Effect of loading type on fatigue properties of high strength bearing steel in very high cycle regime
W. Li,T. Sakai,Q. Li,L.T. Lu,P. Wang
DOI: https://doi.org/10.1016/j.msea.2011.03.020
2011-01-01
Abstract:Very high cycle fatigue tests under axial loading at frequencies of 95 Hz and 20 kHz were performed to clarify the effect of loading type on fatigue properties of a high strength bearing steel in combination with experimental result of this steel under rotating bending. As a result, this steel represents the single P - S - N (probabilistic-stress-life) curve characteristics for surface-induced fracture and interior inclusion-induced fracture, just like that under rotating bending. However, fatigue strength is lower, where the run-out stress at 10 9 cycles is evaluated to be 588 MPa, less than that under rotating bending with about 858 MPa. Occurrence probability of larger and deeper inclusion-induced fracture is much higher than that under rotating bending. Furthermore, the formation process of fine granular area (FGA) is independent of the type and frequency of loading, which is very slow and is explained as the crack nucleation process under the special dislocation mechanism. The stress intensity factor range at the front of FGA, Δ K FGA , is approximately regarded as the threshold value controlling the stable propagation of interior crack. For the control volume of specimen under axial loading, the estimated value of fatigue limit by FGA is similar to experimental run-out stress value at 10 9 cycles, but that by inclusion is larger. However, the corresponding estimated results under rotating bending are all conservative. Keywords Very high cycle fatigue Loading type Bearing steel Interior crack initiation Fatigue limit 1 Introduction Fatigue fracture of metallic material in very high cycle regime has aroused wide concern in the past ten years, especially for some high strength steels with tensile strength more than 1200 MPa [1–8] . Because of two competitive fracture modes consisting of surface-induced fracture and interior-induced fracture, fatigue S - N data for these high strength steels sometimes exhibit a considerable amount of scatter, and that can represent the duplex S - N curves characteristics [1] or the stepwise S - N curves characteristics [2] . Although the essence of this special S - N property is not well understood as yet, some uncertainty factors such as the size [3] and spatial distribution [4] of metallurgical defect, the control volume [5] and surface treatment condition [6] of specimen, the loading type [7] , and the experimental environment [8] , etc, are considered to have a close relation to the occurrence of these two fracture modes. It is well known that the interior inclusion-induced fracture with fish-eye is a remarkable fracture mode for some high strength steels in very high cycle regime. Sometimes, a characteristic rough area indicating fine granular morphology and named as the “fine granular area (FGA)” by authors [9] , or the “granular bright facet (GBF)” by Shiozawa et al. [10] , or the “optically dark area (ODA)” by Murakami et al. [11] , occurs in the vicinity of inclusion. In this paper, it is called the FGA. Now, it is generally believed that the formation process of FGA plays a very important role in the process of interior crack initiation and propagation of high strength steel in very high cycle regime. Unfortunately, no one can come up with a convincing explanation on its formation mechanism as yet, especially under different loading type. Thus, it is very necessary to make a further study on clarifying the relationship between the type and frequency of loading, and FGA, and deeply understanding the crack initiation and propagation process within FGA and the fracture mechanical property of FGA in the process of entire crack propagation, etc. However, high strength steels have been widely used as structural material to realize the small size of the product and to improve the structural reliability. In addition to the effect of external environment, it is apparent that fatigue strength of high strength steel must be heavily relied on the intrinsic property of material such as the inclusion size, and the applied loading type. At the same time, the difference of loading type is also represented in the difference of control volume. For example, the control volume under rotating bending (RB) is much less than that under axial loading (AL), which has been better defined in early works of authors [12] . Besides, although the FGA does not exist before fatigue test, as a key transition region of interior crack propagation its size must be also related to fatigue strength. Therefore, the reliability evaluation on fatigue strength of high strength steel based on these factors is a subject very worthy to be studied. In the present study, a high strength bearing steel was studied to clarify its very high cycle fatigue properties under AL including the conventional axial loading (CAL) with usual frequency and ultrasonic axial loading (UAL). At the same time, in combination with experimental result of the same batch of steel under RB [12] , the effect of loading type on very high cycle fatigue properties was discussed, with emphasis on the probabilistic S - N characteristics, the crack initiation and propagation mechanism within FGA, and the evaluation on maximum inclusion size and fatigue limit. 2 Experimental procedure 2.1 Material and specimen Material used in this study was a high strength bearing steel (GB: GCr15), whose chemical composition (mass percentage) is: 0.99C, 0.25Si, 0.30Mn, 0.013P, 0.005S, 1.45Cr, 0.10Cu, 0.06Ni and 0.02Mo. Firstly, from the annealed material, specimens were machined into the shape of hourglass with a certain amount of finishing allowance. Next, they were quenched in vacuum (1113 K × 15 min + oil cooling) and tempered (473 K × 120 min + air cooling). The surfaces of specimens under CAL and UAL were all grinded in a direction perpendicular to the axis of specimen by the grade 600–2000 abrasive paper to the final shapes as shown in Fig. 1 . The minimum diameters of specimens are all 3 mm, and the round-notch radiuses are 15 mm and 31 mm, respectively. The corresponding elastic stress concentration factors are evaluated to be 1.04 and 1.03, respectively. The Vickers hardness (HV) distribution along the cross section of specimen is almost uniform, about 703 kgf/mm 2 . The tensile strength and yield strength are about 2300 MPa and 1617 MPa, respectively. Furthermore, based on observation by means of scanning electronic microscope (SEM) under the conditions of Nital-etching and electro-polishing, the microstructure of heat-treated material is tempered martensite structure where a number of spherical carbide particles are randomly distributed in the matrix, as shown in Fig. 2 . 2.2 Fatigue testing method A resonant testing machine (TESTRONIC-250) and an ultrasonic testing machine (USF-2000) were adopted to carry out the very high cycle fatigue test of GCr15 steel under AL with stress ratio R = −1, whose frequencies are 95 Hz and 20 kHz, respectively. Fatigue tests were all performed in an open environment at room temperature. During the process of UAL, the intermitted loading was applied to avoid the “self-heat” of specimen with fixed pulsing period is 150 ms. However, the pausing period is dependent on the applied stress amplitude, i.e. the higher stress amplitude corresponds to the longer pausing period. Meantime, the compressed air was used to improve the cooling of specimen so as to keep the surface temperature of specimen below 303 K. The detailed testing process is seen in Ref. [13] . After the experiment, fracture surfaces of all the failed specimens were carefully observed by the optical microscope (OM) and SEM, paying an attention to the crack initiation sites and crack initiation and propagation mechanisms. 3 Experimental results and discussions 3.1 S - N and P - S - N characteristics The S-N diagram of GCr15 steel under CAL and UAL is shown in Fig. 3 . Based on the observations of all fracture surfaces, fatigue fracture of specimen is clearly divided into two competitive fracture modes: surface-induced fracture and interior-induced fracture. Furthermore, it should be noted that the obtained S - N data under CAL have a good agreement with that under UAL, which means that the effect of frequency is very small on fatigue strength of this steel. Therefore, considered all S - N data as a whole, the S - N curves of GCr15 steel under AL, corresponding to surface-induced fracture and interior-induced fracture, can be indicated by a dash line and a solid line in Fig. 3 , respectively. They all exhibit the continuously declining trend and conventional fatigue limit cannot be observed. For the sake of comparison, the obtained S - N curves and data points of this steel under RB, similarly corresponding to these two fracture modes, are also shown in Fig. 3 . Under RB, fatigue fracture of this steel can be clearly divided into three regions: (i) the surface fracture region (about N < 10 6 ), (ii) the co-existence region of surface fracture and interior fracture (10 6 ≤ N ≤ 10 7 ) and (iii) the interior fracture region ( N ≥ 10 7 ). By contrast, the results under both loading types all indicate that the interior-induced fracture is the predominant fracture mode of this steel in ultra-long life region of N > 10 7 . However, the co-existence region under AL is much wider (about 10 3 ≤ N ≤ 10 7 ) and there is hardly the solo surface fracture region. In other words, even in the short life region, the interior-induced fracture still can take place. Furthermore, another obvious difference is that fatigue strength under AL is much lower than that under RB. Based on the interior S - N curves fitted by least square method, the run-out stress value at 10 9 cycles of this steel under AL, σ run-out , is evaluated to be 588 MPa, much less than that under RB with about 858 MPa. This is all attributed to the different stress distribution on the minimum cross-section of specimen under these two loading types. Apparently, there exists the effect of stress gradient under RB. Under the same stress amplitude, the stress is uniform on the entire cross-section under AL, whereas the stress is the largest on the surface layer and gradually decreases to zero at the center of cross-section under RB. Since the material defects such as the non-metallic inclusions or some inhomogeneous microstructures are supposed to be randomly distributed in the material, the number of critical defects likely causing the fatigue crack under AL, also considered as “damage area”, becomes much larger than that under RB. Therefore, from the consideration of probability and statistics, if the number of critical defects becomes large, fatigue strength must tend to decrease and the possibility of interior-induced fracture also become larger. In addition, compared with the duplex S - N curves characteristics exhibited by other high strength steels under AL [1,6] , the separation of S - N curves corresponding to these two fracture modes for GCr15 steel under AL is not so distinct. Therefore, the single probabilistic S - N curve is proposed to describe the very high cycle fatigue property of GCr15 steel under AL just like that under RB. Based on the statistical method in JSMS-SD-6-04 [14] , the nonlinear S - N curve model is used to establish the P - S - N curves of GCr15 steel under AL, corresponding to the failure probabilities of P ( N f ) = 1%, 10%, 50%, 90% and 99%, as shown in Fig. 4 . 3.2 Observation of fracture surfaces As above-mentioned, fatigue fracture of specimen under AL can be classified into surface-induced fracture and interior-induced fracture just like that under RB. The surface-induced fracture is mainly induced by the surface fine flaw or inclusion. The interior-induced fracture is entirely induced by the interior inclusion. A typical interior fracture surface under CAL observed by OM is shown in Fig. 5 (a) , it can be seen that an isolated fish-eye is situated on the fracture surface and the inclusion is basically situated at the center of fish-eye. At the same time, a dark rough area indicating fine granular morphology occurs around the inclusion, which is the FGA and is also the reason why it is called the ODA. Its higher magnification photograph is shown in Fig. 5 (b). Furthermore, the FGA is scarcely observed on the fracture surface with fatigue life less than 10 6 cycles under CAL or under UAL, as shown in Fig. 5 (c) and (d), respectively. The same experimental phenomena are also observed for this steel under RB, as shown in Fig. 5 (e) and (f). This means that the formation of FGA is independent on the type and frequency of loading. Furthermore, it can be clearly found that the location of inclusion causing fatigue fracture under AL is more close to the centre of fracture surface than that under RB. The detailed discussion will be carried out in the later section. 3.3 Formation mechanism of FGA The formation progress of FGA can be divided into three stages: (i) the formation of fine granular layer, (ii) the nucleation and coalescence of mirco-debonding and (iii) the formation of penny-shape crack. Firstly, because of the different thermal contraction coefficient between the inclusion and the adjacent martensite matrix, the severe stress concentration inevitably takes place on the interface between them under the longtime action of cyclic loading. The adjacent martensite laths begin to yield and induce the plastic deformation, and that leads to the non-uniform multiplication and movement of many dislocations in the vicinity of inclusion. However, owing to the obstruction of the densely distributed carbide particles as hard phase, these newly generated dislocations incessantly pile-up at boundaries between hard carbide particles and martensite laths, and can not move, which also results in the occurrence of stress concentration at boundaries. The degree of stress concentration tends to increase with the incessant piling-up of dislocation. In order to keep the coordination of integral deformation in cycle loading, the phenomenon of fragmental crystallite takes place in these martensite laths due to the extremely inhomogeneous characteristics of plastic deformation of each martensite lath resulting from the orientation difference. Consequently, the martensite laths in the given region around inclusion are gradually segmented into many fine packets and the dislocation cells are formed. Then the extensive polygonization and misorientation of many fine sub-grains take place in succession, which gradually cause a fine granular layer to occur in the given region. It should be noted that this phenomenon is only limited to this given region around inclusion due to the effect of stress concentration. Then, when the martensite laths are distinctly segmented and misorientation of newly generated fine sub-grains exceed a threshold value, the cell walls and interfaces between fine granular layer and matrix tend to separate, and some segments of micro-debonding gradually originate along the boundaries between fine granular layer and matrix. With the continuous initiation of micro-debonding, some of them coalesce to one another. When the nucleation and coalescence of the micro-debonding entirely spreads all over the fine granular layer, a penny-shape crack is formed around the inclusion, namely, a characteristic rough area is formed as the fine granular area (FGA). 3.4 Discussion based on fractography Based on the fractography, two crack size parameters in units of μm are defined to discuss the interior inclusion-induced fracture mechanisms under AL and RB. Among them, d inc denotes the depth of inclusion from the centre of inclusion to the nearest edge of fracture surface, and a r e a denotes the sizes of inclusion and FGA, defined as the square root of respective area, such as a r e a inc and a r e a FGA (including a r e a inc ). Fig. 6 shows the relationship between d inc and N f . The values of d inc under AL are in the range of 30–430 μm regardless of N f , whereas that under RB are only in the range of 15–125 μm and approximately tend to increase with the increase of N f . Furthermore, the distribution of the former is much wider and covers the entire life region, but that of the latter is only limited to the longer life region of N > 10 6 . This means that the occurrence probability of deeper inclusion-induced fracture under AL is much larger than that under RB, which is also attributed to the different stress distribution on the cross-section of specimen. Fig. 7 shows the relationship between a r e a inc and the real stress amplitude applied on the inclusion, σ at . Under AL, the sizes of inclusions are in the range of 13–36 μm regardless of σ at , in which the sizes of surface inclusions are in the range of 13–29 μm, and the sizes of interior inclusions without FGA and with FGA are in the range of 28–36 μm and 17–35 μm, respectively. Compared with the size of inclusion under RB, the size of surface inclusion is almost the same, but the size of interior inclusions is clearly larger. Especially, the size of interior inclusion with FGA is much larger than that under RB about in the range of 5–15 μm. This further proves the possibility that the larger interior inclusions do not always induce fatigue crack under RB due to the effect of stress gradient. In addition, Fig. 8 shows the relationship between a r e a FGA and σ at . The values of a r e a FGA under AL are in the range of 56–70 μm and thus larger than that under RB with about 17–45 μm. Considered as a whole, the values of a r e a FGA tend to decrease with the increase of σ at , as indicated by a dashed line in Fig. 8 . 3.5 Discussions based on fracture mechanics Based on the measured sizes of inclusion and FGA under AL, the stress intensity factor ranges at the front of inclusion and FGA in units of MPam 1/2 , Δ K inc and Δ K FGA , were calculated by the following equations [15] : For surface inclusion: (1) Δ K sur-inc = 0.65 σ a π a r e a inc For interior inclusion or FGA: (2) Δ K int-inc, or FGA = 0.5 σ a π a r e a inc, or FGA where Δ K sur-inc and Δ K int-inc denote the stress intensity factor ranges at the front of surface inclusion and interior inclusion, respectively. Fig. 9 shows the relationship between Δ K inc and N f . The values of Δ K sur-inc under AL are in the range of 4.25–4.92 MPam 1/2 regardless of N f , somewhat less than that under RB. As like the change trend of Δ K int-inc with N f under RB approximately indicated by a dashed line in Fig. 9 , the values of Δ K int-inc under AL also tend to decrease with the increase of N f , as indicated by a solid line in Fig. 9 . Thereinto, the values of Δ K int-inc without FGA and with FGA are in the range of 4.07–5.03 MPam 1/2 and 2.48–3.34 MPam 1/2 , respectively, which are similar to corresponding values of Δ K int-inc under RB. Furthermore, compared with the threshold stress intensity factor range of surface long crack for GCr15 steel, Δ K th , about 4–5 MPam 1/2 [16] , it can be found that the values of Δ K sur-inc and Δ K int-inc without FGA are almost the same, but the values of Δ K int-inc with FGA are clearly smaller. Furthermore, Fig. 10 shows the relationship between Δ K FGA and N f . The values of Δ K FGA under AL are in the range of 3.99–5.16 MPam 1/2 , very similar to that under RB about 4.07–5.58 MPam 1/2 , and approximately regardless of N f . Considered as a whole, the average value of Δ K FGA of GCr15 steel is evaluated to be 4.67 MPam 1/2 , also similar to the value of Δ K th . Therefore, Δ K FGA can be approximately regarded as the threshold value controlling the stable propagation of interior crack regardless of loading type, and fatigue property within FGA is explained as the interior crack nucleation process. In summary, the interior crack initiation and propagation mechanism induced by inclusion can be summarized as follows: if the size of the inclusion is large enough and exceeds the critical size corresponding to Δ K FGA , the induced crack can directly enter into the stable crack propagation stage, i.e. the formation period of fish-eye outside of inclusion. Simultaneously, this means that the FGA never occurs in the vicinity of such large sized inclusion. Conversely, if the size of the inclusion is relatively smaller, less than the critical size corresponding to Δ K FGA , the corresponding crack initiation and propagation mechanism is slightly complicated. Firstly, it is assumed that there is an interior threshold stress intensity factor range, Δ K int-th . Only when the value of Δ K int-inc corresponding to the smaller inclusion is greater than the value of Δ K int-th some micro-cracks can begin to initiate, coalesce and propagate, that is, the beginning of FGA formation. As the FGA is formed, the crack enters into the stable crack propagation stage, i.e. the formation period of fish-eye outside of FGA. Subsequently, as the fish-eye is formed, the crack enters into the unstable propagation stage until the final catastrophic fracture takes place [1,12] . In addition, since the smaller inclusion corresponding to Δ K int-th can be approximately equivalent to a small crack, the study on crack nucleation process within FGA can fall into the category of small crack propagation [17] . A prediction equation is proposed to evaluate the value of Δ K int-th based on the a r e a model proposed by Murakami and Endo [18] , defined as follows: (3) Δ K int-th = C ′ HV + 120 a r e a 1 / 3 where the units of Δ K int-th , HV and a r e a are defined as MPam 1/2 , kgf/mm 2 and μm, respectively, but parameter C′ is given as a dimensionless constant and determined as the intercept of the regression line for the plots of Δ K int-th /(HV + 120) versus a r e a diagram. Furthermore, the crack within FGA can be considered as the small crack [19] . Thus, based on Eq. (3) , the values of Δ K int-th and Δ K FGA under AL can be evaluated by the following equations: (4) Δ K int-th = 1.22 × 10 − 3 HV + 120 a r e a inc 1 / 3 (5) Δ K FGA = 1.37 × 10 − 3 HV + 120 a r e a FGA 1 / 3 The relationships between the estimated values of Δ K int-th and Δ K FGA , and their respective sizes are indicated by two solid lines in Figs. 11 and 12 , respectively. They are not constant and tend to increase with the increase of respective sizes, just like the statistical results under RB indicated by two dashed lines in Figs. 11 and 12 . However, it should be noted that even though the sizes of inclusion and FGA under AL are larger, the estimated value ranges of Δ K int-th and Δ K FGA are almost the same. Therefore, considered as a whole, the interior crack growth rate within FGA can be approximately evaluated by the following crack growth law [20] : (6) d a d N = C ( Δ K − Δ K int -th ) m where C and m are environmentally sensitive material-constants. Combined with the Eq. (2) , the Eq. (6) can be rewritten as: (7) Δ K inc − Δ K int-th m ( N f a r e a inc ) = 2 C ( m − 2 ) 1 − a r e a inc a r e a FGA ( m / 2 ) − 1 Since more than 90% of total fatigue life in very high cycle regime is consumed in the period of FGA formation [21] , so the fatigue life in this stage can be approximately equal to the total fatigue life. Thus, the relationship between Δ K inc and N f /( area inc ) 1/2 is indicated by a solid line in Fig. 13 and the interior crack growth rate equation within FGA of GCr15 steel can be expressed as: (8) d a d N = 6.85 × 10 − 11.17 ( Δ K − 2.47 ) 2.29 As a result, the value of Δ K int-th is evaluated to be 2.47 MPam 1/2 , which means that crack cannot propagate under the condition of Δ K < 2.47 MPam 1/2 . Under AL, the corresponding size of inclusion is evaluated to be 14.88 μm based on Eq. (4) , which can be considered as the maximum critical inclusion size causing crack non-propagation when the applied stress amplitude is less than or equal to the above-evaluated run-out stress value under AL. 3.6 Evaluation of fatigue limit Based on the statistics of extreme values (SEV) method [22] , the maximum sizes of inclusion and FGA in a given control volume of GCr15 steel, V (units: mm 3 ), was predicted. Firstly, it is confirmed that the measured sizes of inclusion and FGA under AL can be well characterized by the extreme value distribution just like the statistical result under RB, as shown in Fig. 14 . Take the case of inclusion, let X V denote the characteristic value of maximum inclusion size under AL in units of μm, the return period, T , and the cumulative probability of X V , P ( X V ), are given by: (9) T = V V 0 (10) P ( X V ) = 1 − 1 T where V 0 is the volume of standard inspection plane, S 0 , with a certain thickness. Under AL, S 0 corresponds to the area of minimum cross-section of specimen, about 7.06 mm 3 , and its thickness value is approximately defined as the mean size of inclusion, about 27.23 μm. So the value of V 0 is evaluated to be 192.38 × 10 −3 mm 3 . Therefore, the value of X V can be evaluated by the following equation: (11) X V = 24.42 − 6.94 ln − ln 1 − 192.38 1000 V The more detailed solution process is seen in Refs. [12,21] . Similarly, the characteristic value for maximum FGA size, X ′ v , under AL in units of μm is given by: (12) X ′ v = 52.78 − 7.54 ln − ln 1 − 400.29 1000 V Fig. 15 shows the predicted maximum sizes of inclusion and FGA as a function of V for GCr15 steel under AL and RB. They all exhibit the monotonously increasing trend with the increase of V , but the result under AL is larger. For the hourglass-shaped specimen used in this study, its control volume under AL is expressed as [21] : (13) V = 0.25 π l ( d ) 2 where d (=3 mm) is the diameter of minimum cross-section of specimen, l is the length of variable cross-section. Herein, the value of l is defined as the length at which the stress value on the cross-section is 0.9 times that on minimum cross-section at l /2. Therefore, the values of l under CAL and UAL are evaluated to be 3.12 mm and 4.48 mm, respectively. The corresponding values of V are about 22 mm 3 and 32 mm 3 , respectively, which are all much larger than that under RB, about 2.53 mm 3 [12] . Based on the Eqs. (11) and (12) , the corresponding maximum sizes of inclusion and FGA under CAL are evaluated to be 57 μm and 83 μm, and that under UAL are evaluated to be 60 μm and 86 μm, which are all larger than that under RB, about 41 μm and 59 μm [12] , respectively. Next, based on the above predicted maximum size of inclusion or FGA, fatigue limit, σ w , corresponding to interior inclusion-induced fracture can be evaluated by the following equation [23] : (14) σ w = 1.56 ( HV + 120 ) ( a r e a ) 1 / 6 It should be noted that fatigue limit herein evaluated by Eq. (14) is the threshold stress for non-propagation of crack originating from an initial defect or crack for the metallic materials containing small defects or cracks [22] . Fig. 16 shows the estimated values of σ w as a function of V for GCr15 steel under AL and RB. By contrast, the estimated values of σ w by inclusion or FGA under AL are somewhat less than that under RB. Corresponding to the control volumes of specimen in this study, the estimated values of σ w by inclusion and FGA under CAL are evaluated to be 654 MPa and 615 MPa, very similar to that under UAL about 649 MPa and 611 MPa, respectively. This means the effect of control volume of specimen under these two types of AL on fatigue strength of this steel is almost the same, which also is an important reason why the fatigue S - N data obtained under two types of AL shows a good consistency. Furthermore, it should be noted that the estimated value of σ w by FGA is very similar to the experimentally estimated value of σ run-out under AL (about 588 MPa), whereas that by inclusion is somewhat larger. This again confirms that fatigue design based on fatigue limit evaluated by the size of inclusion sometimes involves a certain amount of risk in very high cycle regime. However, the values of σ w estimated by inclusion and FGA under RB, 690 MPa and 654 MPa, are all less than the experimentally estimated value of σ run-out under RB (about 858 MPa), which seems to be conservative. This is mainly because fatigue strength of high strength steel under RB is not only dependent on the size of inclusion or FGA but also dependent on their distributions on the cross-section under the effect of stress gradient. Therefore, the more reasonable estimation of fatigue limit of high strength steel in very high cycle regime should be further studied. 4 Conclusions Main conclusions obtained in this study are summarized as follows: 1. Under axial loading, GCr15 steel represents the single probabilistic S - N curve characteristics in very high cycle regime, corresponding to surface-induced fracture and interior inclusion-induced fracture. The run-out stress value at 10 9 cycles is evaluated to be 588 MPa, less than that under rotating bending with about 858 MPa. 2. Regardless of loading type and testing frequency in this study, FGA scarcely occurs in the fatigue life less than 10 6 cycles. Its formation process is very slow and can be explained as the crack nucleation process under the special dislocation mechanism. Δ K FGA can be approximately regarded as the threshold value controlling the stable propagation of interior crack. 3. Occurrence probability of larger and deeper inclusion-induced fracture under axial loading is much higher than that under rotating bending. The values of Δ K int-inc tend to decrease with the increase of fatigue life, and the value of assumed Δ K int-th is evaluated to be 2.47 MPam 1/2 . 4. Based on a r e a model, the estimated value of σ w by FGA under axial loading is similar to the experimental run-out stress value at 10 9 cycles, whereas that by inclusion is larger. However, the estimated results for inclusion and FGA under rotating bending are all conservative. References [1] T. Sakai Y. Sato N. Oguma Fatigue Fract. Eng. Mater. Struct. 25 2002 765 773 [2] M.D. Chapetti T. Tagawa T. Miyata Mater. Sci. Eng. A 356 2003 227 235 [3] Z.G. Yang J.M. Zhang S.X. Li G.Y. Li Q.Y. Wang W.J. Hui Y.Q. Weng Mater. Sci. Eng. A 427 2006 167 174 [4] K.S. Ravi Chandran S.K. Jha Acta Mater. 53 2005 1867 1881 [5] Y. Furuya Scripta Mater. 58 2008 1014 1017 [6] K. Shiozawa M. Murai Y. Shimatani T. Yoshimoto Int. J. Fatigue 32 2010 541 550 [7] I. Marines-Garcia P.C. Paris C. Bathias Mater. Sci. Eng. A 468–470 2007 120 128 [8] M. Nakajima K. Tokaji H. Itoga H.N. Ko Fatigue Fract. Eng. Mater. Struct. 26 2003 1113 1118 [9] T. Sakai M. Takeda N. Tanaka M. Kanemitsu N. Oguma K. Shiozawa Trans. JSME 67A 2001 1805 1812 [10] K. Shiozawa L.T. Lu S. Ishihara Fatigue Fract. Eng. Mater. Struct. 24 2001 781 790 [11] Y. Murakami T. Nomoto T. Ueda M. Ohtori J. Soc. Mater. Sci. Jpn. 48 1999 1112 1117 [12] W. Li T. Sakai Q. Li L.T. Lu P. Wang Int. J. Fatigue 32 2010 1096 1107 [13] H. Mayer Int. J. Fatigue 28 2006 1446 1455 [14] JSMS-SD-6-04 Standard Evaluation Method of Fatigue Reliability for Metallic Materials-standard Regression Method of S - N Curves 2004 JSMS Tokyo [15] Y. Murakami S. Kodama S. Konuma Trans. JSME 54 1988 688 695 [16] S.L. Zheng L.P. Zhang Coal Technol. 21 2002 58 59 [17] M.D. Chapetti T. Tagawa T. Miyata Mater. Sci. Eng. A 356 2003 236 244 [18] Y. Murakami M. Endo Int. J. Fatigue 16 1994 163 182 [19] Y.B. Liu Z.G. Yang Y.D. Li S.M. Chen S.X. Li W.J. Hui Y.Q. Weng Mater. Sci. Eng. A 497 2008 408 415 [20] M. Klesnil P. Lukas Fatigue of Metallic Materials 1980 Elsevier Science New York [21] K. Shiozawa Y. Morii S. Nishino L. Lu Int. J. Fatigue 28 2006 1521 1532 [22] Y. Murakami Metal Fatigue: Effects of Small Defects and Nonmetallic Inclusions 2002 Elsevier Science Amsterdam & Boston [23] Y. Murakami T. Nomoto T. Ueda Fatigue Fract. Eng. Mater. Struct. 22 1999 581 590