Journal of Materials Processing Technology 119 (2001) 158±164
Fundamental aspects for the evaluation of the
fatigue behaviour of cold forging tools
B. Falk*, U. Engel, M. Geiger
Chair of Manufacturing Technology, University of Erlangen-Nuremberg, Egerlandstrasse 11, D-91058 Erlangen, Germany
Abstract
The service life of cold forging tools is mainly limited by high-cycle fatigue (HCF). The prediction of the expectable tool life is an
essential precondition for an optimisation of the tool resistance against fatigue and the evaluation of optimisation measures during early
design stages. In this context, the numerical process simulation (primarily the ®nite element method) enables the calculation of the load
conditions at the critical tool areas. Nevertheless, many investigations show that numerous fundamental aspects must be taken into
consideration to obtain calculated lifetime values with suf®cient accuracy. One goal of this publication is to discuss aspects concerning the
®nite element models and the evaluation of several in¯uences concerning the lifetime calculation. Finally, a close connection between the
fatigue strength of powder-metallurgical tool steels and the size of the stressed volume has been quanti®ed based on four-point bending
tests. In the framework of the investigations presented, these results have been applied to carry out an advanced volume-dependent lifetime
calculation. # 2001 Elsevier Science B.V. All rights reserved.
Keywords: Cold forging; Finite elements; High-cycle fatigue
1. Introduction context, the prediction of the reachable tool lifetime in early
design stages is in the foreground. Numerous examples from
For the production of cold forged parts with near-net- industry show in a clear manner that practicable lifetime
shape attributes, the quality of the tool system is responsible calculations for cold forging tools should be focused on the
for an essential portion of costs for the ®nished components. accurate prediction of the number of load cycles until the
To ensure a production with high ef®ciency, two funda- initiation of fatigue cracks [1]. The additional calculation of
mental demands have to be ful®lled. On the one hand, a high the crack propagation period does not normally lead to a
lifetime of the applied tool system (mainly limited by the signi®cant increase in the prediction accuracy.
occurrence of wear and high-cycle fatigue (HCF)) should be Furthermore, the results gained from previous investiga-
aspired to achieve reduced tool expenses. On the other hand, tions show that lifetime predictions based on ®nite element
the lifetime scatter has to be reduced to guarantee an analyses are characterised by unacceptable inaccuracies [2].
increased reliability of the cold forging process itself. One reason for this fact can be ascribed to the process
Due to shorter time periods for tool development, the parameters which will be incorporated into the simulation
requirements mentioned above can only be accomplished model. In many cases, these parameters are quanti®ed
with the application of the numerical process simulation insuf®ciently or the description does not consider the sta-
(e.g. ®nite element techniques). Nevertheless, if numerical tistical character of the process parameters itself (e.g. scatter
methods are used to obtain accurate information about the of ¯ow stress of the workpiece material). In practice, a
process-dependent load, which represents the basis for a simple possibility for the estimation of effects caused by the
subsequent lifetime calculation, different aspects concern- scatter of in¯uence parameters is given by the so-called
ing the simulation and the properties of the tool system have `worst case' and `best case scenarios'. One further reason for
to be considered. Therefore, one intention of the investiga- inaccurate results (which are the basis for the following
tions described below is to illustrate the possibility for the lifetime calculation procedures) gained from the numerical
application of numerical process simulation methods. In this process simulation is affected substantially by the choice of
the ®nite element mesh. Especially for three-dimensional
*
Corresponding author. Tel.: 49-9131-85-8317;
material ¯ow analysis, the applied element type (e.g. ®rst-
fax: 49-9131-36-403. order tetrahedral or ®rst-order hexahedral elements) affects
E-mail address: fa@[Link] (B. Falk). the results. First-order tetrahedral elements, in particular, are
0924-0136/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 4 - 0 1 3 6 ( 0 1 ) 0 0 9 3 4 - 7
B. Falk et al. / Journal of Materials Processing Technology 119 (2001) 158±164 159
overly stiff and should be avoided whenever possible [3]. 2.1. Local stress and local strain approach
Unfortunately, powerful mesh generators for the automatic
generation and the remeshing of hexahedral meshes are The most employed damage approach is given by the
indeed not wide-spread in commercial FEM-codes. There- evaluation of the stress components which are assumed to be
fore, it is nearly impossible to assess the numerical effects responsible for the initiation of fatigue cracks. Especially for
due to the employment of tetrahedral elements. Based on brittle materials like high-speed tool steels (hardness >56
this fact, the evaluation of different damage concepts or HRC), the normal stress hypothesis can be applied to trace
signi®cant in¯uence parameters is only feasible for applica- back the fatigue damage to the maximum principal stress
tions or examples with hexahedral elements or two-dimen- [5]. This advantage provides a direct correlation between
sional discretisations, which have been applied for the load and lifetime data gained from fatigue diagrams. In this
investigations discussed in the following. context, the consideration of mean stress components sm
Experimental results show, that accurate lifetime calcula- becomes extremely important when the regarded cold
tions should also consider the fatigue strength of the applied forging tool fails due to HCF. According to MORROW,
tool material as a function of the stressed volume [4]. the so-called local strain approach, which depends on the
Especially for the consideration of the volume-dependence strain amplitude 12 De, considers the in¯uence of mean stress
of tool materials, the ®nite element code DEFORMTM has effects for uni-axial load conditions (Eq. (1)). Nevertheless,
been extended by a user-de®ned subroutine to enable an this kind of view leads to dif®culties for the treatment
automatic lifetime calculation. of multi-axial load conditions. Due to this circumstance,
According to the facts described above, the objectives MORROW's concept should only be applied when the
of the following investigations are: (1) to give a condensed failure cause can be assigned to one dominating strain
overview of damage concepts which can be appended easily component [6].
to FEM-codes; (2) the application of a volume-based 1=b
damage approach for lifetime predictions; (3) besides some DeE
2Nf (1)
aspects concerning the interpretation of the results gained 2 s0f sm
from the numerical process simulation, also the comparison
of lifetime approaches for an industrial tool system.
2.2. Local energy approach
2. Lifetime prediction Ð analytical procedures According to [7], one can assume that such kind of
damage concepts are limited to few problems and may lead
In general, deterministic lifetime approaches are based on only in speci®c cases to lifetime calculations with suf®cient
the use of damage concepts which describe a mathematical accuracy. In order to avoid this disadvantage for the inves-
relationship between the maximum load and the number of tigation of cold forging tools, which are mostly characterised
load cycles until the initiation of fatigue cracks for low-cycle by complex load conditions, an advanced energy-based
fatigue and HCF. The material parameters required are damage concept should be used. Taking into account the
ascertained by means of fatigue tests (e.g. uni-axial reasonable assumption that plastic strain partitions are neg-
strain-controlled fatigue test). The choice of the most sui- ligible for HCF the modi®ed elastic strain energy density
table damage parameter which allows the correct reduction DW e represents a suitable damage parameter (Eq. (2)).
of three-dimensional load conditions to an equivalent one-
dimensional value presents one fundamental dif®culty DW e 12 12 Ds sm 12 De em (2)
for the application of fatigue damage concepts. In order DW h 38 Dsh Deh (3)
to use a computer-based lifetime approach and to choose
1=2b
suitable damage parameters, the following aspects seem to 2EDW e
be signi®cant: 2Nf (4)
s0f 2
The determination of the existing load conditions in the
Nevertheless, DW e contains also the hydrostatic strain
maximum loaded region of the tool, e.g. uni-axial or
energy density DW h (Eq. (3)) which takes no influence on
multi-axial load conditions.
the damage behaviour. This leads to the conclusion that only
The availability of reliable fatigue data for the regarded
the deviatoric partition of DW e , the so-called effective
tool material. Moreover, detailed information about the
strain energy density DW eff, is responsible for the fatigue
test conditions (e.g. hydrostatic stress partition).
crack initiation. In this context, it is essential to distinguish
Information about the accuracy of the load values gained
between the hydrostatic stress conditions of the real cold
from the numerical process simulation to ensure that the
forging tools and those existing in the test specimens which
calculation of damage parameters is as perfect as possible.
have been applied to determine the material parameters for
Considering these presuppositions, a successful adoption of the description of the fatigue behaviour (e.g. the cyclic stress
different damage approaches becomes possible. coefficient s0f and the cyclic stress exponent b). Due to this
160 B. Falk et al. / Journal of Materials Processing Technology 119 (2001) 158±164
Fig. 1. Correct application of energy-based damage parameters.
fact, it becomes obvious that a lifetime prediction based on die of the tool system fails because of the initiation of
DW e may only lead to convenient results if the differences surface cracks in the area of HCF. The investigation con-
of the hydrostatic partitions between test specimens and the cerning the failure cause could also show that no surface
real tool are taken into consideration. Fig. 1 gives a brief modi®cation (e.g. increase in roughness in the critical area)
overview of a practicable method for the correct determina- took place during the entire service life. Especially, the last
tion of DW e . Finally, the correlation between the number presupposition is extremely important for the assessment of
of load cycles 2Nf and the maximum load, given by DW e, the prediction accuracy of different lifetime approaches.
is represented by Eq. (4). Fig. 2 shows the die insert and the damaged tool component.
This component failed after an average value of 7900 load
cycles. The recording of the lifetime data in industry led to
3. Application of lifetime procedures minimum lifetime values of 5500 and maximum lifetime
values of 11 100, respectively. That means the lifetime
In the following, the application of the local stress scatter is small compared to other examples. Moreover, in
approach is explained. For this purpose, a tool system which Fig. 2 the chosen ®nite element model of the tool system is
is used for the commercial production of axisymmetric parts given to present the mesh densities and the entire construc-
for automotive dumping systems is taken as an example. The tion of the tool system. Especially, in the critical die area, a
Fig. 2. Tool break down caused by HCF and finite element model for the simulation-based lifetime prediction.
B. Falk et al. / Journal of Materials Processing Technology 119 (2001) 158±164 161
high mesh density has been generated to achieve more the presupposition of suf®cient mesh densities in the areas
reliable results. The material behaviour of the billet is where high stress gradients occur, the calculation of nodal
assumed to be rigid-viscoplastic. Based on the justi®ed values, which requires additionally the implementation of a
presupposition that no plastic strains are observable, a user-de®ned subroutine, is negligible.
linear-elastic material behaviour has been introduced for
the tool components. 3.2. Lifetime calculation
3.1. Finite element analysis The results gained from the numerical process simulation
show clearly that the load conditions in the critical die area
The material ¯ow analysis and the calculation of the tool are rather complex. Moreover, the maximum principal stress
load distribution has been carried out by means of the FEM- takes only small values. The application of this damage
code DEFORMTM2D. For the calculation of energetic parameter would lead to the wrong conclusion that no
damage parameters, the FEM-code has been extended by fatigue-caused damage is going to occur. Therefore, the
a user-de®ned subroutine. As a major result of the numerical lifetime prediction based on the stress conditions should
process simulation, Fig. 3 shows the distribution of the concentrate on the effective stress. In general, the use of
modi®ed elastic strain energy density DW e and the effec- energetic damage parameters is also possible, but less pre-
tive stress seff in the fatigue critical area of the die. Both diction accuracy should be expected. One reason for this
damage parameters, DW e and seff, show distinct max- assumption is given by the procedure applied for the calcu-
imum values in the area where the initiation of fatigue cracks lation of the energetic damage parameters. Using the max-
starts. The evaluation of the maximum principal stress, imum value of 1994 MPa of the effective stress which has
which is absolutely identical to the axial stress component been observed at the critical tool area, the lifetime calcula-
in the critical die area, has shown that this component is not tion leads to a value of 8935 load cycles implying a failure
responsible for the fatigue crack initiation. Due to this fact, probability P of 50% (cf. Fig. 4). As mentioned above, the
the tool damage is mainly affected by the shear stress real lifetime values are ranged between 5500 and 11 100
component, which is made plain by the distribution of load cycles. This result shows a fairly good agreement
DW e . For the subsequent lifetime calculation, it seems between the predicted lifetime and the real lifetimes. Never-
to make sense to transfer the elemental values of the damage theless, there are several fundamental effects which may
or load parameters to the boundary nodes of the ®nite prevent such accurate results. One is given by the volume-
element model to achieve results with higher accuracy. In dependence of the fatigue properties of tool steels with
this case, this kind of method enables even a load reduction higher hardness values. In the following, the theoretical
of about only 1.05%. Nevertheless, this circumstance is methodology and the application of the volume-based life-
equivalent to nearly 16% higher lifetime values. Under time approach is explained in detail.
Fig. 3. Evaluation of the tool load distribution based on different damage parameters.
162 B. Falk et al. / Journal of Materials Processing Technology 119 (2001) 158±164
Fig. 4. Correlation between the maximum bending stress sb,max and number of load cycles Nf ascertained by four-point bending test.
4. Volume-based lifetime approach especially for tool components with higher hardness levels.
The tool steel AISI M2 (1.3343) shows a comparable fatigue
The damage approaches mentioned above assume only a behaviour (cf. Table 1).
dependence between the lifetime and the maximum tool The tool material AISI M3 has been tested for the two
load, which can be quanti®ed by the use of suitable damage different hardness values (60 and 64 HRC), both of which
parameters. As an example, the results in Fig. 5 show clearly represent typical heat treatments for cold forging dies. The
the in¯uence of the stressed volume size on the fatigue presentation in Fig. 5a shows the number of load cycles Nf
strength of the powder-metallurgical high-speed steel AISI vs. the maximum bending stress sb,max for three different
M3 (1.3344; ASP 23), which comes in the foreground for failure probabilities P (5, 50, 95%). In Fig. 5b, the applied
high hardness values (e.g. 64 HRC). For smaller hardness experimental set-ups I and II for carrying out the four-point
values (60 HRC), a considerable growth of the lifetime bending tests are presented. In order to provide the variation
scatter can be observed. The lifetime values for a failure of the stressed volume size of the tested rectangular speci-
probability of P 50% are only slightly higher. Therefore, men (cross-section size: bh 9:0 6:0 mm2 ), the bending
it seems to be necessary to consider these circumstances rolls distance has been modi®ed as shown. The lifetime
Fig. 5. Influence of the stressed volume size on the fatigue strength: (a) experimental results for the tool material AISI M3 (1.3344; ASP 23); (b)
experimental set-up.
B. Falk et al. / Journal of Materials Processing Technology 119 (2001) 158±164 163
Table 1 Table 2
Volume-dependence of the tool material AISI M2 (60 HRC) Volume-based lifetime concept Ð material parameters for a failure
probability P 50%
Failure probability P (%)
Reference Reference volume, Reference stress, Gradient, Weibull
5 50 95 lifetime, Nref Vref (mm3) sref (MPa) m (MPa) slope, e
Number of load cycles Nf 10 000 540 1979 131 28 276
Experimental set-up I 3645 14183 55191
Experimental set-up II 4136 20917 105784
values of the specimens for different failure probabilities of take essential in¯uence on the accuracy of the lifetime
the tool steel AISI M2 (60 HRC), evaluated by the use of the calculation, because the applied ®nite element mesh ®ts
arcsin HP-method, are summarised in Table 1. In general, it the load gradients with smaller elements especially in the
is possible to implement a methodology in FEM-codes to critical area (cf. Fig. 3). The lifetime calculation based on
consider the phenomenon of volume-dependence. Using the volume approach shows, moreover, that the predicted
Eq. (5), the lifetime of one single ®nite element Nele can solution converges against the value based on the local stress
be calculated. In this case, the failure probability is de®ned approach for increasing values of `e'. That means for less
implicitly by the reference lifetime Nref and the stressed volume-dependence of the tool material. This fact leads to
volume Vref of the bending specimen. According to [4], it is the conclusion that the supposed volume-dependence gained
convenient to introduce the concept of a reference volume from the experimental work is mainly caused by surface
Vref de®ned by an equivalence in elastic strain energy imperfections. One can assume that the determined effect
(Eq. (6)). The parameter we is the elastic strain energy comes in the foreground for tool components with a nearly
and wmax
e represents its maximum value at the whole struc- homogeneous distribution of the critical load close to the
ture. For the four-point bending test, Vref can be calculated as fatigue endangered surface areas. In contrast, the tool system
de®ned in Eq. (7). investigated is characterised by load conditions which are
The parameter e represents the Weibull slope and m is the concentrated on a small area.
gradient of a linear function which describes the correlation
between the bending stress and number of load cycles (cf.
also Fig. 4). According to Eq. (8), the lifetime of the whole 5. Concluding remarks
component Ncomp can be calculated from the lifetime values
Nele of the single ®nite elements. The service life investigation for cold forging tools pre-
1=e sented in this paper has shown that comparatively simple
Nref Vref damage approaches could result in lifetime predictions with
Nele (5)
exp sref sele =m Vele suf®cient accuracy. For the tool system investigated already,
R the effective stress leads to these accurate predictions. One
we dV
Vref V max (6) general presupposition is the generation of a suitable ®nite
we element model. A further general presupposition which has
Vref 13 la bh (7) been guaranteed in the framework of this deterministic
investigation is the availability of precise material data to
As a concluding remark, it should be mentioned that this describe the fatigue strength. These material data should be
kind of analysis is limited to stress levels below the yield combined with information about the failure probability in
point. It can be used especially for HCF where stresses are each case. The volume-based damage approach, which has
lower and the occurrence of failure is less deterministic and been applied in this presentation, leads to less accurate
more probabilistic [8]. results for the example discussed. Nevertheless, it can be
e X n expected that this concept features increasing advantages for
1 1 e
(8) practical problems which are characterised by tool materials
Ncomp ele1
Nele
with higher hardness values. Furthermore, the data of the
Finally, a lifetime prediction which depends on the stressed parameter `e' ascertained from four-point bending test,
volume has been carried out to compare the predicted results which should describe the speci®c volume-dependence of
with those ones obtained on basis of the local stress a brittle tool material, seems to be in¯uenced also by surface
approach. For the employment of the volume-based damage imperfections.
concept, a lifetime of 20 584 has been calculated for a failure
probability P of 50%. The applied values of the parameters
required (cf. Eqs. (5)±(8)) are summarised in Table 2. Acknowledgements
For each element, or the respective volume which is
de®ned by each element, homogeneous load conditions The investigations presented in this publication were
are presumed. According to [8], this assumption does not carried out within the research project Development of a
164 B. Falk et al. / Journal of Materials Processing Technology 119 (2001) 158±164
Software Module for the Determination of Tool Life in Cold [3] ABAQUS/Standard User's Manual, Vol. II, Version 5.6, Hibbitt,
Forging in the framework of the Bavarian Joint Research Karlson & Sorensen, Inc., pp. 14.1.1±14.1.2.
[4] J. Lemaitre, A Course on Damage Mechanics, 2nd Edition, Springer,
Program for simulation technologies (FORSIM), funded by Berlin, 1996.
the Bavarian Research Foundation and by industry. [5] H. Berns, Neuere Entwicklungen bei Werkzeugwerkstoffen der
Kaltmassivumformung, VDI-Berichte, Vol. 810, VDI-Verlag, DuÈssel-
dorf, 1990, pp. 77±86 (in German).
[6] G. Sines, G. Ohgi, Fatigue criteria under combined stresses or strains,
References J. Eng. Mater. Technol. 103 (4) (1981) 82±90.
[7] H. Zenner, R. Heidenreich, R. Richter, Bewertung von Festigkeitshy-
[1] M. Geiger, B. Falk, U. Engel, Prediction of service life of cold forging pothesen fur kombinierte statische und schwingende sowie synchron
tools failured by fatigue, Prod. Eng. VI (1999) 2. schwingende Beanspruchung, Z. Werkstofftech. 14 (1983) 391±406.
[2] B. Falk, U. Engel, M. Geiger, Estimation of tool life in bulk metal [8] R. August, E.V. Zaretsky, Incorporating finite element analysis into
forming based on different failure concepts, J. Mater. Process. component life and reliability, Reliability, Stress Analysis and Failure
Technol. 80/81 (1998) 602±607. Prevention, DE-Vol. 30, ASME, New York, 1991, pp. 71±77.
