Sie sind auf Seite 1von 11

Available online at www.sciencedirect.

com

ScienceDirect
Procedia Engineering 133 (2015) 309 319

6th Fatigue Design conference, Fatigue Design 2015

FKM Guideline : strengths, limitations and experimental validation


Samuel Kchlin*
Emerson Leroy Somer, Engineering & Development, CS 10015, 16915 Angoulme

Abstract

The FKM guideline for analytical strength assessment of mechanical components has gained an increasing interest for the last
decade in the industry, because it describes a general procedure directly applicable in an industrial design office. It allows in
particular a sound interpretation of finite element analysis, making an efficient bridge between stress results and a trustworthy
safety margin.
The guideline unfortunately doesnt give any comparison to experimental results, requesting from the user a blind confidence in
the procedure and in the included (indeed extensive) database.
Experimental comparison on a cast iron sheave and on mild steel laminate parts are presented in the following, in order to
emphasize on one hand the power of the methodology, and on the other hand specific limitations which cannot be suspected at
first sight. Based on these examples, a comparison with other well-known critera is eventually discussed, showing the efficiency
of the FKM procedure in the product design stage.

2015 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
2015 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review
Peer-review under
under responsibility
responsibility of CETIM.
of CETIM

Keywords: FKM-Richtlinie; fatigue assessment; analytical strength assessment; safety factor; experimental comparison; grey cast iron;

1. Introduction

Mechanical dimensioning has greatly benefitted from finite element simulation tools. Nowadays, getting a
realistic stress value in a structure is not always straigthforward, and requires an adequate modeling, but is generally
well mastered by engineers. This is however only the first half of the task, since a stress value must eventually be

* Corresponding author. Tel.: +33 5 45 64 54 87


E-mail address: samuel.koechlin@emerson.com

1877-7058 2015 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of CETIM
doi:10.1016/j.proeng.2015.12.674
310 Samuel Kchlin / Procedia Engineering 133 (2015) 309 319

compared with some strength limit. For such an evaluation, the engineer is left to his own judgement, which induces
him very often to make inappropriate comparisons with general material tables or to follow a rule of thumb.
The main questions arising in this process are :
x which kind of stress must be used : principal stress, Von Mises stress, etc ?
x shall we ignore finite element red spots, i.e. localized stress peaks?
x what about non-linear material behaviour?
x how to determine influence factors : size, roughness, heat treatment, etc
x how to take into account a limited number of cycles ?
x safety margin definition is not unique, which one shall we take depending on load possible variation?
x which value for the safety factor ? Something between 1 and much more ?

To deal with these questions, a few guidelines exist for some specific applications, but no general one for
machine parts in the mechanical industry, except the FKM guideline (reference [5], english version of the german
original). It appeared in Germany in the 90s, it is now at its 6th edition, and offers a structured and detailed
framework for a sound mechanical strength assessment.

2. Main features of the FKM guideline

The guidline makes a double distinction: between static and fatigue assessment on one hand, and between use of
nominal stress (typically coming from analytical calculation) or local stress (usually given by FEA) on the other
hand.
Basically, strength assessment needs only linear elastic FEA results. However, a partial yielding is allowed in
static assessment: except for simple geometries, it needs therefore a basic elasto-plastic calculation
Each type of stress (e.g. bending and torsion) is treated individually. The combination takes place in the last stage
of the procedure only, on individual degrees of utilization .
Stress gradient around hot spots is taken into account and smoothes out the local peak stresses in fatigue
calculation.
Fatigue limite is given as a ratio of ultimate static strength, for each type of material
Not the least feature of the guideline is an extended material database, along with general S-N curves for several
classes of materials, in such a way that the engineer has all needed material values at hand.
Some restrictions are clearly mentioned: non-proportional stresses are beyond the scope of the guideline, except
for synchronous stresses (e.g. shaft with constant torsion and rotating bending), which are treated in a specific way
with an equivalent mean stress.
The various possible definitions of the maximum allowable stress (i.e. the safety margin definition) are integrated
in the mean stress factor calculation (KAK). Different safety factor levels are set depending on failure criticity and
other parameters.
The guideline is clearly devised in the spirit of a standard, so as to give a common basis for mechanical
dimensioning. This brings in itself significant value to design offices, instead of having as many strength calculation
methods as calculation engineers, each giving a different result.
What clearly lacks in the document (but may for sure need an additional volume) is a comparison to experimental
cases. Reference [6] mentions a large amount of experimental comparisons, but they have not been published
outside german research groups. It is always good practice to consider cautiously even state-of-the-art calculation
results, especially when one wants to dispense with costly experimental validation, which is the purpose of such a
guideline. In the following are therefore presented some experimental comparisons in order to evaluate qualities and
weaknesses of the method.
Samuel Kchlin / Procedia Engineering 133 (2015) 309 319 311

3. Static strength assessment: example of a cast iron sheave

3.1. Industrial problem

On elevator motors, a grey cast iron sheave monted on the motor shaft transmits the movement to the cabin through
the cable. The sheave itself and its connexion to the shaft are of chief concern regarding security. On one hand, the
full torque must always be transmitted (since the key works only in case of emergency braking); on the other hand,
the stress level in the sheave due to shaft conical fit must never reach the material strength limit.

3.2. Mechanical model

Material is a grey cast iron EN-GJL250. A 3D finite element model taking into account the whole geometry with the
conical fit was compared to a simple 2D plain strain model of the hub (Fig. 1) : it appeared that the 2D model would
give a better insight into the inexpected behaviour of the part, without loosing much precision.
A well known feature of grey cast iron is its non-linear behaviour, even for very low stresses [7]: the tensile curve
does not show any linear part. This is taken into account either in the finite element material model, or indirectly in
the FKM guidline through a non-linear correction factor KNL, which is supposed to adjust the linear elastic stress to
a more realistic value.

Fig.1. Sheave geometry and 2D finite element model (1 st principal stress with 0.2mm diametral interference)

Fig.2. Example of sheave broken around a hole


312 Samuel Kchlin / Procedia Engineering 133 (2015) 309 319

3.3. Experimental results

Shaft insertion tests were conducted on several sheaves, and led in each case to a crack located around an extraction
hole (Fig. 2).

3.4. Calculation models

FKM takes the local stress values of a linear elastic model, and correct it in 3 ways :
x non linear behaviour of grey cast iron
x correction according to stress type : compressive strength bonus for cast meterial (iron and
aluminum), and shear strength correction
x allowable yielding, depending on an acceptable plastic strain and the degree of multiaxiality
For grey cast iron, the evaluation follows the Rankine criterion combined, since the last release, with a limit on the
hydrostatic stress.

An other popular method was also applied for comparison: the critical distance method [8] evaluates the stress not at
the hot spot, but at a given distance of it, without taking into account any local yielding. The evaluation distance is:

where Kc is the fracture toughness and lim the tensile strength for a smooth and non cracked specimen : depending
on the effective cracking micromechanism, this value can be the usual tensile strength Rm or a higher value. We
take here the Rm and Kc values given in the standard [4]. With Kc=632 MPa.mm1/2 and Rm= 250MPa we get a
critical distance L=2mm, which means the evaluation must be done at 1mm from the edge.

Fig.3. Local stress field calculated with a linear elastic model : between 0.2 and 1.3mm from the edge,
stress near the threaded hole is larger than near the fillet
Samuel Kchlin / Procedia Engineering 133 (2015) 309 319 313

3.5. Discussion

Since FKM bases its calculation on the local stress, it gives the most critical point in the keyway fillet, where the
stress is the largest. Moreover, since we have a plain strain situation the degree of multiaxiality is 0.49 and we thus
have a plastic notch factor of 1.03 : this means that very little plastic deformation is tolerated. The comparison is
then made with the conventional yield stress Rp0,1 , that is 165MPa, taking into account a non-linear correction of
9% (KNL=1.1) and a size factor of 0.83. We get eventually an extremely pessimistic estimation of the part strength.
The critical distance evaluation, on the contrary, gives a more realistic evaluation, explaining the failure near the
threaded hole (see Fig. 3). It is not clear however if the stress limit is equal to the conventional tensile strength;
according to [8], it may be higher for such a material.
We can conclude from this example that the guideline is very conservative for grey cast iron parts, when very
localized stress concentrations arise. Two additional point must be underlined :
x for grey cast iron, the guideline requires in the most favorable case a safety factor of 1.84, and for most
practical situations largely more than 2. This additional provision is based on the large strength scatter
encountered in cast iron parts.
x peak stress determination is strongly dependant on finite element mesh size. This problem is quite well
delt with in the fatigue assessment method through the stress gradient correction. In the static strength
assessment however, it seems to be a weak point of the method for non ductile materials.

4. Fatigue strength assessment : examples of mild steel laminate parts

4.1. Industrial problem

For manufacturers of electric rotating machines, the last decade has brought a new challenge on the mechanical
design of electric motors. On one hand the development of power electronics has led to an increasing use of variable
speed drives, which means also time-varying centrifugal forces. On the other hand the optimization of magnetic
structures has explored new types of rotors, whose design is often very far from the sturdy induction motor: that
means considerably higher stress arise in rotor material.
This evolution, combining higher and variable stress, has made a fatigue assessment of new electric machine
mandatory.
The rotor of an electric machine consists mainly of a laminate stack. The laminate material is primarily chosen
for its magnetic properties; mechanically it can be described as a mild steel with very low content of carbon and
large grain size (about 0.1mm). Laminates are usually between 0.3 and 0.65mm thick, with a preference for thinner
material for high speed machines. They are cut by a punching tool, before being stacked to the given rotor length.
Since the machine is intended to run more than 10 years at variable speed, we must assess its high-cycle fatigue
strength with sufficient confidence.

4.2. Mechanical model

Laminate parts are calculated with plane stress assumption.


To define the material model, tensile behaviour has been identified experimentally on laminate strips. For
elastoplastic analysis, an isotropic hardening is assumed.
Centrifugal load is much larger than magnetic forces, which are neglected in the simulation.
Geometry and finite element model of the rotor are given in Fig.4 and Fig.5 below ; only half a pole is modeled
for obvious symmetry reasons. An example of laminate sample for tensile fatigue tests is shown in Fig.6.
314 Samuel Kchlin / Procedia Engineering 133 (2015) 309 319

Fig. 4. Complete rotor laminate used for endurance tests

Fig. 5. Finite element model of a rotor laminate (6poles) : 1st principal stress caused by centrifugal force.

4.3. Experiments

Numerous tensile fatigue tests have been run in laboratory (I2M Bordeaux) on different types of laminate
samples. Stress ratio has been set to 0.1, since buckling prevents any possibility of applying a compressive load.
Endurance tests on rotors consist in running up and down a motor cyclically between stop and a definite speed,
thus creating a pulsating centrifugal load. Since acceleration ramps cannot be reduced less than several seconds, we
have a load frequency of a few cycles per minute. To run an endurance motor until failure lasts therefore several
months.
Samuel Kchlin / Procedia Engineering 133 (2015) 309 319 315

4.4. Fatigue strength calculation on a rotor

The main elements of the calculation are the Kt-Kf ratio and the influence of the number of cycle (through the so-
called variable amplitude fatigue strength factor).
Regarding the former, the last release of the guideline offers 2 different calculation methods: the traditional one,
based on Stielers stress gradient formulae, and a more complex one, called Material-mechanical, developed in the
last decade.
The Stieler method relies on the estimation of the relative stress gradient G (see Fig.6):





The material-mechanical method uses a product of 3 factors; one of them, which characterizes crack propagation,
has no influence for a low strength steel. The second one takes into account localized yielding, allowing for a fixed
plastic strain limit of 0.02% through the Neuber correction. The last one is a statistical correction factor nst, which
represents the influence of surface material defect distribution on fatigue crack initiation, according to a Weibull
model.

The calculation of the highly stressed surface (or edge in a 2D-model) A_st is explained in Fig.6. It is related to
the highly stressed area of a reference sample Aref_st. A detailed explanation in the case of a 3D model can be found
in [3]. The Weibull coefficients kst given in the guideline are provisional however, and a value of 30 is suggested for
every kind of steel.

The number of cycle is taken into account through a S-N curve. The guideline gives a very general curve for
every kind of steel. We made use of experimental fatigue results instead, in order to build a specific S-N curve for
exactly the laminate sort we also had in the endurance motors. From the S and N values, a maximum likelihood
estimation allowed to get 2 of the 3 parameters defining the curve: the slope of the Basquin law in the limited
endurance domain, and the fatigue limit. The knee-point has been set at 1 million cycles, as recommended in the
guideline. One can note a very good fit with experimental data in Fig.7.

A further effect, out of the scope of the guideline, is the statistical stacking effect. The strength of one laminate
and a stack of several hundreds, of one pole and a complete 4- or 6-pole rotor, are not identical. This is due to the
fact that the strength value is statistically distributed: each laminate will have a different strength, around the mean
value estimated by calculation or sample tests. If we consider the laminates in parallel, the whole stack has a
strength value which is, on the whole, lower than the individual mean value [1].
This statistical coefficient proves to depend only on the coefficient of variation of the fatigue limit.

For comparison purpose, a uniaxial fatigue calculation has been also performed: it is based on an elastoplastic
simulation, with mean stress correction according to Gerber, applied on the amplitude of the local stress.
316 Samuel Kchlin / Procedia Engineering 133 (2015) 309 319

Fig. 6. Relative stress gradient G and highly stress zone A st (perpendicular to and along the edge respectively), illustrated on a fatigue test
sample)

Fig. 7. S-N curve for punched laminate, from fatigue tests conducted at I2M-Bordeaux [2]
Samuel Kchlin / Procedia Engineering 133 (2015) 309 319 317

4.5. Fatigue strength calculation on a notched sample

Based on fatigue limits for smooth samples, either polished or punched, the calculated fatigue limit has been
compared to the experimental one for different notched sample types: samples cut by electric-discharge-machine
(EDM) without further polishing, samples cut by punching and 2 different notches.
The guideline takes into account the surface roughness. It makes sense to include it in the calculation derived
from polished sample fatigue limit, as an edge-roughness, although the definition related to surface roughness Rz
is in this case not appropriate. In Table 1 below, a somewhat arbitrary value of 0.9 has been chosen for comparison
purpose.

4.6. Comparison on a rotor

Fig. 8 : endurance tests results and prediction with different methods

FKM prediction shows good agreement with experimental results (Fig.8). We can also observe that the
experimental lifetimes fall approximately halfway between the prediction given by the traditional gradient method
according to Stieler and the more recent material-mechanical method included in the last release (2012).
The uniaxial prediction, which is a purely local stress comparison, is largely too conservative.

4.7. Comparison on laminate samples

As shown in Table 1, results on notched fatigue samples are not as good as rotor endurance tests. Here the FKM
method as well as the uniaxial evaluation underestimates considerably the fatigue strength of notched specimen. The
new material-mechanical method, beeing less conservative, gives a slighly better match, but is still largely too
conservative. However, calculated values are in good accordance with experimental results for specimen cut by
electric-discharge-machine (EDM), if the calculation is based on polished specimen values (first column).
318 Samuel Kchlin / Procedia Engineering 133 (2015) 309 319

Table 1. Allowable stress values according to different methods and assumptions. Stress values are given in percentage of
experimental fatigue limit for the notched specimen, so a perfect match would be a value of 100. A stress larger than 100
indicates a non-conservative method.
Notched sample type EDM, Kt=2 Punched, Kt=2 Punched, Kt=3
Relative stress gradient [mm-1] 0.86 0.26 0.84
Experimental fatigue limits used for calculation

unnotched polished specimen R=0.1 x

unnotched punched specimen R=0.1 x x x x

Support number calculation according to Stieler Stieler Mat.Mech. Stieler Mat.Mech.


Calculation results
FKM without roughness correction [%] 112 64 77 54 64

FKM with roughness correction [%] 106

Uniaxial elastoplastic [%] 110 65 52

Fig. 9 : rotor laminates broken by fatigue on endurance test : rupture occurred not always in the peak stress area (left fillet)

4.8. Discussion

To account for the different scores obtained on one hand for the rotor, on the other hand for notched samples, two
explanations can be suggested.
Firstly, there might be a more important influence of laminates stacking.
Secondly, stress concentration is different on the rotor, compared to the notched samples: it is both lower and
more localized in the case of the rotor, since we have Kt=1.37 and a stress gradient of 0.91 mm-1. Therefore, a
significant local yielding appears on the notched samples, whereas it is negligible in the rotor notch for the stress
range of interest. The gradient correction (i.e. the so-called support effect or Sttzwirkung) appears to be
insufficient in the case of significant stress concentration (Kt=2 or more).
Samuel Kchlin / Procedia Engineering 133 (2015) 309 319 319

Moreover, recent research works [2] have shown the importance of edge defects on fatigue crack initiation for
such a material and such a fabrication process (punching). Indeed, as illustrated in Fig.9, rupture occurred not
always in the peak stress area, but also at some distance from it: this indicates that macroscopic stress field alone
may not control the failure.
This specific aspect is most probably not correctly described in the general FKM evaluation, even in the new
statistical correction factor, whose values have been set from a literature review (according to [6]).

5. Conclusion

The comparisons presented here have clearly shown the limitation of a general calculation method like the FKM
guideline. The examples above are quite specific: a grey cast iron part with localized peak stress, and very thin
laminate parts. For such cases, dedicated methods must be applied, but they need an intensive experimental
calibration.
The comparison with other methods (critical distance or simple uniaxial evaluation with traditional mean stress
correction) raises a critical issue: which material data must be used if no specific experimental values are available?
This is especially a problem with the critical distance method, for which the parameters are scarcely documented.
Multiaxial fatigue criteria are very powerful, but only if their parameters are known with confidence, which is
seldom the case in the design phase. Furthermore, with a two parameter criterion, only one influence factor can be
taken into account (e.g. the mean stress effect). Any additional effect (roughness, local yielding, size effect, etc)
needs additional test data.
When costly experimental investigation is not realistic (this is actually the most frequent case in many areas of
mechanical industry), the FKM guideline gives a general framework and a valuable working base, even if it seems
usually quite conservative. Like any calculation tool, it must always be regarded with the critical eye of the
professional.

References

[1] H.E. Daniels, The statistical theory of the strength of bundles of threads, Proceedings of the Royal Society of London. Series A, Mathematical
and Physical Sciences, Vol. 183, No. 995 Jun. 18, 1945, pp. 405-435
[2] H. Dehmani, Ch. Brugger, Th. Palin-Luc, Charles Mareau, Samuel Koechlin, Fatigue behavior of punched thin Fe-Si sheets, 15th
International ASTM/ESIS Symposium on Fatigue and Fracture Mechanics, 2015
[3] A.Diemar, R. Thumser, J.W.Bergmann, Statistischer Grsseneinfluss und Bauteilfestigkeit, eine neue Methode zur Ermittlung von
Spannungsintegralen, MaterialPrfung, 2004
[4] NF EN1561 : European standard - Norme europenne : Fonderie Fonte graphite lamellaire, 2011
[5] FKM (Forschungskuratorium Maschinenbau), Analytical Strength Assessment of Components, VDMA Verlag, 6th revised edition, 2012
[6] E. Kullig, M. Vormwald, O. Hertel, A. Esderts, K. Hinkelmann, Weiterentwicklung des Ermdungsfestigkeitsnachweises, Tagung Deutscher
Verband fr Materialforschung und -prfung e.V. DVM Darmstadt, 2010
[7] Manuel des Fontes Moules (Iron Castings Handbook), Editions techniques des Industries de la Fonderie, 1983
[8] D.Taylor, The Theory of Critical Distances, A New Perspective in Fracture Mechanics, Elsevier, 2007

Das könnte Ihnen auch gefallen