Metal Science and Heat Treatment

Vol. 52, Nos. 9 10, 2010

UDC 669.017:621.78

A METHOD FOR DETERMINING CONSTANT C

IN THE HOLLOMON PARAMETER
B. M. Shlyakman,1 O. N. Yampolskii,1 and D. V. Ratushev1
Translated from Metallovedenie i Termicheskaya Obrabotka Metallov, No. 9, pp. 48 51, September, 2010.

A method for determining constant C in the Hollomon temperature-time parameter is suggested, which ensures the highest correlation between the properties of a material and this parameter. The use of this value of C
improves approximation of experimental data and accuracy of prediction.

Key words: temperature time parameter, Hollomon, heat treatment, mechanical properties.
It has been suggested in [2] to use a constant value
C = 20 in parameter (1). Then (1) is called a Larson Miller
parameter.
In the actual fact constant C can take various values in a
quite wide range. In [3] processing off the results of tests for
long-term strength shows that constant C determined by formula (2) varies depending on the grade of the steel and on
the test conditions from 7 to + 56, which makes the use of
C = 20 for description of these tests not always appropriate.

INTRODUCTION
The Hollomon temperature time parameter [1] is often
used for describing dependences of the properties of materials on the temperature and time conditions of their treatment.
This parameter makes it possible to combine the temperature
and the time in one variable and thus to reduce the number of
experiments and to simplify processing of experimental data.
As a rule, the Hollomon parameter is written in the form
H = T (C + log t)/1000,

(1)

METHOD FOR COMPUTING CONSTANT C

where T is the Kelvin temperature, t is the time (commonly

in hours), and C is a constant dependent on the material.
Constant C entering the Hollomon parameter is determined from experimental data with the help of the following
formula [1]
C=-

T1 log t1 - T2 log t 2
,
T1 - T2

We suggest that constant C should be determined on the

basis of all the experimental data of the sample independently of the availability of data with the same properties.
Then the determined value of constant C is the best for approximating experimental data.
If we assume that the properties of the material and the
Hollomon parameter are correlated, the best approximation
variant will be the one where the correlation coefficient has a
maximum absolute value. Let us determine the value of C
from the condition of maximum correlation coefficient.
It is known that the coefficient of correlation between
two arrays X and Y is computed by the formula:

(2)

where T1 , T2 , t1 , t2 correspond to two similar values of

characteristics of the material obtained at two different temperatures T and times t. If the test sample contains several
pairs with the same values of characteristics of the material,
C is chosen as an arithmetical mean of several values computed by formula (2).
If the sample is not large, experimental data with same
values of characteristics of the material may be absent. Then
it is impossible to determine constant C with the help of formula (2). In addition, the experimental error may noticeably
affect the value of C when the sample is small.
1

r=

n xy - x y
n x 2 - ( x)

n y 2 - ( y)

(3)

Let Y be an array of values of one of the properties of the

material and X be the Hollomon parameter H. Substituting
(1) into (3) and making elementary transformations we find
r=

TK OMZ-Izhora JSC Research Center, St. Petersburg, Russia

(e-mail: BSh@inbox.ru).

aC + b
m gC 2 + 2dC + e

(4)

451
0026-0673/10/0910-0451 2010 Springer Science + Business Media, Inc.

452

B. M. Shlyakman et al.
HB

TABLE 1. Values of Constant C in the Hollomon Parameter as Computed by Different Formulas

300

C = 36.32

280

Constant C computed by formula

R
V
W
T
S
U
*

13.4
15.0
15.9
17.1
18.0
19.6

260

260

240

240

12.4
13.3
14.3
13.3
13.6
15.9

220
200

b
C = 20

280

(5)

Steel*
(2)

HB
300

220
200

R = 0.969

180
35

36

37

38

R = 0.901

180
19

20

21

22

sr , P

sr , P

950

950

C = 32.99

C = 20

Conventional notation.

where

850

850

750

750
R = 0.977

a = n Ty - T y,

650

b = n T log t y - T log t y,

650
31

32

33

34

s0.2 , P
800

g = n T 2 - ( T ) ,
2

35

H
C = 30.66

(ad - bg ) C + ae - bd
r
=
= 0.
C m ( gC 2 + 2dC + e ) 3
whence
ae - bd
.
bg - ad

C = 20

29

30

31

R = 0.940
32

33

Differentiating (4) with respect to C and equating the obtained value to zero we derive an equation for determining
constant C that ensures the highest value of the correlation
factor, i.e.,

C=

800

R = 0.977

m = n y 2 - ( y) .

(5)

We can see that this simple enough formula allows us to

compute C using all experimental data without constraint.
PROCESSING OF EXPERIMENTAL DATA
Let us compare the values of C obtained by formulas (5)
and (2). Table 1 presents the results of processing of experimental data given in work [1] that studies the effects of various tempering modes on the hardness of carbon steels. It can
be seen that the values of C computed by formula (5) correlate with the values computed by formula (2).
Let us see how the results of processing of experimental
data depend on the use of constant C computed by formula (5)

22

600

500

21

700

600

e = n T 2 log 2 t - ( T log t) ,

20

s0.2 , P

700

d = n T 2 log t - T log t T ,

R = 0.928
19

500

19

20

21

22

Fig. 1. Dependence of mechanical properties of steel

X12CrMoWVNb on Hollomon parameter H for various tempering
modes (hardening from 1050C, tempering at 680 720C for
2 40 h): a) constant C computed by equation (5); b ) constant
C = 20.

as compared to C = 20. Figure 1 presents dependences of the

values of the hardness, rupture strength, and conventional
yield strength of steel X12CrMoWVNb (the DIN classification) after hardening from 1050C and tempering at 680
720C on the Hollomon parameter H. It can be seen that in
case (a) the experimental data are less scattered and have a
higher correlation coefficient than in case (b ). The computed
values of constant C differ somewhat. If we want to have a
common constant for various properties, we may take their
mean arithmetical value. In the case considered the arithmetical mean is C = 33.15 and the coefficient of correlation remains virtually unchanged (differs by no more than 0.001).
A common constant can also be obtained by normalizing
all the data with respect to Y and uniting them into one sample. Specifically, in the case of normalizing by the formula
y* =

y - y min
,
y max - y min

(6)

and uniting three arrays (the hardness, the rupture strength,

and the conventional yield strength) into one array, computa-

A Method for Determining Constant C in the Hollomon Parameter

s, P

s, P

200

200
650
705
760
815

100
50

650
705
760
815

100
50

871
982

871
982

10

22

24

26

28

30

TABLE 2. Prediction of the Time before Failure of Stainless Steel

Tested for Long-Term Strength (according to the data of [4])
Test method

Time before failure, h

Computation

T, C

s, MPa
C = 20

10
5
20

453

5
13

b
15

17

19

Fig. 2. Dependence of the long-term strength of stainless steel on

Hollomon parameter H (according to the data of [4]): a) C = 20;
b ) C = 12.36 as computed from equation (5).

tion of C by formula (5) gives C = 33.32, i.e., we obtain a

value virtually corresponding to the arithmetical mean.
Figure 2a presents the results of tests of stainless steel
for long-term strength [4]. The points lie on a straight line for
each temperature. We can also draw one common line, which
is justifiable when the volume of the experimental data is
high enough. If the data are scarce, especially if they have
been obtained at close temperatures, the error due to such an
approximation increases markedly. The author of [4] manages to place all the points on one straight line using a linear
combination of T and log t as the temperature time parameter; in fact, he uses another parameter. However, a similar
effect can be obtained if we choose a correct value of C in the
Hollomon parameter.
Figure 2b presents the results of processing of the same
experimental data, where C has been computed by formula (5). In this case not only the accuracy of the approximation
within the interval is improved but also the accuracy of prediction outside the interval. Indeed, if we take all the data
presented in Fig. 2a except for the last ones obtained at
982C, plot the approximating dependences for them, and
predict the time before failure t for the temperature of
982C, we can evaluate the accuracy of the prediction.

982
982
982
982

20.7
14.5
9.0
8.3

4.7
34
493
776

C found from
formula (5)

20.5
80
495
674

Experiment

22
100
520
580

Table 2 presents the results of such computations. It can

be seen that the use of constant C computed with the help of
formula (5) gives a more accurate prediction of the time before failure.
CONCLUSIONS
We suggest a method for determining constant C in the
Hollomon parameter, which ensures the best correlation between the properties of a material and this parameter. Constant C is determined for the whole of the sample of experimental data. The use of C determined by the method suggested improves approximation and extrapolation of experimental data.
REFERENCES
1. J. H. Hollomon and L. D. Jaffe, Time-temperature relation in
tempering steel, Trans. Am. Inst. Min. Met. Engrs., 162,
223 249 (1945).
2. F. R. Larson and A. Miller, A time-temperature relationship for
rupture and creep stress, Trans. ASME, 74(5), 765 775 (1952).
3. G. Bandel and H. Gravenhorst, Methods of processing of experimental results, in: Study of Refractory Steels and Alloys [Russian translation], GNTI, Moscow (1960), pp. 27 40.
4. F. J. Clauss, An examination of high-temperature stress-rupture
correlating parameters, Proc. ASTM, 60, 905 927 (1960).