International Journal of Fatigue 26 (2004) 959–966

www.elsevier.com/locate/ijfatigue

The effect of inclusions on the fatigue behavior of fine-grained

high strength 42CrMoVNb steel
Z.G. Yang a,
, G. Yao a, G.Y. Li a, S.X. Li a, Z.M. Chu b, W.J. Hui c, H. Dong c,
Y.Q. Weng c
a
Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences, Shenyang 110016, PR China
b
Beijing Research Institute of Mechanical and Electrical Technology, Beijing 100083, PR China
c
Central Iron and Steel Research Institute, Beijing 100081, PR China

Received 23 April 2003; received in revised form 8 October 2003; accepted 28 January 2004

Abstract

The fatigue properties of high strength 42CrMoVNb steel with three fine grain sizes produced by different heat treatment pro-
cedures are studied in this paper. The experimental results illustrate that there is no horizontal asymptote in fatigue S–N curves
for smooth specimens at 106–107 cycles regime, hence the conventional fatigue limit eliminates. The fatigue crack initiation sites in
smooth specimens were related closely to fatigue life, and most of the fracture origins initiated at the inclusions. The diameters
and locations of these inclusions were measured, and their morphology was observed in SEM. The critical size of inclusion and
the appropriate grain size of prior austenite, which should be controlled in practical production, are obtained by analyzing the
experimental data.
# 2004 Elsevier Ltd. All rights reserved.

Keywords: High strength 42CrMoVNb steel; Fine grain; Inclusion; Fatigue strength; Fatigue crack initiation site

1. Introduction steels consist of a few large ones and clouds of small

ones. But these few large inclusions, which are hard to
In order to improve the properties of the steels, a avoid in clean steel, can prove to be catastrophic [5].
surge to seek ultra fine grained technology, purified Therefore, in order to get the best comprehensive
technology and uniformity technology took place properties of high strength steel, how to harmonize the
recently [1]. It is known that grain refinement is a relation of grain size and inclusion size to achieve
promising way to enhance the tensile strength and higher tensile strength and fatigue strength, is currently
toughness of steels. It is demonstrated that the Hall– a hot topic of steel research [6].
Petch relation holds for yield strength and ultimate ten- The materials studied in this work are fine-grained
sile strength when the grain size of steels decreases high strength 42CrMoVNb steel to be used for auto-
down to 1 lm [2–4]. Generally speaking, the higher the motive structures and components. The fatigue pro-
tensile strength, the higher the fatigue strength will be. perty of the steel is one of the key issues for
However, the inclusion plays a very crucial role in the improvements of the mechanical performance. In this
fatigue fracture process for high strength steels. The paper, the fatigue behaviors and micro-fracture char-
application of secondary refining techniques and non- acteristics of these steels are studied in detail, and the
metallic inclusion reduction techniques in steel pro- relations between the properties and microstructures
duction processes has greatly reduced the size and (inclusions and grains) of these steels are discussed.
amount of inclusions in steels. The inclusions in clean
2. Materials, specimens and experimental method


Corresponding author. Tel.: +86-24-8397-8023; fax: +86-24-2389-
1320. Based on 42CrMo, the high strength 42CrMoVNb
E-mail address: [email protected] (Z.G. Yang). steel has been designed, and its main chemical
0142-1123/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2004.01.009
960 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966

composition consists of (wt%) [7]: 0.42 C, 0.17 Si, 0.31 3. Results

Mn, 0.001 S, 0.004 P, 1.10 Cr, 0.98 Mo, 0.32 V, 0.04 Nb,
3.1. S–N curves and fatigue strength
0.01 Al, and the balance Fe. Steel of 18 mm diameter
were produced by hot rolling after smelting in electric The S–N curves of smooth specimens are shown in
furnace, then machined to rough specimens, and heat Fig. 3, for which no horizontal asymptote but gradient
treated. Finally, standard rotating bending fatigue exists at 106–107 cycles regime, hence the conventional
specimens (Fig. 1) were prepared for testing, and corre- fatigue limits cannot be determined. At 107 cycles, the
spondingly the circularly notched rotating bending fatigue strengths of smooth and notched specimens,
fatigue specimens were also prepared for testing of the r1 and r1N, were figured out by the staircase method
notch sensitivity. Both the depth of the notch and the in order to raise the confidence (Fig. 4). Notch sensi-
radius of the notch tip are 0.75 mm. The high strength tivity coefficient q could be calculated from the formula
steels used in the present work are designated as [8]
ADF1-880, ADF1-920 and ADF1-940 in terms of vari- Kf  1
q¼ ð1Þ
ous quenching temperatures. Table 1 shows the heat Kt  1
treatment procedures, prior austenite grain sizes, and here Kf ¼ r1 =r1N , and Kt ¼ 1:89 is a notch stress
tensile properties. The prior austenite grain sizes and concentration factor in the present experiment.
appearances are shown in Fig. 2. All fatigue tests for The fatigue properties of three steels are shown in
smooth and notched specimens were carried out under Table 2. The fatigue strength of smooth specimens does
rotating bending at a rotation rate of 5000 rpm, and not change monotonically with prior austenite grain
the compressed air was used to eliminate the possible size altering but increases monotonically with tensile
‘‘self-heating’’ in specimens during testing. strength (Fig. 5). The ADF1-920 with the highest ten-
sile strength has the highest fatigue strength, however,
the ADF1-940 with the lowest tensile strength has the
highest fatigue strength of notched specimens.

3.2. Fatigue crack initiation

3.2.1. Fractographic observation

Observed by SEM, three kinds of fatigue crack
initiation sites in the fractured specimens were found:
Fig. 1. Specimen configuration for rotating bending test. internal inclusion (I), surface inclusion (SI) and surface

Table 1
Heat treatment procedures, prior austenite grain sizes and tensile properties of 42CrMoVNb

Steel code Heat treatment procedure d (lm) rs (MPa) rb (MPa) rs/rb

v
ADF1-880 920 C vacuum furnace preheating ð30 minÞ þ oil cooling þ electricity rapid 1.7 1370 1440 0.95
v v v
heating (70–100 C/s) to 880 C þ oil quenching three times þ 600 C
tempering (2 h)
v
ADF1-920 920 C vacuum furnace preheating ð30 minÞ þ oil cooling þ electricity rapid 4.0 1465 1540 0.94
v v v
heating (70–100 C/s) to 920 C+oil quenching þ 600 C tempering (2 h)
v v
ADF1-940 940 C salt furnace preheating ð30 minÞ þ oil quenching þ 600 C tempering 8.0 1315 1400 0.94
(2 h)

Fig. 2. The appearances of prior austenite grain boundary networks of the steels: (a) ADF1-880, (b) ADF1-920, (c) ADF1-940.
 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966 961

Table 2
Fatigue properties (at 107 cycles)

Materials r1 (MPa) r1N (MPa) q

code

ADF1-880 746  20 421  20 0.87

ADF1-920 767  19 417  10 0.94
ADF1-940 738  22 446  20 0.75

Fig. 3. S–N curves of smooth specimens of 42CrMoVNb after three

heat treatments.

Fig. 5. Monotonic relationship of r-1 to rb for smooth specimens of

three steels.

these inclusions were divorced from the matrix and the

most of them were cracked and broken.
There are several radial ledges around the inclusion
Fig. 4. Experimental result of smooth specimens of ADF1-920 by (Fig. 7), which were formed by propagation and link-
the staircase method. age of primary micro-cracks at different planes, mean-
while the tire patterns emanating from the inclusions
matrix (S). Most fracture origins are internal inclu- were found in a few specimens, the direction of which
pffiffiffiffiffiffiffiffiffi is parallel to the specimen surface (Fig. 8).
sions, and their diameters / (calculated from area)
and depths D were measured, as listed in Tables 3–5.
3.2.2. Factors affecting the type of fatigue crack
The appearance of an inclusion can be seen in Fig. 6.
initiation site
These inclusions, containing Al, Mg, Ca and O, most As seen in Tables 3–5, when rmax 4r1 and Nf < 106
likely consist of Al2O3, MgO and CaO. In addition, all cycles, fatigue cracks mostly initiated at the specimen

Table 3
Fatigue cracking data of ADF1-880

Specimen no. Applied stress, Fatigue life, Stress ratio, Fatigue crack Distance, Inclusion
rmax (MPa) Nf (cycle) rmax/r1 initiation sitea D (lm) diameter, / (lm)

ADF1-880-09 725 5:85 106 0.97 I 260 33

ADF1-880-08 725 1:027 107 0.97 I 146 25
ADF1-880-19 725 1:28 107 0.97 SI 0 18
ADF1-880-18 750 5:14 106 1.01 I 322 37
ADF1-880-07 750 6:60 106 1.01 I 66 18
ADF1-880-06 750 9:14 106 1.01 I 520 45
ADF1-880-11 775 1:68 106 1.04 I 145 40
ADF1-880-12 775 6:40 106 1.04 SI 0 8
ADF1-880-17 775 8:05 106 1.04 I 66 35
ADF1-880-15 800 1:24 106 1.07 I 50 45
ADF1-880-20 850 1:90 105 1.14 S – –
ADF1-880-22 850 2:00 105 1.14 S – –
ADF1-880-23 900 1:27 104 1.21 S – –
ADF1-880-21 900 9:53 104 1.21 SI (0.17 17):(35.15 15)
a
I, internal inclusion; SI, surface inclusion; S, surface matrix.
962 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966

Table 4
Fatigue cracking data of ADF1-920

Specimen no. Applied stress, Fatigue life, Stress ratio, Fatigue crack Distance, Inclusion
rmax (MPa) Nf (cycles) rmax/r1 initiation sitea D (lm) diameter, / (lm)
ADF1-920-06 750 9:97 106 0.98 I 247 40
ADF1-920-08 775 3:025 105 1.01 I 24 44
ADF1-920-05 775 7:79 106 1.01 I 215 29
ADF1-920-10 775 9:91 106 1.01 I 253 36
ADF1-920-01 800 8:93 105 1.04 I 36 17
ADF1-920-04 800 1:84 106 1.04 I 68 23
ADF1-920-12 800 2:72 106 1.04 I 105 22
ADF1-920-14 850 2:27 106 1.11 I 91 38
ADF1-920-16 875 1:70 106 1.14 I 180 29
ADF1-920-15 900 1:03 105 1.17 SI 0 27
ADF1-920-17 900 1:77 105 1.17 SI 0 17
ADF1-920-19 950 1:37 105 1.24 SI 0 21
ADF1-920-18 950 1:40 105 1.24 SI 0 38
ADF1-920-20 1000 8:70 104 1.30 SI 0 10
a
I, internal inclusion; SI, surface inclusion; S, surface matrix.

Table 5
Fatigue cracking data of ADF1-940

Specimen no. Applied stress, Fatigue life, Stress ratio, Fatigue crack Distance, Inclusion
rmax (MPa) Nf (cycles) rmax/r1 initiation sitea D (lm) diameter, / (lm)

ADF1-940-15 725 3:60 106 0.98 I 38 10

ADF1-940-17 725 5:70 106 0.98 I 75 17
ADF1-940-14 750 5:10 105 1.02 I 34 46
ADF1-940-19 750 7:19 105 1.02 I 27 37
ADF1-940-03 750 7:04 106 1.02 I 170 10
ADF1-940-13 775 3:59 105 1.05 SI 0 18
ADF1-940-05 775 2:19 106 1.05 I 187 45
ADF1-940-02 800 3:85 105 1.08 S – –
ADF1-940-01 800 9:98 105 1.08 S – –
ADF1-940-12 800 2:50 106 1.08 I 318 41
ADF1-940-11 825 8:12 104 1.12 S – –
ADF1-940-10 850 1:20 103 1.15 S – –
a
I, internal inclusion; SI, surface inclusion; S, surface matrix.

surface of three steels. Surface cracks started usually at men surface, lm; R, the diameter of working section of
surface inclusions for ADF1-920 and at surface matrix specimen and R ¼ 4000 lm. When only the interior
for other two lower strength steels, except two speci- inclusions are considered and rL =r1 < 1, we can find
mens of ADF1-940, in which cracks initiated at inclu- the fatigue life of all specimens is above 106 cycles. So,
sions inside with extraordinary large sizes and at short the ratio of local stress to fatigue strength is an impor-
distances from the surface. But when rmax r1 and tant factor affecting the type of crack initiation.
Nf > 106 cycles, the fatigue cracks for all the fractured Moreover, under the present experiment conditions,
specimens initiated at interior inclusions. a simple linear relationship between the internal inclu-
According to the stress distribution in the cross sec- sions’ size / and location D can be found
tion of a specimen under rotating bending, we can
assume, instead of rmax, the local stress at inclusion
center, rL, / ¼ 12:5 þ 0:076D ð3Þ
 
D
rL ¼ rmax 1  ð2Þ among filled symbols that represent the local stress
R
rL =r1 < 1, as shown in Fig. 9. This means that the
where rmax is the maximum stress applied, MPa; D, the diameter / of the cracking inclusions should increase
shortest distance from an inclusion center to the speci- with increase in the distance D. But the open symbols
 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966 963

Fig. 9. Relationship between / and D of interior inclusions as

fatigue crack initiation sites at two stress ranges (filled symbols:
rL =r1 < 1; open symbols: rL =r1 > 1).

that represent the local stress rL =r1 > 1 were all

above the line (fitted by solid symbols).

4. Discussion
Fig. 6. An internal inclusion as fracture origin on the fracture sur-
face observed by SEM, showing its global and broken appearance. 4.1. Elimination of fatigue limit for high strength steel
(a) Location of an inclusion; (b) appearance of an inclusion (speci-
with fine grain
men ADF1-920-06#: rmax ¼ 750 MPa, Nf ¼ 9:97 106 cycles,
D ¼ 247 lm, / ¼ 40 lm).
The results presented above show that all three steels
cracked from inclusions when rmax r1 . Their fatigue
lives are within 3:6 106 1:28 107 cycles. There is no
conventional fatigue limit at 107 cycles. Miller and
O’Donnell [9] and Murakami et al. [10–12] have dis-
cussed possible factors which may cause fatigue failure
in the regime of Nf > 107 cycles. According to their
viewpoint, it is not so easy to identify the crucial mech-
anism, among many possible factors for eliminating the
fatigue limit. However, special attention should be paid
to the influence of the hydrogen trapped by non-
metallic inclusions in high strength steels.
Fig. 7. Radial ledges around an inclusion served as crack initiation
site (specimen ADF1-880-17#: rmax ¼ 775 MPa, Nf ¼ 8:05 106 4.2. Influence of grain size on the tensile and fatigue
cycles, D ¼ 66 lm, / ¼ 35 lm). properties of high strength steel

Morris et al. [13] have pointed out that in the case

of martensitic steel the adequate have measure of grain
size was the size of the crystallographically coherent
block or packet of tempered martensites, which was
regarded as the effective size affecting the strength. In
our studies, grain refinement does effectively reduce the
lengths of the laths, so the size of blocks decreases.
According to Hall–Petch’s equation, the strengths of
steels should have increased with decrease in grain
sizes, however, the highest strength is of ADF1-920
with blocks of medium size, not of ADF1-880 with the
Fig. 8. Tire patterns from an inclusion served as crack initiation site smallest sized blocks (Fig. 10), which means there
(specimen ADF1-880-06#: rmax ¼ 750 MPa, Nf ¼ 9:14 106 cycles, should be other factors influencing the strength. The
D ¼ 520 lm, / ¼ 45 lm). factors affecting the strength may, at the least, include:
964 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966

Fig. 10. Relation between austenite grain size and tensile strength.

(1) For present low alloy steels with microalloy

elements V and Nb, the smaller size of laths leads to
the increase of the specific surface of laths, which can
decrease the amount of microalloy element in the
interior of the laths, then precipitation carbides reduce
and the strength of matrix reduces. As compared with
ADF1-940, no obvious carbide precipitated in ADF1-
880 after tempering, and acicular carbide in ADF1-940
perpendicular to lath longitudinal direction can resist
the cleavage fracture along the direction of lath, as
Fig. 12. Schematic diagram illustrating variant crack initiation sites
shown in Fig. 11. (2) On increasing the times of quench-
in finer or coarser grains (a) Finer grains and (b) Coarser grains.
ing and tempering, the dislocation density becomes
greater. Such a condition can enhance the strength but
at the sametime can effectively obstruct the precipi- schematically given in Fig. 12. Under larger tensile
tation of carbides, which then reduces the strength. stress, the smaller grains can withstand less strain in
More quenching and tempering times will bring more terms of their strength and more uniform plastic defor-
vacancies in the steel that can also reduce the strength. mation, so the strain will be imposed on the inclusion,
2 more vacancies in steel can cause the premature which can be so serious that micro-cracks form finally
fragmentation of inclusion and decohesion between one in this inclusion and cause its breakage. Compara-
inclusion and the surrounding grains [14]; this is a tively, because the localized plastic deformation will
favorable factor for the early formation of micro- occur in the coarser grains rather easily, the slip lines
cracks. Thus, the strength of ADF1-880 is not and slip steps are able to impinge the interface between
improved when compared to that of ADF1-940. the matrix and inclusion. Then the micro-cracks form
Besides, the grain size may affect micro-crack directly at the interface rather easily. In both the cases
initiation during tensile testing by other mechanisms, as cracking inclusions will lead to lower tensile strength of
the steels. The appropriate grain size may not cause the
micro-cracks to initiate in the interior of the inclusion
and at the interface between the inclusion and the
metal matrix as easily as in finer or coarser grains.

4.3. Inclusion and grain size control

The size and location of inclusions in high strength

steel components under rotating and bending loading
should be controlled strictly in terms of their important
influence on the fatigue properties. Fig. 13 represents
the relation between sizes and locations of inclusions as
crack initiation sites with fatigue life above 106 cycles,
in which we can find three approximately parallel lines.
Each line denotes the critical size of inclusion as the
Fig. 11. The appearances of precipitates in: (a) ADF1-940 and (b) crack initiation site, above which the larger the dis-
ADF1-880. tance to the line the shorter the fatigue life of speci-
 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966 965

that the appropriate prior austenite grain size for

ADF1 steel is around 4 lm.

5. Conclusions

1. The S–N curves of smooth specimens of three steels

have no horizontal asymptotes in the 106–107 cycles
regime, and the conventional fatigue limits cannot
be determined. The fatigue strength of smooth speci-
mens does not change monotonically with alteration
in the prior grain size but increases monotonically
with tensile strength.
Fig. 13. Relationship between / and D of inclusions (Nf > 106
cycles). 2. Three kinds of fatigue crack initiation sites in the
fractured specimens were found: internal inclusion
(I), surface inclusion (SI) and surface matrix (S).
mens, and below which the fatigue failure is very diffi- When rmax r1 , fatigue cracks of the most frac-
cult to take place. tured smooth specimens initiated at interior inclu-
We can obtain the safe critical values of initiations in sions; when rmax 4r1 , the cracks generally occurred
these high strength steels from Fig. 13: (1) when in the surface matrix.
D ¼ 0 100 lm, / ¼ 8 14 lm; (2) when D ¼ 100 300 lm, 3. Grain refinement has complex influence on fatigue
/ ¼ 14 28 lm; and (3) when D > 300 lm, / > 30 lm. strength of high strength steels. There is an appro-
For ADF1-880, ADF1-920 and ADF1-940, if fatigue priate grain size that increases the fatigue strength.
crack initiation site is not from the inclusion, the The inclusion plays a very important role in fatigue
maximum size of inclusion will be less than 8, 15.5 and
strength of high strength steels, and there is a critical
7.5 lm, respectively. ADF1-880 and ADF1-940 have
size of inclusions, below which the fatigue fracture
more rigorous requirement of inclusion size than
origins will not be initiated from the inclusions.
ADF1-920. In fatigue testing, the micro-crack initiation
may be divided into two stages. At first the inclusion or Acknowledgements
the interface between the inclusion and the matrix
cracks, and then the micro-crack is initiated in the This work was supported by the project ‘‘Fundamen-
adjacent grains. ADF1-880 has the finest grains tal research on new generation of iron and steel in
(1.7 lm) and most of its fatigue life is consumed during China’’ (No. G19980615). The authors wish to thank
the period of crack initiation in the matrix near the Professor W.C. Yu, Dr. Z.X. Yin, Ms. H. Zhou,
inclusion due to uniform deformation occurring in the Senior engineers W.K. Gong and H.H. Su for their
finer grains. However, once the fatigue crack is formed experimental supports.
in the matrix, it will lead to failure quickly because of
its smaller fatigue crack growth threshold and higher
crack growth rate [15]. ADF1-940, which possesses the
coarsest grains (8 lm) can bring about more localized References
slip activities relatively, which causes the fatigue crack,
to initiate in the matrix grains that are close to the [1] Weng YQ. Development of ultrafine grained steels in China.
inclusion rather easily. Although it has larger fatigue In: Proceedings of the Workshop on New Generation Steel (NG
STEEL’2001), 13–16 November. Beijing (China): The Chinese
crack growth threshold and higher resistance to crack
Society for Metals; 2001. p. 1–7.
growth, the fatigue strength is still close to ADF1-880. [2] Niikura M, Yokota Y, Sato K, Shirota T. Ultra refinement of
For ADF1-920 with a medium grain size (4 lm), the grain size and its effect on mechanical properties. In: Proceedings
more difficult micro-crack initiation than ADF1-940 of the Workshop on New Generation Steel (NG STEEL’2001),
(8 lm) in the matrix adjacent to the inclusion and the 13–16 November. Beijing (China): The Chinese Society for
Metals; 2001. p. 100–3.
larger crack growth threshold than ADF1-880 (1.7 lm) [3] Takaki S, Kawasaki K, Kimura Y. Mechanical properties of
made it the sample with the highest fatigue strength. It ultra fine grained steels. J Mater Process Technol 2001;117:359.
is obvious that fatigue crack initiates more easily in [4] Shin DH, Kim DJ, Choo WY. Grain refinement of a commer-
coarser grains but grow faster in smaller grains. Assur- cial 0.15% C steel by equal-channel angular pressing. Scripta
edly, there is an appropriate grain size that makes both Mater 1999;41:259.
[5] Zhou S, Murakami Y, Beretta S, Fukushima Y. Experimental
the initiation and the growth of fatigue crack relatively investigation statistics of extremes for three-dimensional
difficult. The highest fatigue strength may be reached distribution of non-metallic inclusions. Mater Sci Technol 2002;18:
around this grain size. In the present study, it seems 1535–1543.
966 Z.G. Yang et al. / International Journal of Fatigue 26 (2004) 959–966

[6] Choo W-Y. Present developing status of high performance steel [10] Murakami Y, Namoto T, Ueda T. Factors influencing the mech-
in Korea. In: Proceedings of the Workshop on New Generation anism of superlong fatigue failure in steels. Fatigue Fract Eng
Steel (NG STEEL’2001), 13–16 November. Beijing (China): The Mater Struct 1999;22:581–90.
Chinese Society for Metals; 2001. p. 16–26. [11] Murakami Y, Namoto T, Ueda T, Murakami Y. Fatigue Fract
[7] Hui WJ, Dong H, Wang MQ, Chen SL, Weng YQ. Effect of Eng Mater Struct 2000;23:893–902.
heat treatment on mechanical properties of second-hardening [12] Murakami Y, Namoto T, Ueda T, Murakami Y. Fatigue Fract
high strength 42CrMoVNb steel. In: Proceedings of the Work- Eng Mater Struct 2000;23:903–10.
shop on New Generation Steel (NG STEEL’2001), 13–16 [13] Morris Jr. JW, Guo Z, Krenn CR, Kim YH. The limits of
November. Beijing (China): The Chinese Society for Metals; strength and toughness in steel. ISIJ Int 2001;41(6):599–611.
2001. p. 375–8. [14] Kuhlmann-Wilsdorf D, Thomason PF. The role of vacant lattice
[8] Suresh S. Fatigue of materials [Wang ZG, Trans. in Chinese]. sites in the low-amplitude fatigue failure qt inclusions in steel.
Beijing: National Defense Industry Press; 1999. p. 173. Acta Metall 1982;30:1243–5.
[9] Miller KJ, O’Donnell WJ. The fatigue limit and its elimination. [15] Suresh S. Fatigue of materials [Wang ZG, trans. in Chinese].
Fatigue Fract Eng Mater Struct 1999;22:545–57. Beijing: National Defense Industry Press; 1999. p. 248.