Professional Documents
Culture Documents
t-2, 1995
A T O T A L STKA~N E N E R G Y D E N S I T Y M O D E L O F M E T A L F A T I G U E
K. M. Golos
In this paper a total cyclic strain energy density equal to the sum of ptastic strain energy
and tensile elastic strain energy densities is used as a damage parameter for metal fatigue.
It is shown that the total cyclic strain energy density is a consistent damage parameter
for low- and high-cycle fatigue in the conditions of both uniaxial and muttiaxial cyclic
loading. This parameter is also consistent with the concept of crack initiation and
subsequent propagation. The approach described here is applicable for both ideal
Masing and non-Masing material response. The predictions of the proposed criterion
are compared with the experimental data for medium carbon steel St5. The comparison
has shown good agreement.
INTRODUCTION
During the desig~ process of every new machine or element that will be subjected to cyclic loading
in service, the fatigue characteristics of the proposed design are necessary. Failure criterion should be
selected on the basis of its generality and consistency and should be applied not just to a special
situation, but to all types of loading, geometries, and materials. An analytical fatigue failure theory that
is easy to use would enable the designer to assess the fatigue behavior of his preliminary designs and
modifications before arriving at a final design.
Earlier research efforts were concerned with correlating the fatiggue life of a smooth specimen with
either the stress or plastic strain amplitudes in high- or tow-cycle fatigue, respectively [1, 2]. Sub-
sequently, Manson [3] and Coffin [4] proposed a relationship for correlating high- and low-cycle fatigue
in which the total strain amplitude is divided into elastic and plastic components. In these approaches
the interrelation between the cyclic stress and strain amplitudes, and the fatigue process was usually
overlooked [5-8]. Since the fatigue damage is mainly caused by plastic strain energy, this parameter
was used to describe fatigue damage [8]. However, when the plastic strain range, A a, decreases,
A ap --+ 0, and the corresponding plastic strain energy per cycle AWp + 0. Therefore, it was proposed
to describe the fatigue damage process as a function of mechanical input energy into the material [9-12].
The total strain energ3' density, equal to the sum of the tensile elastic and plastic strain energy densities,
was used as a damage parameter.
In the paper the unified nature of this parameter is considered. The total strain enerD" density
approach is used to describe low- and high-cycle uniaxial fatigue and multiaxial fatigue. Experimental
data for medium carbon steel St5 are presented to show the correlation between the predicted and
observed results.
Warsaw University of Technology, ul. Narbutta 84, 022524 ;vVarsaw, Poland l Published in PrOblemy
Prochnosti, Nos. 1-2, pp. 53-64, January-February, 1995.
1
Fig. 1. A typical hysteresis loop and tensile elastic and plastic
strain energies.
During cyclic loading the energy is dissipated because of plastic deformations. A part of this energy
is converted into heat, and the other part is rendered irrecoverable at every cycle due to the plastic
strain energy absorption. The plastic strain e n e r ~ per cycle, zXWP~is the area of the hysteresis loop,
and it does not vary appreciably with cycles in the case of strain-controlled tests [5-15]. However, when
the strain range, A e, decreases, A g - + 0 and the corresponding plastic strain energy density
A W p --+ O. In this case, the macroscopic (bulk) response of the material is quasielastic, although at the
microscopic (gain) level plastic deformation may occur. Therefore, it is assumed that damage due to
cyclic loading can be modelled as a function of the absorbed plastic strain energy per cycle and that
part of the elastic energy which facilitates the crack growth; see Fig. 1. Thus, both tensile elastic and
plastic parts of the strain energy per cycle have to be determined, i.e.,
+
AW t = AW p + AWe . (1)
The expression for calculating the plastic strain energy for non-Masing material was developed in
[13-14]. A skeleton curve different from that of cyclic curve is defined (Fig. 2). This curve is obtained
from matching the upper branches of the hysteresis loops through translating each loop along its linear
response. The equation of the skeleton curve, where the origin corresponds to the hysteresis loop with
the minimum elastic part, can be expressed as
9 A~J* (h ~, )l/n*
Ae - E + 2 ~/2K'* , (2)
Ar A~ skeleton curve
~o
AC
r
0 strain range
1-n
A W P :- , (A a - 5 G o) A e ~ + 5 ~0 A ~ , (3)
l+n
where
is the increase in the proportional stress due to non-Masing behavior of the material. Knowing the
equation of the skeleton curve and the applied stress and plastic strain range, 5 <Y0can be determined
from (4). The ideal Masing description is obtained from (3) by setting n* = n', K* = K', and 3 o0 = 0.
The recoverable tensile elastic strain energy density is given by
+
Therefore, the total damaging energy input per unit material volume is
A power law relationship exists between the damage parameter, AW t and the number of cycles to
failure, Nf, i.e.,
35 Strain Energy Density Model of Metal Fatigue
AW t = ~ N~ + c , (7)
where constant C is the portion of the tensile elastic energy input which causes no damage, i.e.,
1
c =~ (%m) 2, (8)
where (51irais the fatigue limit of the material. For most materials, the value of the fatigue limit, (51ira,
is listed in material handbooks, and thus C can be readily calculated from (8).
To determine a and ~ at least two tests are required.
To extend the concept of the total strain energy density per cycle to the multiaxial stress case, we
have to calculate the elastic and plastic parts of the strain energy during cyclic loading.
The cyclic elastic strain energy density for the positive stress parts of the cycle can be calculated
from
where
xT (sTaxH((STY)+ + (sTx
"~2;'max= (51maxo"2max"''/-/((51max,).t.1,,,-
(G~nax) + (52119824(5~n3
ax
H (O~2)H
ax
(G~nax) +
+%max%max//(%max)H(%max)
are the first and second invariants of the stress tensor, and H is the Heaviside function.
The cyclic plastic strain energy density, AW P, can be calculated from
where
p,1/2 ,
9 1/2, A ap = (2/3 A Ei]P A Eij)
A 0.* = (3/2 kSij* AS(/)
Golos 36
[MPa] ~
500, ./J
/ S
J
/ III
400
300
L/
steel St5
i00,
i i ~ 4 s
and K* and n* are the material parameters, in general, the functions of multiaxial stress state. According
to the J2 theory, the equivalent plastic strain can be defined as follows:
Combining Eqs. (11) and (12), we obtain the effective skeleton cyclic stress-strain curve for
multiaxial proportional loading:
The cross-hatched area NO*M in Fig. 2, representing the increase of the plastic strain energ3r per
cycle due to non-Masing behavior, can be evaluated in the following manner. For the coordinate system
with the origin at L, the plastic strain ener~w density for loading from point L to O* is given by
AWlp = I ~ (15)
A W t i' b[~/,. 3]
o steel St5
101 ~ p---(
Wuniaxial
<)
>
I00: ~ ~
i0-i , , , ,I , , , ,! ! , , ,I , i iiI I i z ~I f
102 103 104 105 106 Nf [cycle
Fig. 4. Total strain energy density per cycle, AW t, versus the number of cycles to
failure, Nf, 9 = - v (uniaxial cyclic loading).
Then the plastic strain energy density of the hysteresis loop N O * K L M N i s given by
where 6 o0 is the increase in the elastic part due to multiaxial loading history, a measure of the multiaxial
cyclic hardening (or softening), and is given by
where
is the effective stress range, and subscripts (2) and (1) denote the maximum and minimum value of the
deviator stress components.
The increase in the proportional range 5 Go may also be estimated from the effective stress on the
cyclic curve (A or/2 for each controlled strain) and the minimum proportional stress o0, i.e.,
Golos 38
TABLE 1. Chemical Compositions in wt% for St5 Medium Carbon Steel
,,,,,, ,,
AWt=Awe + A W P = A W e =. .2E1[
. . . (ITax)2 - 2(1 + v)/~2ax] +
AW t = ~:(p) N ? + C, (22)
e2
p- (23)
(a 1 - a3)/2 "
~:(p) = ap + b, (24)
It is noted that the proposed multiaxial fatigue criterion (25) can be expressed as
The proposed criterion is hydrostatic pressure sensitive (/~1ax) and has an invariant property, i.e.,
it is a frame-indifferent criterion.
39 Strain Energy Density Model of Metal Fatigue
A medium carbon steel St5 has been selected for the analysis. Stock materials for the specimens
were a 20-ram-diameter bar and a 36-mm-thick plate. The chemical composition of the investigated
material is given in Table 1.
In the initial fatigue tests, the specimens were subjected to fully, reversed strain-controlled q(cles
of a sinusoidally varying axial strain. The average applied strain rate for uniaxial tests was 5-10.~ s-1.
Monotonic and cyclic strain-stress curves for St5 steel are shown in Fig. 3. Figure 4 shows the cyclic
strain energy density, AW t, plotted against Nf for St5 medium carbon steel. The values of ~ ~ C, as
well as other parameters for the above material are listed in Table 2. The best fit line to experimental
results (Eq.(7)) is also shown in Fig. 4.
Multiaxial fatigue tests were conducted with in-plane strain ratio 9 = A e/A ea = 0.0. Two principal
strains, aa and at, were controlled during the tests. Then, stable hysteresis energy was taken to compare
and analyze the fatigue results. For the analyzed material the area of the hysteresis loop for both
AWt'I[MJ/m3] steelSt5
p=~ct/ACa=O 0
I01
~O ~ 0 68 + 0.15
100.
lo'll102 I I I l l
103
! I I Jl
104
I I I I |
105
I I I I
106 Nf [cycler--
I I J I t|
Fig. 5. Total strain energy density, AW t, versus the number of cycles to failure, Nf,
p=0.
Golos 40
directions does not vary appreciably during the cyclic life. To obtain the total strain energy density, the
positive tensile elastic strain energy was added to the plastic strain energy. The variation of the total
strain energy density, AW t, when plotted against the number of cycles to failure, Np falls on an almost
straight line for p = 0 in log-log scale; see Fig. 5. The solid line in this figure represents theoretical
prediction.
The correlation with the experimental data is good for both low- and high-cycle regimes. One can
observe the unifying nature of the proposed damage parameter.
CONCLUSIONS
In this study the total strain energy density theory of metal fatigue has been investigated.
A form of the cyclic strain energy density equal to the sum of plastic and tensile elastic strain
energy densities is used as a damage parameter for fatigue failure.
The unifying nature of this parameter is demonstrated. It is shown that the total cyclic strain energy
density is a consistent damage parameter for low- and high-cycle uniaxial and multiaxial fatigue.
The predictions of the proposed unified method are compared with the experimental results for
St5 medium carbon steel and are shown to be in good agreement.
REFERENCES
14. D. Lefebvre and E Ellyin, "Cyclic response and inelastic strain energ!r in fatiga~e failure," Int. J..
Fatigue, 6. no 1, 9-15 (1984).
15. K. Golos, "Badania zameczeniowe stali St5 w warunkach cyclicznych obnciazen blokowych," VI
Konferencja naukowa-Sterowanie, naped, wytrzymalosc zmeczeniowa i projektowanie maszyn
budowlanych, Warszawa (1994).