You are on page 1of 9

International Journal of Fatigue 26 (2004) 1725 www.elsevier.

com/locate/ijfatigue

Evaluation and comparison of several multiaxial fatigue criteria


Ying-Yu Wang, Wei-Xing Yao
Department of Aircraft Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China Received 16 January 2003; received in revised form 29 April 2003; accepted 30 May 2003

Abstract In this paper several multiaxial fatigue criteria are reviewed. The criteria are divided into three groups, according to the parameters used to describe the fatigue life or fatigue strength of materials. They are stress criteria, strain criteria and energy criteria. Their predictive capabilities are checked against the experimental data of six materials under proportional and nonproportional loading. Among the stress criteria, the criterion of Lee is in the best agreement with the test data. Among the strain criteria, the Kandil, Brown and Millers criterion has the best correlation with the experimental data of the materials employed. The Farahanis criterion yields the most satisfactory result among the energy criteria. Its fatigue life correlation for 1045HR steel and 304 stainless steel fell within factors of 2 and 3, respectively. 2003 Elsevier Ltd. All rights reserved.
Keywords: Multiaxial fatigue; Nonproportional loading; Stress criteria; Strain criteria; Energy criteria

1. Introduction The components of engineering structures such as aircraft and automobiles usually undergo multiaxial loading. The cyclic stressstrain responses under multiaxial loading, which depend on the loading-path, are very complex and the fatigue behavior of materials and structures is very difcult to be described. Multiaxial fatigue criteria, whose aim is to reduce the complex multiaxial loading to an equivalent uniaxial loading, are very important in the study of multiaxial fatigue. Up to now, many researchers have proposed multiaxial fatigue criteria suitable to different materials and different loading conditions. There is not yet a universally accepted model in spite of a great number of criteria. Even though there are extensive reviews of multiaxial fatigue criteria, which were presented by Garud [1], Brown and Miller [2], You and Lee [3], Papadopoulos [4], Macha and Sonsino [5] and others, few critiques on multiaxial fatigue criteria with a lot of experimental data yet exist. In the present paper, the multiaxial fatigue criteria are reviewed and the most criteria published in literatures are classied into three categories, namely stress criteria,

strain criteria and energy criteria. Furthermore, in order to evaluate the capabilities of some popular criteria to correlate multiaxial fatigue experiments for both proportional and nonproportional loading conditions, fatigue experimental data of six different materials are quoted. Some various criteria are critically examined and the comparisons are made.

2. Stress criteria 2.1. Brief review of stress criteria Gough and Pollard [6,7] proposed two equations for metals under combined in-phase bending and torsion, namely the ellipse quadrant for ductile metals, Sb f 1
2

St t 1

(1)

and the ellipse arc for brittle metals, St t 1


2

Sb f 1

f t

1 1

f Sb 2 f 1 t

1 1

(2)

Corresponding author. Fax: +86-25-4891422. E-mail address: wxyao@nuaa.edu.cn (W.-X. Yao).

Considering the phase difference between loadings, Lee [8] modied the ellipse quadrant of Gough as follows

0142-1123/$ - see front matter 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0142-1123(03)00110-5

18

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

Nomenclature b, c axial fatigue strength exponent and axial fatigue ductility exponent, respectively b,c torsional fatigue strength exponent and axial fatigue ductility exponent, respectively E, G Youngs modulus and shear modulus, respectively f 1, t 1 fatigue limits in reversed bending and torsion, respectively amplitude of the second invariant of the stress deviator J2,a NP, NE predicted life and experimental life, respectively Sb, St bending and torsional stress amplitudes, respectively shear fatigue strength tA,B a material constant relating to the additional hardening b, k, l, S, C material constants e equivalent strain range between time A and time B e22, g21 the normal and shear strain ranges on the critical plane, respectively emax range of the maximum strain normal strain range on the critical plane en gmax range of the maximum shear strain s22, t21 the normal and shear stress ranges on the critical plane, respectively normal stress range on the critical plane sn t shear stress range on the critical plane tmax maximum shear stress range on the critical plane normal strain on the plane of gmax en en strain normal to plane of g en the normal strain excursion between adjacent turning points maximum shear strain gmax maximum shear strain on the plane p/4 to surface g smax, smax maximum absolute values of shear and normal stresses on the critical plane, respectively 12 22 amplitude of equivalent stress sa,eq s f, e f axial fatigue strength coefcient and axial fatigue ductility coefcient, respectively mean value of the hydrostatic stress sH,m sH,max maximum value of the hydrostatic stress maximum normal stress on the critical plane sn,cr mean stress on the critical plane sn,m mean stress normal to the critical plane sm n maximum normal stress on the critical plane smax n ultimate tensile strength su yield stress sy shear stress amplitude on the critical plane ta,cr t f, g f torsional fatigue strength coefcient and axial fatigue ductility coefcient, respectively q angle of the cycle path orientation with respect to the principle axis coefcient of nonproportionality j phase difference between applied torsion and bending

sa,eq

Sb 1

f 1K 2t 1

x 1/x

(3)

where x = 2(1 + bsinj), K = 2St / Sb. Carpinteri and Spagnoli [9] used the maximum normal stress and the shear stress amplitude on the critical plane as parameters to modify Goughs quadrant equation. Sines [10] proposed a popular high-cycle fatigue criterion J2,a ksH,m l (4)

Kakuno and Kawada [11] suggested separating the effects of the amplitude and mean value of the hydrostatic stress. Crossland [12] proposed that the inuence of the hydrostatic stress must appear in the fatigue formula by its maximum value. Findley [13], Matake [14] and McDiarmid [15] used the shear stress amplitude and the maximum value of the normal stress on the critical plane as parameters and respectively proposed a similar fatigue criterion as follows (5) ta,cr ksn,cr l

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

19

Findley [13] determined the critical plane by maximizing a linear combination of shear stress amplitude and maximum value of the normal stress. However, Matake [14] and McDiarmid [15] proposed that the critical plane is the plane on which the amplitude of the shear stress attains its maximum. McDiarmid [15] considered difference of crack propagation and dened k and l as follows k tA,B ,l 2su tA,B (6)

Papadopoulos [16] proposed a fatigue limit criterion which could be applied in the case of constant amplitude multiaxial proportional and nonproportional loading in the eld of high-cycle fatigue. The fatigue limit criterion were written as ta,cr ksH,max l (7)

2.2. Critique on stress criteria In the present section, four stress criteria, namely the criteria of Gough, Lee, Sines and McDiarmid are selected for application against some experimental data. The experimental data related to in-phase and 90out-ofphase loading are retrieved from the following publications: Papadopoulos [4] and Carpinteri and Spagnoli [9]. Correlations of experimental data with predicted parameter are shown in Figs. 14. Involved material parameters are reported in Table 1. Fig. 1 illustrates that the Goughs criterion does not t in with the test data under nonproportional loading. The prediction of this criterion under proportional load-

Fig. 2.

The Lees criterion.

Fig. 3.

The Sines criterion.

Fig. 1.

The Goughs criterion.

ing showed a deviation less than 10%, but the prediction of this criterion under nonproportional loading showed a deviation greater than 10%. From Fig. 2, one can observe that the Lees criterion has a good life prediction capability under both proportional loading and nonproportional loading. The prediction of the Sines criterion under proportional loading showed a deviation less than 10%, see Fig. 3. The errors between experi-

20

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

from the multiaxial fatigue criteria, and x equal f 1 or t 1 depending on different criterion. The average absolute errors of the four stress criteria are presented in Table 2. Considering all sides, the criterion of Lee is in the best agreement with the test data. Its average absolute error is less than 5% for all the materials and all kind of loading conditions employed. However, in the Lees criterion there are material constants, which can be determined by a lot of experimental data, so the application of the Lees criterion is limited. The Goughs criterion is valid for a lot of materials under proportional loading and its calculation is relatively simple.

3. Strain criteria 3.1. Review of strain criteria Brown and Miller [17] proposed a theory based on a physical interpretation of mechanisms of fatigue crack growth. Kandil, Brown and Miller (KBM) [18] then proposed a specic form for it gmax
Table 1 Material properties Material f 1 (MPa) 313.9 235.4 398 t 1 (MPa) 196.2 137.3 260 f 1/t
1

Fig. 4.

The McDiarmids criterion.

Sen

(10)

Lohr and Ellison [19] dened the critical plane as a plane p/4 to surface. The proposed fatigue parameter is g
su (MPa) 704.1 518.8 1025

ke n

(11)

Socie et al. [20] modied the formulation of Eq. (10) by including the effect of mean stress as follows gmax en sn,m E C (12)

Hard steel [9] Mild steel [9] 42CrMo4 [4]

1.6 1.71 1.53

mental data and the predictions of the Sines criterion under nonproportional loading are bigger. The deviation degrees are different depending on the material, namely the error of the mild steel is the smallest, the error of the hard steel is bigger and the error of the 42CrMo4 is the biggest. Fig. 4 shows that the McDiarmids criterion has a good correlation within 5% for hard steel and mild steel, but has a nonconservative correlation for 42CrMo4. This coincides with the result in the paper [15]. The error index I is dened to measure the relative difference between the equivalent stress (sa,eq) and the fatigue limits (f 1 or t 1), that is I I sa,eq x (%) x 1 ni
n

Fatemi and Socie [21] changed the normal strain term in Eq. (10) to a normal stress and gave the following criterion gmax 1 smax n k sy C (13)

Table 2 The average absolute error of four stress criteria Criterion The average absolute error I (%) Proportional loading Hard steel Gough Lee Sines McDiarmid 2.4 2.4 2.7 2.6 Mild steel 4.0 4.0 3.5 3.2 Nonproportional loading Mild steel 10.9 5.0 4.8 5.3 42CrMo4

(8) (9)

42CrMo4 Hard steel 9.0 9.0 8.5 8.7 4.7 1.5 12.5 9.4

|Ii|
1

where sa,eq is the amplitude of the equivalent stress got

10.7 4.2 27.8 30.2

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

21

Additional cyclic hardening developed during out-ofphase loading is included in the normal stress term in the above parameter. Shang and Wang (SW) [22] proposed a path-independent parameter, which is based on the critical plane approach. This parameter is made with the maximum shear strain range gmax and the normal strain excursion e n between adjacent turning points. It is given as eeq / 2 e
2 n

1 3

gmax 2

2 1/2

(14)

Under proportional loading the parameter agrees with Von Mises effective strain. Borodii and Strizhalo (BS) [23] take into account the inuence of the different cyclic path and proposed the following reduced strain range enp (1 ksinq)(1 a ) e (15)

3.2. Critique on strain criteria The results predicted by existing strain criteria are compared with experimental data found in literature for SAE-1045 steel reported in Kurath et al. [24], 304 stainless steel reported in Socie et al. [25] and 6061 aluminum alloy tested by Itoh et al. [26]. The corresponding strain paths are in-phase and 90 out-of-phase straining for three materials. Results are shown in Figs. 57. Table 3 lists the material properties used. Correlation for three strain criteria is within a factor of three in life except for the data under out-of-phase

Fig. 6. Predicted versus experimental lives for 304 stainless steel: strain criteria.

Fig. 7. Predicted versus experimental lives for 6061Al alloy: strain criteria.

loading. The error index E is dened to measure the deviation between predicted lives and experimental lives, that is
Fig. 5. Predicted versus experimental lives for 1045HR steel: strain criteria.

log(NP / NE)

(16)

22

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

E
0.445 0.353

1 ni

|Ei|
1

(17)

The mean absolute errors of the three strain criteria analyzed are presented in Table 4. It is shown that the KBMs criterion has a universal good correlation over low to high lives for 1045HR steel, 304 stainless steel and 6061Al alloy. The SWs criterion has a good correlation within a factor of 3 except for 1045HR steel. On the contrary, the BSs criterion has a good correlation for 1045HR steel.

0.097 0.121

0.413 0.413

4. Energy criteria 4.1. Review of energy criteria

Torsional cyclic properties

t f (MPa)

505 709

Smith et al. (SWT) [27] proposed an experimental damage parameter as follows W smax n emax 2 (18)

G (GPa)

79 82.8

0.260 0.171 0.225

For tensile mode failures, Socie [28] modied the SWT parameter by considering that the parameters, which control damage, were the maximum principal strain range and the maximum principal stress on the maximum principal strain plane. Chen et al. [29] proposed different damage parameters for the tensile mode failures and the shear mode failures. For the tensile mode failures, Socies model were modied by including the shear components for the tensile mode failures W emax s1 1 g1 t1 (19)

0.445 0.402 0.629

For the shear mode failures, the following criterion were given W gmax t en sn (20)

(MPa)

0.092 0.114 0.089

Axial cyclic properties

Fatigue properties are given by referenced papers. Fatigue properties are calculated from test data.

There are other developments of the SWT approach proposed by Chu et al. [30], Liu [31] and others. Glinka et al. [32] proposed a fatigue parameter by using the sum of elastic and plastic energy densities in the critical plane to be

E (GPa)

202 185 80

948 1000 528

Table 4 The average absolute errors of three strain criteria Material The average absolute error E (%) Proportional loading KBM 1045HR steel 20 304 Stainless 31 6061Al alloy 8 SW 23 29 8 BS 23 38 8 Nonproportional loading KBM 42 16 18 SW 72 32 7 BS 29 64 45

Table 3 Material properties

1045HRa 304a 6061Alb

Material

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

23

g12 s12 2 2

e22 s22 2 2

(21)

By considering the mean stress effect, Glinka et al. (GWP) [33] modied Eq. (21) in the following form W g12 s12 1 max 2 2 1 s12 / t 1
f

1 s

max 22

/s

(22)
f

Farahani (F) [34] proposed a new parameter that does not use an empirical tting factor. The proposed parameter is the sum of the normal energy range and the shear energy range calculated for the critical plane W 1 ( sn en) s fe f (1 s m / s f) n ( tmax gmax) t fg f (23)

Pan Wen-Fang et al. (PHC) [35] proposed that the inuence of strain energy on the fatigue life in shear direction should be different from its inuence in normal direction. They modied the fatigue strain energy density parameter proposed by Glinka et al. [32] in Eq. (21) W g21 t21 2 2 k1k2 e22 s22 2 2 (24)

where k1 = g f / e f and k2 = s f / t f are two weight constants for strain and stress amplitudes, respectively. 4.2. Critique on energy criteria The test data related to in-phase and 90 out-of-phase sinusoidal loading in [24,25] are used to evaluate the criteria of Glinka et al. (Eq. (21)), Farahani and Pan Wen-Fang et al. (Pan). Predicted lives versus experimental lives are shown in Fig. 8 and the average absolute errors of the three energy criteria analyzed are presented in Table 5. The Fs criterion yields the most satisfactory result. Its correlation for 1045HR steel and 304 stainless steel are within factors of 2 and 3 in life, respectively. The PHCs criterion takes second place. The GWPs criterion has a good correlation for 1045HR steel. In addition, Fig. 8(a) shows that scatter of data is noticeably greater at longer lives.

Fig. 8. Predicted versus experimental lives: the criteria of Glinka, Farahani and Pan.

5. Conclusion 1. Among the stress criteria, the Lees criterion is most promising. Its average absolute error is less than 5% for the materials and loading conditions employed. However, in the Lees criterion there are material constants, which can be determined by a lot of experi-

mental data, so the application of the Lees criterion is limited. The Goughs criterion is relatively simple to use but it can only be used under proportional loading. 2. Among the strain criteria, the KBMs criterion has a good correlation over low to high lives for 1045HR steel, 304 stainless steel and 6061Al alloy.

24

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

Table 5 The average absolute errors of three energy criteria Material The average absolute error E (%) Proportional loading GWP 1045HR steel 12 304 Stainless 54 F 14 29 PHC 23 39 Nonproportional loading GWP 28 39 F 28 24 PHC 54 41

3. The Fs criterion leads to the most satisfactory result among the energy criteria. Its fatigue life correlation for 1045HR steel and 304 stainless steel fell within factors of 2 and 3, respectively. However, the scatter of data estimated by using the energy criteria is noticeably greater at longer lives. 4. Critique and comparison on multiaxial fatigue criteria rely on multiaxial fatigue experimental test, but due to the complexity and expense, especially the tests of high-cycle fatigue are very few. It is difcult to do a comprehensive and systemic comparison on the existing multiaxial fatigue criteria. Therefore, enhancing the study on the multiaxial fatigue test is important to multiaxial fatigue research.

Acknowledgements The authors gratefully acknowledge the nancial support of National Doctoral Foundation of China under grant 20020287022.

References
[1] Garud YS. Multiaxial fatigue: a survey of the state-of-the-art. J Test Evaluat 1981;9(3):16578. [2] Brown MW, Miller KJ. Two decades of progress in the assessment of multiaxial low-cycle fatigue life. In: Amzallag C, Leis B, Rabbe P, editors. Low-cycle fatigue and life prediction, ASTM STP 770. Philadelphia: ASTM; 1982. p. 48299. [3] You BR, Lee SB. A critical review on multiaxial fatigue assessments of metals. Int J Fatigue 1996;18(4):23544. [4] Papadopoulos IV, Davoli P, Gorla C, Fillippini M, Bernasconi A. A comparative study of multiaxial high-cycle fatigue criteria for metals. Int J Fatigue 1997;19(3):21935. [5] Macha E, Sonsino CM. Energy criteria of multiaxial fatigue failure. Fatigue Fract Eng Mater Struct 1999;22:105370. [6] Gough HJ, Pollard HV. The strength of metals under combined alternating stress. Proc Inst Mech Engrs 1935;131:318. [7] Gough HJ, Pollard HV. Properties of some materials for cast crankshafts, with special reference to combined alternating stresses. Proc Inst Automobile Engrs 1937;31:82193. [8] Lee SB. A criterion for fully reversed out-of-phase torsion and bending. In: Miller KJ, Brown MW, editors. Multiaxial fatigue, ASTM STP 853. Philadelphia: ASTM; 1985. p. 55368.

[9] Carpinteri A, Spagnoli A. Multiaxial high-cycle fatigue criterion for hard metals. Int J Fatigue 2001;23:13545. [10] Sines G. Behaviour of metals under complex stresses. In: Sines G, Waisman JL, editors. Metal fatigue. New York: McGraw-Hill; 1959. p. 14569. [11] Kakuno H, Kawada Y. A new criterion of fatigue strength of a round bar subjected to combined static and repeated bending and tortion. Fatigue Eng Mater Struct 1979;2:229. [12] Crossland B. Effect of large hydrostatic pressures on the torsional fatigue strength of an alloy steel. In: Proceedings of the International Conference on Fatigue of Metals, Institution of Mechanical Engineers, London. 1956. p. 13849. [13] Findley WN. A theory for the effect of mean stress on fatigue of metals under combined torsion and axial load or bending. J Eng Ind, Trans ASME 1959;81:3016. [14] Matake T. An explanation on fatigue limit under combined stress. Bull JSME 1977;20:25763. [15] McDiarmid DL. A general criterion for high cycle multiaxial fatigue failure. Fatigue Fract Eng Mater Struct 1991;14:42953. [16] Papadopoulos IV. Long life fatigue under multiaxial loading. Int J Fatigue 2001;23:83949. [17] Brown MW, Miller KJ. A theory for fatigue failure under multiaxial stressstrain conditions. Proc Inst Mech Engrs 1973;187:74555. [18] Kandil FA, Brown MW, Miller KJ. Biaxial low cycle fatigue fracture of 316 stainless steel at elevated temperatures. The Metal Society, London 1982;280:20310. [19] Lohr RD, Ellison EG. A simple theory for low cycle multiaxial fatigue. Fatigue Fract Eng Mater Struct 1980;3:117. [20] Socie DF, Waill LA, Dittmer DF. Biaxial fatigue of inconel 718 including mean stress effects. In: Miller KJ, Brown MW, editors. Multiaxial fatigue, ASTM STP 853. Philadelphia: ASTM; 1985. p. 736. [21] Fatemi A, Socie DF. A critical plane to multiaxial fatigue damage including out-of-phase loading. Fatigue Fract Eng Mater Struct 1988;11(3):14965. [22] Shang DG, Wang DJ. A new multiaxial fatigue damage model based on the critical plane approach. Int J Fatigue 1998;20(3):2415. [23] Borodii MV, Strizhalo VA. Analysis of the experimental data on a low cycle fatigue under nonproportional straining. Int J Fatigue 2000;22:27582. [24] Kurath P, Downing SD, Galliart D. Summary of non-hardened notched shaft round robin program. In: Leese GE, Socie DF, editors. Multiaxial fatigue: analysis and experiments, SAE, AE-14. Warrendale: SAE; 1989. p. 1232. [25] Socie D. Multiaxial fatigue damage models. J Eng Mater Tech 1987;109:2938. [26] Itoh T, Nakata T, Sakane M, Ohnami M. Nonproportional low cycle fatigue of 6061 aluminum alloy under 14 strain paths. In: Macha E, Mroz Z, editors. Proceedings of the Fifth International Conference on Biaxial/Multiaxial Fatigue and Fracture, TU Opole, Cracow, vol. I, 812 September 1997. p. 17387. [27] Smith KN, Watson P, Topper TH. A stressstrain function for the fatigue of metals. J Mater 1970;5(4):76778. [28] Socie DF. Multiaxial fatigue damage models. J Eng Mater Tech 1987;109:2938. [29] Chen X, Xu S, Huang D. A critical plane-strain energy density criterion for multiaxial low-cycle fatigue life under non-proportional loading. Fatigue Fract Eng Mater Struct 1999;22:679 86. [30] Chu CC, Conle FA, Bonnen JJ. Multiaxial stressstrain modeling and fatigue life prediction of SAE axle shafts. In: McDowell DL, Ellis R, editors. Advances in multiaxial fatigue, ASTM STP 1191. Philadelphia: ASTM; 1993. p. 3754. [31] Liu KC. A method based on virtual strain-energy parameters for multiaxial fatigue life prediction. In: McDowell DL, Ellis R, edi-

Y.-Y. Wang, W.-X. Yao / International Journal of Fatigue 26 (2004) 1725

25

tors. Advances in multiaxial fatigue, ASTM STP 1191. Philadelphia: ASTM; 1993. p. 3754. [32] Glinka G, Plumtree A, Shen G. A multiaxial fatigue strain energy parameter related to the critical plane. Fatigue Fract Eng Mater Struct 1995;18(1):3746. [33] Glinka G, Wang G, Plumtree A. Mean stress effects in multiaxial fatigue. Fatigue Fract Eng Mater Struct 1995;18:75564.

[34] Farahani AV. A new energy critical plane parameter for fatigue life assessment of various metallic materials subjected to in-phase and out-of-phase multiaxial fatigue loading conditions. Int J Fatigue 2000;22:295305. [35] Pan WF, Hung CY, Chen LL. Fatigue life estimation under multiaxial loadings. Int J Fatigue 1999;21:310.

You might also like