FATIGUE LIFE OF FORGED, HARDENED AND TEMPERED
CARBON STEEL WITH AND WITOUT NORMALIZING
A. Zabett1*, R. Irankhah1, M. Miri Disfani1, I. Zohur Karimi1, M. Hashemi2
1- Ferdowsi University of Mashad, Mashad, Iran
2- Part Sazan forging Co. Mashad, Iran
A common process of auto-parts production consists of forging, normalizing, quench and
tempering. Normalizing treatment is usually employed to improve machinability of steel, and
to homogenize and refine the microstructure. For simple parts, production cost would reduce
if normalizing stage were eliminated. In this paper, effects of normalizing on properties of a
carbon steel has been studied. Experiments have been carried out at industrial conditions on
forged, quenched and tempered parts. Results show minor effect of normalizing on
microstructure, hardness and tensile properties. However, specimens normalized after forging
indicate improved fatigue life at different stress amplitudes. Forty samples were tested for
each heat treatment at three different stress amplitudes. Statistical analysis of the results
proves meaningful reduction in number of cycles to failure for specimens without normalizing
treatment. It is believed that more number of cracks in these specimens and more coalescence
of cracks cause the early failure of specimens. Disperse ferrite areas in microstructure seem to
be the preferred initiation sites and different distribution of ferrite in the microstructure of
samples leads to different fatigue lives.
Keywords: fatigue life, heat treatment, normalizing, microstructure and carbon steel.
Forging the steel parts followed by heat treatment is a common process in auto-part
manufacturing. The heat treatment involves normalizing, austenitizing, quenching and
tempering. Normalizing treatment before hardening is mainly for homogenization and
refinement of microstructure and softening the steel. Homogenization and microstructure
refinement improve mechanical properties and prevent distortion or cracking of parts with
complicated shapes during hardening, while softening reduces machining costs. If the part
does not need machining, and has a simple shape, normalizing treatment may be eliminated
from the process to decrease manufacturing costs. For this decision, the effect of normalizing
on the mechanical properties of the part should be considered. It is the aim of this paper to
investigate these effects on mechanical properties of a carbon steel, with an emphasis on the
fatigue life of the parts.
* Corresponding author, email@example.com
2. EXPERIMENTAL PROCEDURE
The part selected for this study is a ball pin made of CK45 steel (see table 1). Steel bars with
38 millimeter diameter cut in 250 millimeter pieces were heated to 1250°C in an induction
furnace and forged in two steps by a 1000 ton mechanical press. 120 forged parts were
selected, from which, 60 parts were normalized and all 120 parts were hardened and tempered
altogether. Normalizing took place at 890°C for 130 minutes and cooled in air. All parts were
austenitized at 890°C for 120 minutes and quenched in oil and then tempered at 570°C for 75
minutes. All treatments were followed based on manufacturing procedures at industrial
Table 1. Chemical composition of CK45 steel.
Elements C Si Mn P S Cr Ni Mo
Wt% 0.44 0.22 0.64 0.006 0.012 0.15 0.11 0.05
Hardness of all parts was measured. One metallographic sample was taken after each
treatment. For the purpose of simple reference, abbreviated name for each condition is listed
in table 2. Tensile properties were determined according to ASTM E8m for two QT and two
NQT samples using a 200kN universal testing machine. Prior austenite grain size was
measured for specimens based on ASTM E112, using a quantitative image processing
Table 2. Treatment condition of samples
Condition Abbreviated Name
1 As received AR
2 As forged AF
3 Forged and normalized N
4 Forged and hardened (quenched) Q
5 Forged, normalized and hardened (quenched) NQ
6 Forged, hardened and tempered QT
7 Forged, normalized, hardened and tempered NQT
Fatigue tests were carried out for twenty QT and twenty NQT samples using a rotary bending
fatigue testing machine. A drawing of fatigue test piece is shown in figure 1. Fatigue test
pieces were made from parts with hardness between 60.2 to 61.4 HRA. The test pieces were
machined with a CNC machine and grinded with silicon carbide paper grades 400, 600, 1000
and 2500 and finally polished to a bright surface with one micron diamond paste.
RESULTS AND DISCUSSION
Hardness of samples after each treatment is given in table 3. Hardness numbers are the
average value of five readings on each part. The difference between parts hardened with and
without normalizing (NQ and Q) is about 0.7HRA while after tempering (QT and NQT) the
difference reduces to about 0.1HRA. However, the hardness measurement taken on all QT
and NQT parts shows wider range of hardness values. A T-test† was performed on 96 samples
to show any significant difference between the two sets of parts. This test confirms NQT
samples with an average hardness of 61.45HRA are harder than NQT with an average
hardness of 60.31HRA. Although this difference in hardness is very small compared with
* MIP is a registered trade mark metallographic image processing software developed at Ferdowsi University of
† Student T-test or paired variance test
acceptable range for hardness of these parts and does not represent a significant change in
mechanical properties, it confirms a difference resulted from normalizing treatment.
The prior austenite grain size of samples after each treatment is given in table 4. Figures 2 to
11 show the microstructure of different samples. Small changes of prior austenite grain size
and little differences in microstructure of QT and NQT sample are observed. Therefore, it is
expected to see little changes in mechanical properties as well. This is true for tensile
properties as summarized in table 5.
Table 3. Average hardness at different condition.
Condition AR AF N Q NQ QT NQT
(HRA) 58.04 58.44 54.68 62.92 62.22 61.78 61.88
Table 4. Prior austenite grain size, ASTM E112.
Condition AR AF N Q NQ QT NQT
grain size 9.2 9.3 9.4 9.9 9.6 10 10.3
Table 5. Tensile properties of two QT and twoNQT samples.
Yield Stress Tensile Strength Elongation
Specimen number N/mm2 N/mm2 %
QT-1 584 861 17.7
QT-2 578 851 16.7
NQT-1 584 850 17.2
NQT-2 588 844 16.7
S-N curve is drawn for QT and NQT samples in figure 12 and table 6 summarizes the fatigue
test results. From figure 12 the difference between the lives of QT and NQT samples is
evident. Nevertheless, in order to see the statistical significance of this difference and to
assure the changes of average lives at different stress levels do not lie within the normal
variations common to fatigue testing results, T-test was performed on the fatigue lives.
Table 6. Fatigue test results.
Tests Average Life Standard
QT 583 8 45635 15181 17.3
QT 534 8 101399 27881 N/A
QT 516 4 137837 48125 9.7
for 583 20 45635 14775
NQT 583 8 54018 11184 11
NQT 534 6 119552 37531 8.5
NQT 516 6 194089 68817 5.6
for 583 20 54635 13528
* Average number of cracks counted on the fracture surface.
The results of T-test at each stress level did not confirm a significant difference between
average lives of QT and NQT samples (see table 7). On the other hand, in all three stress
levels similar difference is distinguished, as average lives of QT samples in all three stress
levels are shorter than those of NQT samples. Since the purpose of using T-test is to
differentiate between QT and NQT conditions in general, a normalizing formula was
employed to test fatigue data altogether. For this, average fatigue life at 583MPa was taken as
a reference value, and the lives of samples tested at 534 and 516MPa were divided by the
ratio of their average life to the reference value. In this way, normalized lives of twenty QT
samples and twenty NQT samples were used for T-test. The results given in table 7, verify
significant change in average life of the QT samples over that of the NQT specimens.
Table 7. Two-Sample T-Test and CI: NQT, QT
Stress Amplitude T-test Results
583 MPa Estimate for difference: 8383.75
95% CI for difference: (-6141.46, 22908.96)
T-Test of difference = 0 (vs not =): T-Value = 1.26 P-Value = 0.232 DF = 12
534 MPa Estimate for difference: 18152.8
95% CI for difference: (-23860.3, 60165.9)
T-Test of difference = 0 (vs not =): T-Value = 1.00 P-Value = 0.348 DF = 8
516 MPa Estimate for difference: 56252.2
95% CI for difference: (-31216.1, 143720.4)
T-Test of difference = 0 (vs not =): T-Value = 1.52 P-Value = 0.172 DF = 7
Normalized for 583 MPa Estimate for difference: 8383.75
95% CI for difference: (-692.30, 17459.80)
T-Test of difference = 0 (vs not =): T-Value =1.87 P-Value = 0.069 DF = 37
Fracture surfaces of all samples were examined with the aid of a stereomicroscope. Figures 13
and 14 show typical fracture surfaces of two samples tested at 583 MPa. Number of cracks on
the final fracture surface of each sample was counted (see table 6). Average number of cracks
increases as the stress amplitude increases. At the same stress amplitude the average number
of cracks is higher for QT specimens. Examination of the specimen surfaces for cracks other
than those on the final fracture surface agrees with the fact that the number of cracks which
could be detected on the surface of QT specimens is larger than that of NQT specimens. More
number of cracks initiated in QT samples followed by coalescence reduces the life of QT
samples compared with those of NQT specimens. This can be related to more preferable crack
initiation sites in QT samples. It is well established that ferrite is the usually the crack
initiation site in ferritic-pearlitic steels[1-4]. Many researchers have shown that ferrite grains
in ferritic-pearlitic steels[5-8] and even in duplex steels[9-10] are the favorable sites for crack
nucleation. G.Z. Kovalchak, et. al.  reported reduction in fatigue life with increase of ferrite
in microstructure. In QT samples more disperse ferrite areas could very well be the reason for
the more number of cracks and the shorter lives. Homogenized microstructure of normalized
samples (NQT) seems to be less vulnerable to crack initiation and results in longer lives.
Estimated ferrite in QT samples is about 12-15% while 7-10% ferrite is observed in NQT
samples. This reasonably explain the difference in lives of the two sets of samples.
SUMMARY AND CONCLUSIONS
A number of experiments at industrial conditions were conducted to investigate the effects of
normalizing treatment on mechanical properties of auto parts which were forged, hardened
and tempered. The results show elimination of normalizing treatment has little effect on
hardness and microstructure of simple parts made of medium carbon steel and does not
change the tensile properties. However, meaningful changes of fatigue life at medium cycle
are recognized. Samples made from parts without normalizing treatment show shorter lives
and more number of cracks in fatigued specimens. More number of cracks on the surface and
fracture surface of QT samples seems to be the reason for shorter lives of these specimens.
Based on microstructural examination it is assumed that more disperse ferrite grains in QT
samples are the cause of more crack initiation and more coalescence in these samples lead to
shorter lives. More detailed experimental work is needed to examine the crack initiation area
and to investigate the effect of normalizing on reducing the number of cracks in forged,
hardened and tempered carbon steels.
1. ASM Metals handbook, Vol. 1, Properties and selection: Irons, steels and high-
performance alloys, 1990, ASM International Publication.
2. ASM Metals handbook, Vol. 11, Failure analysis and prevention, 1986, ASM
3. ASM Metals handbook, Vol. 12, Fractography, 1987, ASM International Publication.
4. ASM Metals handbook, Vol. 1, Fatigue and Fracture, 1996, ASM International
5. K. Endo and H. Goto, "Initiation and propagation of fretting fatigue cracks", Wear,
Vol. 38, 1976, pp. 311-324.
6. G.Z. Kovalchak, et. al., "Effect of widmanstutten ferrite on some properties of hypo-
eutectoid steel" Metal Science and Heat treatment, Vol. 21, No. 2, Feb. 1979.
7. N. Narasaiah, K.K. Ray, "Small crack formation in a low carbon steel with banded
ferrite–pearlite structure", Mats, Sci, and Eng. A, Vol. 392, 2005, pp. 269–277.
8. J.J.F. Bonnen, F.A. Conle, T.H. Topper, Int. J. of Fatigue, Vol. 23, 2001, pp. S385–
9. I. Alvarez-Armas, et. al., Int. J. of Fatigue, Vol. 29, 2007, pp. 758–764.
10. P.C. Chakraborti and M.K. Mitra, Int. J. of Fatigue, Vol. 28, 2006, pp. 194–202.
Figure 1. Drawing of fatigue test specimen
Fig. 2. Prior austenite grains of quenched
and tempered (QT) sample. Fig. 3. Prior austenite grains of normalized,
quenched and tempered (NQT)
Fig. 4. Microstructure of as forged (AF)
sample. (100X) Fig. 5. Microstructure of as forged and
normalized (N) sample.
Fig. 6. Microstructure of forged, quenched
and tempered (QT) sample. (100X) Fig. 7. Microstructure of forged, normalized,
quenched and tempered (NQT)
Fig. 8. Microstructure of forged, quenched
and tempered (QT) sample. (200X) Fig. 9. Microstructure of forged, normalized,
quenched and tempered (NQT)
Fig. 10. Microstructure of forged, quenched
and tempered (QT) sample. (500X) Fig. 11. Microstructure of forged, normalized,
quenched and tempered (NQT)
Fig. 12. S-N curve for QT and NQT samples.
Stress Amplitude (MPa)
Number of Cycles to Failure
Fig. 13. Fracture surface of QT sample. Fig. 14. Fracture surface of NQT sample.