ. Finally the result of the simplified method according to Schneider is applied on a high preloaded bolt connection (R = 0.71)
and compared with the result achieved with the advanced material model.
©
© 2014
2014 The Authors.
Published byPublished by Elsevier
Elsevier Ltd. Ltd. under CC BY-NC-ND license.
Open access
Selection
Selectionand
andpeer-review
peer-reviewunder
underresponsibility of of
responsibility thethe
Politecnico di Milano,
Politecnico Dipartimento
di Milano, di Meccanica.
Dipartimento di Meccanica

Keywords:Finite Element, Chaboche, Döring, damage parameter, PSWT, PJ, bolted connection, threaded connection, cyclic creep, cyclic stress
relaxation,damage assessment, parameter identification, nonlinear optimization.

1. Introduction

A prerequisite for the application of local strain concepts is the exact knowledge of the local stress-strain
information. The practical access to such local stress-strain curves by measurement on threaded connections is

1877-7058 © 2014 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of the Politecnico di Milano, Dipartimento di Meccanica
doi:10.1016/j.proeng.2014.06.235
120 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128

technically unwieldy and usually not feasible. In addition to this fact the cyclic plasticity phenomena are not described
well in classical material models typically implemented in commercial FE software. Hence in this work a numerical
tool is created to enable a new improved access for applying local concepts by using an advanced material model.
This numerical approach and the main components of the work containing the description of material properties and
the FE analysis are displayed in Fig. 1.

Nomenclature

α
back stress tensor

 material elasticity (stiffness) tensor
ɂ strain tensor

ε
elastic strain tensor

ε plastic strain tensor
n normality tensor


deviatoric stress tensor

ɐ stress tensor
ȟɂ୮୪ǡୣ୤୤ effective plastic strain range
ȟɐୣ୤୤ effective stress range
ε1 normal strain
εa amplitude of the strain
φ yield function (Mises)
λ
plastic multiplier
ɐ
 


ɐ
 





ɐ yield stress
ɐ



ɐ
maximum stress (stress-relaxation test)

E Young’s modulus
ᇱ plastic tangent modulus
 number of backstress tensors
PJ damage parameter according to Vormwald
PSWT damage parameter according to Smith, Watson, Topper
 load ratio
 index 1 to M
݊Ԣ cyclic hardening exponent
p accumulated plastic strain
r(0) radius of the initial yield function (deviatoric yield stress)
 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128 121

Fig. 1. Numerical approach

2. Determining the material properties - parameter identification - Döring model

2.1. Advanced material model

The material models used in this work belong to the group of phenomenological models, which are based on the
observed behaviour of materials. The description of cyclic plasticity incorporates effects as cyclic hardening or
softening and cyclic creep. This is achieved by the implementation of a more complex material model proposed by
Döring et al. [5]. The decisive advantage in this environment is the possibility to determine the transition from the
initial monotonic stress-strain curve to the cyclic behaviour of the material. The major components of continuum
mechanical model according to Döring are presented in a compact compilation in equations (1) to (20).

1. Von Mises yield function:

Ԅ൫‫ܛ‬ǡ હǡ ” ሺ଴ሻ ൯ ൌ ሺ‫ ܛ‬െ હሻǣ ሺ‫ ܛ‬െ હሻ െ  ሺ” ሺ଴ሻ ሻଶ (1)
2. Decomposition of strain:
ઽ ൌ ઽ‫ ܔ܍‬൅  ઽ‫ܔܘ‬ (2)
3. Flow rule:
μԄ
†ઽ‫ ܔܘ‬ൌ †ɉ ൌ †’ ή ‫ܖ‬ (3)
μો
4. Hooke's law:
†ો ൌ  ۳ ‫ ׷ ܔ܍‬ሺ†ઽ െ †’ ή ‫ܖ‬ሻ (4)
5. Kinematic hardening:
୑

હ ൌ  ෍ હሺܑሻ
(5)
୧ୀଵ

†” ሺ୧ሻ
†હሺܑሻ ൌ  … ሺ୧ሻ ή ൣ” ሺ୧ሻ ‫ ܖ‬െ   ሺ୧ሻ હሺܑሻ ൧†’ ൅  હሺܑሻ ǡ ‹ ൌ ͳǡʹǡ ǥ ǡ 
” ሺ୧ሻ (6)
ሺ௜ሻ ஧ሺ೔ሻ
ฮߙ ฮ
 ሺ௜ሻ ൌ  ቆ ሺ௜ሻ ቇ
ȁ‫ ݎ‬ȁ
(7)

ሺ୧ሻ ሺ୧ሻ ሺ୧ሻ

ɖሺ୧ሻ ൌ  ɖ଴ ൅ ൫ɖ଴ ൅ ͲǤͳ൯൫…஧ െ ͳ൯ ή ൫ͳ െ  ห‫ۺ ׷ ܖ‬ሺܑሻ ห൯ (8)

ሺ୧ሻ ƒ஧
ɖ଴ ൌ ሺ୧ሻ  ή ൭ͳ ൅  ଶ൱ (9)
൫ͳ ൅ „஧  ୑಍ ൯
122 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128

ʹ ͳ ʹ ᇱሺ୧ሻ െ  ᇱሺ୧ାଵሻ
… ሺ୧ሻ ൌ  ඨ ୮୪ሺ౟ሻ ǡ ” ሺ୧ሻ ൌ  (10)
͵ɂ ͵ … ሺ୧ሻ

஢ሺ౟ሻି஢ሺ౟షభሻ ஢ሺభሻ ିఙಷ

ᇱሺ୧ሻ ൌ  ሺ౟ሻ ሺ౟షభሻ ǡ ᇱሺଵሻ ൌ  ሺభሻ , ‹ ൌ ʹǡ ǥ ǡ 
க౦ౢ ିக౦ౢ க౦ౢ (11)
6. Isotropic hardening:
ሺ୧ሻ ሺ୧ሻ
„ ή ൫”୘ െ ” ሺ୧ሻ ൯†’ห”୘ െ ” ሺ୧ሻ ห ൑ ͳ
†” ሺ୧ሻ ൌ  ൝ ሺ୧ሻ ሺ୧ሻ ሺ୧ሻ (12)
„ ή ൫”୘ െ ” ൯ห”୘ െ ” ሺ୧ሻ ห†’ห”୘ െ ” ሺ୧ሻ ห ൒ ͳ
ሺ୧ሻ

ሺ୧ሻ ሺ୧ሻ ሺ୧ሻ

ሺ୧ሻ ሺ୧ሻ ƒଵ ƒଶ ƒଷ
”୘ ൌ  ”ஶ ή ቆͳ ൅  ൅ ൅ ቇǡ ‹ ൌ Ͳǡ ͳǡ ʹǡ ǥ ǡ  (13)
ሺͳ ൅  „ଵ ’ሻଶ ሺͳ ൅  „ଶ ’ሻଶ ሺͳ ൅  „ଷ ’ሻଶ

ሺ଴ሻ ሺ଴ሻ ሺ଴ሻ

”ஶ ൌ  ”ஶǡ଴  ή ൣ ή “ ୒ ൫ ୑಍ ൯ ൅  ሺͳ െ ሻ ή  “ ୔ ൫ ୑಍ ൯൧ (14)
ሺ୧ሻ
ሺ୧ሻ ƒ୩
ሺ୧ሻ ሺ୧ሻ
”ஶ ൌ  ”ஶǡ଴  ή  “ ୔ ൫ ୑಍ ൯ ή  ቆͳ ൅  ቇǡ ‹ ൌ ͳǡ ʹǡ ǥ ǡ  (15)
ሺͳ ൅  „୩ ɖሺ୧ሻ ሻଶ
ሺ୧ሻ
ሺ୧ሻ ƒ୔
“ ୔ ൫ ୑಍ ൯ ൌ ͳ ൅  ଶǡ ‹ ൌ Ͳǡͳǡ ʹǡ ǥ ǡ  (16)
൫ͳ ൅  „୮  ୑಍ ൯
ƒ୒
“ ୒ ൫ ୑಍ ൯ ൌ  “ ୒ǡ଴  ή ൭ͳ ൅ ଶ൱ (17)
൫ͳ ൅  „୒  ୑಍ ൯
7. Memory area:
ன
† ୑಍ ൌ  ൣɄ ή  െ  ሺͳ െ ሻ ή ൫…୑಍  ୑಍ ൯ ൧ ή ‫’†ۄ כܖ  ׷ ܖۃ‬
(18)
8. Non-proportionality tensor according to Tanaka:

†୧୨୩୪ ൌ  …୘ ή ൫୧୨ ୩୪ െ  ୧୨୩୪ ൯†’ (19)

୩୪ ୧୨୩୪ ୧୨୬୫୬ౣ౤
 ൌ  …୅ ή ሺ୘ െ ሻ†’ǡ ୘ ൌ  ඨͳ െ (20)
୧୨୩୪ ୧୨୩୪

2.2. Results of the parameter identification

A good description of the material phenomena with the Döring model (10 back stress tensors) is given by
identification of 103 parameters. The model-specific calculation is performed with software realized in Fortran
language (Fig. 1). The main program of optimization minimizes the objective function. At the beginning of the
optimization task, a suitable starting vector must be chosen, so that the optimization process can proceed without
convergence problems. The optimization is carried out in five steps in the following order.

1. Discretization of the initial load curve and the stabilized stress-strain curve.

2. Division of the initial load curve in two parts at a specified strain limit.

3. Optimization of the first part with the strains greater than or equal to a specified strain limit.

‹ ˆଵ ሺšଵ ሻ ൌ ͵ˆ୊୐ଵ ሺšଵ ሻ ൅  ˆ୙୐ଵ ሺšଵ ሻ ൅ ˆ୰ୣ୪ୟ୶ଵ ሺšଵ ሻ

୶భ (21)
•Ǥ –Ǥšଵ  ‫  א‬ɖଵ

4. Optimization of the second part with the lower strains.

‹ ˆଶ ሺšଶ ሻ ൌ ͵ˆ୊୐ଶ ሺšଶ ሻ ൅  ˆ୙୐ଶ ሺšଶ ሻ ൅ ˆ୰ୣ୪ୟ୶ଶ ሺšଶ ሻ

୶మ (22)
•Ǥ –Ǥšଶ  ‫  א‬ɖଶ
 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128 123

5. Determination of the parameters for the cyclic creep.

‹ ˆଷ ሺšଷ ሻ
୶య (23)
•Ǥ –Ǥšଷ  ‫  א‬ɖଷ

The nonlinear optimization problem of main concern here is determining a vector of decision variables š୧ ‫א‬
Թ୬ ሺ‹ ൌ ͳǡ ʹǡ ͵ሻ, which represent the parameters in the Döring model. The solution is called optimal, when the objective
function ˆ୧ ሺǤ ሻǣԹ୬  ՜ Թ is minimized, while x is restricted and belongs to a defined feasible set ߯ ‫  ك‬Թ௡ , which is
defined by inequality constraints ߯ ൌ  ሼ‫ݔ‬௜ ߳Թ௡ ǣ݃௜ ሺ‫ݔ‬௜ ሻ  ൑ Ͳǡ ݃ሺǤ ሻǣԹ୬  ՜  Թ୫ ሽ. The objective functions ˆଵ ሺšଵ ሻand ˆଶ ሺšଶ ሻ consist
of three weighted parts. For example, the function ˆଵ ሺšଵሻ ൌ ͵ˆ୊୐ଵ ሺšଵሻ ൅  ˆ୙୐ଵ ሺšଵ ሻ ൅ ˆ୰ୣ୪ୟ୶ଵ ሺšଵሻ, where ˆ୊୐ଵ ሺšଵ ሻ describes least
squares based on the experimental data of the initial load curve and the simulated initial load curve. The function
ˆ୙୐ଵ ሺšଵ ሻ determines the method of least squares based on experimental data of the initial unload curve and the simulated
initial unload curve and ˆ୰ୣ୪ୟ୶ଵ ሺšଵ ሻ defines the method of least squares based on the experimental stress relaxation and
the simulated stress relaxation. The weight factor 3 in front of ˆ୊୐ଵ ሺšଵ ሻ and ˆ୊୐ଶ ሺšଶ ሻ emphasizes that this part of the
objective function has a particular importance.
Although the problem is stated in simple terms, its solution may become very difficult due the analytical properties of
the functions involved and the number of variables n and m of constrains and variables. Fig. 2, Fig. 3 and Fig. 4 display
the quality of the obtained material parameter set by comparison of the simulation with the experimental data.

Fig. 2. Initial monotonic stress-strain curve and initial unload curve (material 34CrNiMo6, simulation and experimental data)

Fig. 3. Stress-relaxation curves R = -1, (material 34CrNiMo6, simulation and experimental data)
124 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128

Fig. 4. Cyclic creep (material 34CrNiMo6, simulation and experimental data)

3. FE analysis

3.1. FE model

The constructed threaded connection model M16 is in accordance with DIN 13 [4] and the associated tolerance
classes 6H for nut threads and 6g for bolt threads are maintained. The FE model of the investigated screw joint is
realized as an axisymmetric model (2D) by using linear elements (CAX4) with full integration. The friction coefficient
between the nut and the bolt is defined by μ = 0.12 with the nonlinear contact algorithm "Penalty" (tangential
behaviour) and "Augmented Lagrange" (normal behaviour). The result of a convergence analysis of the equivalent
stress and equivalent strain is that in the notch root (fillet radius) of the bolt should be defined 30 elements, along the
screw flank 38 elements and along the nut flank 23 elements. A comparison of a 2D-model and 3D-model (bolted
connections) is given in [6]. Results of this study recommend that for the preliminary investigation, a 2D modeling is
sufficient for the application of the local concept. The two elastic-plastic material models are used in the FEM analysis
with 10 back stress tensors (additive decomposition of the total back stress tensor in 10 back stress tensors). The
parameters were determined by experimental results (material 34CrNiMo6). Both parts, the screw and nut are defined
with the same material 34CrNiMo6.

3.2. FE results - local stress-strain curves

Fig. 5 and Fig. 6 describe the local situation in the first load-bearing thread of the bolt, evaluated at a node with
the maximal stress-concentration σ1. The results with the Döring model describe the transition from the initial load
curve to the transient material behaviour. A combination of stress relaxation and cyclic creep is observed as an answer
on the cyclic load ሺɐ୫ ് Ͳሻ. The material model according to Chaboche as implemented in the commercial software
Abaqus is not able to describe the initial load curve and the transition to the transient material behaviour, because the
Chaboche parameters are determined by optimization based on the cyclic stabilized stress-strain curve. Furthermore,
the results with the Chaboche model show a larger stress relaxation and higher cyclic creep values. These differences
between the Döring and Chaboche model increase with higher load ratios (Fig. 6). The benefit of the ability to describe
the transition from the initial load curve to the cyclic material behaviour is promising a numerical approach to describe
 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128 125

the reduction of the preload stress and other useful investigations of threaded connections. The dissimilar description
of the stress-strain-curves by the two material models will cause different results in the fatigue assessment.

Fig. 5. Local stress-strain curves – first load-bearing bolt thread – R = 0.06

Fig. 6. Local stress-strain curves – first load-bearing bolt thread – R = 0.71

4. Damage Assessment applying the local strain approach

4.1. Local concept by applying the damage parameters PSWT and PJ

One of the reasons for the non-applicability of using the nominal approach concepts is that the determination of
the nominal cross-section in general threaded connections is often not possible. At this point the local concept seems
to be promising, but there is only little experience in the application for threaded connections. The damage assessment,
126 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128

according to the local strain approach, is done by the damage parameters PJ and PSWT. Further details on the calculation
procedure can be found in [7] and [9]. The damage parameter PSWT [8] is defined by equation (24).

ୗ୛୘ ൌ  ඥሺɐୟ ൅  ɐ୫ ሻ ή ɂୟ ή  (24)

The results show that there are significant differences between the damage values after 55 load cycles, if one uses
different material models. The stress-strain curves describe with the Chaboche model a different stress relaxation and
cyclic creep behaviour compared with the results using the Döring model. The differences between the damage values
in the case study 1 (R = 0.06) are lower compared with the case study 2 (R = 0.71). In the case study 1, the mean-
stress is converging against a similar value with increasing load cycles, and the hysteresis show a similar shape, while
the cyclic creep is higher by using the Chaboche model. If the load cycle number increases until the stabilized state is
achieved, the damage ratio will increase in both case studies, especially in the case of the higher load ratio. Solely
from the definition (24) of the damage parameter PSWT and the contemplation of the stress-strain curves (Fig. 6) can
be seen that the application of the Chaboche model leads to lower damage values by the increased relaxation compared
with the advanced material model. The mean stress values are lower and are decreasing continuously until the stable
state. On the other hand the mean stress of the advanced material model decreases slightly. The parameter PSWT is the
most widely used damage parameter, though predictions after PJ-models are preferred, because the physical model is
more elaborate, especially the derivation is based on the cyclic J-integral [9, 10].

ȟɐଶୣ୤୤ ͳǤͲʹ
୎ ൌ  ቆͳǤʹͶ ή ൅ ή ȟɐୣ୤୤ ή ȟɂ୮୪ǡୣ୤୤ ቇ (25)
 ξԢ

According to experimental results in [11] by applying the damage parameter PJ for life prediction of threaded
connections are slightly more conservative in the fatigue endurance limit compared with the results by applying PSWT.
The properties of the damage parameter PJ and its description in the context of the local concept are explained in [9].
In this paper the results of the local strain approach based on the damage parameter P J are shown in Table 1. It should
be mentioned, that the presented case studies do not represent the stabilized state (here: 111 cyclic load steps).

Table 1 Damage assessment, 111 load cycles, based on the damage parameter PJ

Case study 1 Case study 2

Material model Döring Chaboche Material model Döring Chaboche
Load ratio R = 0.06 R = 0.06 Load ratio R = 0.71 R = 0.71
Damage value 55
Damage value
55
¦ Di 0.0281731 0.01300831 ¦ Di 0.00068371 0.00020741
i 1 i 1

55 55

¦ Di, Döring ¦ Di,Döring

i 1 i 1 3.296
2.166
Damage ratio 55 Damage ratio 55

¦ Di, Chaboche ¦ Di, Chaboche

i 1 i 1

4.2. Simplified method according to Schneider applying the damage parameter PJ

The case study 3 is concerned with the comparison of the simplified method according to Schneider [11] and the
results of the damage assessment by using the advanced material model (Table 2). Both variants perform almost the
same result caused by the 111 load steps. The calculation with the Döring model takes the stress relaxation and cyclic
creep into account, which is in the case of the high load ratio (R = 0.71) relatively small, hence the damage ratio is
nearly 1 (Table 2). The study 3 ( R = 0.71) highlights that the simplified method of Schneider leads to conservative
 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128 127

results with significantly less effort compared with the using of the advanced material model. This simplified method
is not intended for the evaluation of high preloaded threaded connections with low load ratios R.

Table 2 Comparison of the simplified method according to Schneider with the damage assessment by applying the advanced material model

Case study 3

Material model Döring Simplified method / Schneider

Load ratio R = 0.71 R = 0.71

Damage value 0.00068371 0.0006902

5. Conclusions and outlook

The damage assessment based on the local concept depends, among other influences, in particular from the correct
description of the local stress-strain curves. The material models implemented in a commercial software are not
satisfactory to delineate the cyclic plasticity phenomena, so that the applying of an advanced material model is
promising a new access. The effort by using the advanced material model is justified by a new set of possibilities to
explore numerically the physical effects in context with threaded connections. One of these phenomena are, for
example, the transition of the initial monotonic stress-strain curve (first load curve) to the cyclic material behaviour.
The results of the simulation are validated with special notched specimen. These experiments were not presented in
this paper, since only a limited number of completed experiments are available. However, the first validations of the
simulation by the experimental data are very satisfactory. The investigations will continue. The simplified method of
Schneider is recommended in order to avoid the high numerical effort for fatigue life assessments of high preloaded
bolted connections with high mean stresses using advanced material models. The next goal is to identify a simplified
method for threaded connections under low mean stress and low preload, too, where the simplified method is not
recommended. These findings from the research results open a new access to studies of other components, which are
related to the fatigue life predictions and local concepts.

Acknowledgements

The authors express their sincere gratitude to the German Federal Ministry of Economics and Technology
(Bundesministerium für Wirtschaft und Technologie, BMWi) for financial support via AiF (Arbeitsgemeinschaft
industrieller Forschungsvereinigungen „Otto von Guericke“ e.V.) of the research project „Örtliche Bewertung der
Schwingfestigkeit von Gewindeverbindungen II" under the IGF project no. IGF 16920 N.

References

[1] VDI-Richtlinie 2230, Systematische Berechnung hochbeanspruchter Schraubenverbindungen, 2003.

[2] Eurocode 3, Bemessung und Konstruktion von Stahlbauten, 2001.
[3] Forschungskuratorium Maschinenbau (FKM), Analytical Strength Assessment, VDMA Verlag, 2003.
[4] Deutsches Institut für Normung, DIN 13, Grenzmaße für Bolzengewinde, Teil 20 bis 22, Berlin, 1999.
[5] Döring, R., Hoffmeyer, J., Seeger, T., Vormwald, M.: A plasticity model calculating stress-strain under multiaxial nonproportional cyclic
loading, Computational Materials Science 28, 587-596, 2003.
[6] Seybold, R.: Finite-Elemente-Simulation örtlicher Beanspruchungen in Schraubengewinden. PhD-Thesis Technische Universität Darmstadt,
2005.
[7] Dowling, N.E.: Local Strain approach to fatigue. In: Comprehensive structural integrity, Elsevier, vol. 4, pp. 77-94, 2003.
128 D. Panic et al. / Procedia Engineering 74 (2014) 119 – 128

[8] Smith, K.N., Watson, P., Topper, T.H.: A Stress-Strain Function for the Fatigue of Materials. Int. Journal of fatigue S767-778, 1970.
[9] Radaj, D., Vormwald, M., Ermüdungsfestigkeit Grundlagen für Ingenieure, Springer, ISBN 978-3-540-71458-3, 2007.
[10] Vormwald, M., Heuler, P., Seeger, T.: A fracture mechanics based model for cumulative damage assessment as part of fatigue life
prediction. In: H. Amzallag, (Ed.), Automation in Fatigue and Fracture Testing and Analysis, ASTM STP 1231, pp. 219-231, 1994
[11] Schneider, R.: Örtliche Bewertung der Schwingfestigkeit von Gewindeverbindungen. Publication of the „Institut für Werkstoffkunde“ of the
Technische Universität Darmstadt, 2010.