You are on page 1of 6

International Journal of Pressure Vessels and Piping 76 (1999) 741746 www.elsevier.

com/locate/ijpvp

An evaluation of multiaxial fatigue life assessment methods for engineering components


J. Das a,*, S.M. Sivakumar b
b

Steam Turbines Engineering, Bharat Heavy Electricals Limited, Ramachandrapuram, Hyderabad 500 032, India Solid Mechanics Division, Department of Applied Mechanics, Indian Institute of Technology, Chennai 600 036, India Received 21 October 1998; accepted 1 April 1999

Abstract This article deals with the problem of multiaxial fatigue life assessment of engineering components. General aspects of the critical plane damage models are discussed rst. A computer-based procedure for multiaxial fatigue life assessment incorporating critical plane damage models, suitable for use in design evaluations of engineering components based on nite element analysis results, is presented and applied to correlate results from tests conducted on SAE 1045 steel notched shaft specimens. From experimental data correlations, it is found that the second variant of the critical plane models (in which critical planes are dened as the planes experiencing the maximum damage) provides better life estimates than the rst variant (in which critical planes are dened as the planes experiencing the maximum range of strain) or the conventional local strain based life prediction method, even for simple and short loading histories made up of repeating cycles. 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Multiaxial fatigue life assessment; Critical plane damage models; Design evaluations

Nomenclature b c ds dt dscycle dtcycle dsblock dtblock Dsmax Dtmax E G K KH M M0 n nH n x , ny , nz Nf fatigue strength exponent fatigue ductility exponent shear damage parameter tensile damage parameter damage per cycle for shear model damage per cycle for tensile model damage per block for shear model damage per block for tensile model maximum value of dsblock maximum value of dtblock modulus of elasticity modulus of rigidity empirical material constant cyclic strength coefcient bending moment at the notch root bending moment at the right face of the FE model unit normal to the candidate plane cyclic hardening exponent direction cosines of the unit normal n to the candidate plane cycles to failure

Nb T 1 1R 1 ac 1 eqpl 1 Hf g g ac g Hf s s eq sn s ys s maxs

s maxt s Hf t t Hf u, uR
Subscripts s t

blocks to failure applied torque normal strain on the candidate plane resultant strain on the candidate plane largest normal strain amplitude for any plane equivalent plastic strain fatigue ductility coefcient shear strain on the candidate plane largest shear strain amplitude for any plane torsional ductility coefcient normal stress on the candidate plane von Mises equivalent stress maximum normal stress for the current cycle yield stress maximum normal stress on the same plane as g ac maximum normal stress on the same plane as e ac tensile fatigue strength coefcient shear stress shear fatigue strength coefcient orientation angles for the unit normal n to the candidate plane shear tensile

* Corresponding author.

0308-0161/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved. PII: S0308-016 1(99)00053-8

742

J. Das, S.M. Sivakumar / International Journal of Pressure Vessels and Piping 76 (1999) 741746

x, y, z Cartesian co-ordinates a, min, max amplitude, minimum, and maximum values for the current cycle of strain on the candidate plane

The following shear based damage model proposed by Fatemi and Socie [5] has been found to be superior for materials whose damage development was shear dominated:

gac 1 K smaxs =sys


1. Introduction Many engineering components and structures are subjected to multiaxial stresses and strains under service conditions. Most of the engineering components contain stress concentration features or notches. For such notched components, even when the remotely applied load is uniaxial, the local stressstrain state developed at the critical locations or notches (where fatigue failures are likely to occur) is usually multiaxial because of the constraint effect at the notch. In recent years, a number of investigations have shown that the local state of stress or strain does indeed inuence the fatigue strength of a component and uniaxial-based life predictions are very nonconservative [1]. Various multiaxial fatigue life prediction methods based on effective strainstress, plastic workenergy, critical plane approaches, etc., have been proposed, but as yet there is no universally accepted approach. In general, better correlations of multiaxial fatigue data have been obtained using critical plane approaches [2,3]. Crack initiation size plays an important role in the determination of fatigue life of a component. For a specic crack initiation size that is assumed to indicate failure, the material constants are found out from a set of experimental data in which same size is assumed for failure. This reduces the discrepancy in the correlation of model predictions and experimental observations. Notch geometry plays a role in xing the crack initiation criterion though it is not taken into account in this work. Critical plane approaches are based upon the physical observation that fatigue cracks initiate and grow on certain material planes, the orientation of which is determined by both the stresses and strains at the critical location. These planes on which cracks initiate and grow, and hence on which the dominant fatigue damage should occur, are called critical planes. The critical plane approaches postulate that normal stresses or strains on these planes assist in the fatigue damage process. Depending upon strain amplitude, material type, and state of stress, materials generally form one of the two types of cracks - shear cracks or tensile cracks. The following tensile based damage model proposed by Smith, Watson and Topper [4] has been found to be superior in correlating fatigue lives for materials whose damage development was tensile dominated:

t Hf 2Nf b g Hf 2Nf c G

The critical plane for this model is identied as the plane experiencing the maximum shear strain amplitude and the fatigue life is estimated based on the accumulated damage on this plane. For complex variable amplitude loading histories, small modications to the denition of damage parameters have been suggested by Bannantine and Socie [1,6]. The modied tensile and shear models are referred to as the second variant of the critical plane models in this article. In the second variant of tensile and shear models the critical plane is identied as the plane experiencing the maximum damage, and not maximum normal or shear strain amplitude as in the rst variant. Thus in the second variant of the tensile model, the critical plane is identied as the plane for which the damage parameter given by the product 1a sn is maximum, where 1 a is the normal strain amplitude and s n is the normal stress for the current cycle of strain on a candidate plane. Similarly, in the shear model, the critical plane is the plane for which the damage parameter ga 1 K sn =sys is maximum, where g a is the shear strain amplitude and s n is the normal stress for the current cycle of strain on a candidate plane. The material constant K accounts for the extra cyclic hardening in out-of-phase loading and is taken as unity in this work. 2. Implementation procedure The procedure given below is suitable for multiaxial fatigue life estimation of an engineering component subjected to variable amplitude loading. Both tensile and shear failure modes are considered in the procedure reecting the two possible modes of cracking. The entire procedure has been implemented in a computer code, which allows the procedure to be applied quickly. The stepwise procedure incorporating the second variant of critical plane models is outlined below: 1. Carry out elastic FEA for the component with coarse FE mesh under the given loadings to determine the critical locations, i.e. locations of high peak stresses. The level of stresses above the yield stress indicates the presence of plastic strains at these locations. 2. Rene the FE mesh at the critical locations and carry out elasticplastic FEA under the given loadings. Rene the mesh at the critical locations progressively until the results seem to be converging. 3. Obtain time histories of the strain tensor (sum of elastic and plastic components) and the stress tensor at the critical location under consideration from the results of FEA.

1ac smaxt

s Hf 2 2Nf 2b 1 Hf s Hf 2Nf bc E

The critical plane is identied as the plane experiencing the maximum normal strain amplitude and the fatigue life is estimated based on the accumulated damage on this plane.

J. Das, S.M. Sivakumar / International Journal of Pressure Vessels and Piping 76 (1999) 741746

743

1Rz 1z nz 0:5gxz nx gyz ny

10

7. Solve for Step 6 repeatedly for all the time points in the current cycle (for i 1 to ntime) to develop time histories of 1 , g and s on the candidate plane for the current cycle of strain. (Each cycle is divided into ntime 1 subdivisions and 1 , g , and s are calculated for each of the time points leading to the development of the time histories of 1 , g and s .) 8. Determine normal strain amplitude 1 a, shear strain amplitude g a, and maximum normal stress s n for the current cycle of strain on the candidate plane from the 1 , g , and s time histories obtained at Step 7, where:

1a 1max 1min =2

ga gmax gmin =2

11

9. Calculate tensile and shear damage parameters as follows: d t 1a sn


Fig. 1. Orientation of the unit normal n (to a candidate plane) dened with angles u R and u .

ds ga 1 sn =sys

12

4. Consider a candidate plane (say, kth candidate plane) through the critical location with the orientation of the unit normal n to the plane dened by angles u R, u (Fig. 1). Let the starting values be uR 0 ; u 0 : 5. Calculate direction cosines of the normal n to the candidate plane as: nx sin u sin uR nz cos u: ny sin u cos uR 3

10. Solve for fatigue life Nft as per tensile model for the current cycle of strain from the following equation using corresponding material properties (obtained from uniaxial fatigue test data): dt s Hf 2 2Nft 2b 1 Hf s Hf 2Nft bc E 13

11. Solve for Nfs as per shear model for the current cycle of strain from the following equation: ds

6. Calculate normal strain 1 , shear strain g , and normal stress s on the candidate plane at time point i, using time histories of stress and strain tensors as follows:
2 2 1 1x n2 x 1y ny 1z nz gxy nx ny gyz ny nz gxz nx nz 4 2 1=2 g 2 12 R1 2 2 s sx n 2 x sy ny sz nz 2txy nx ny 2tyz ny nz

t Hf 2Nfs b g Hf 2Nfs c G

14

If the torsional fatigue constants are not known, they are estimated from uniaxial fatigue constants as follows [2]:

g Hf 1:661 Hf t Hf =G 1:44s Hf =E

15

2txz nx nz where:
2 2 1=2 1R 12 Rx 1Ry 1Rz

7 8 9

1Rx 1x nx 0:5gxy ny gxz nz 1Ry 1y ny 0:5gxy nx gyz nz

The NewtonRaphson (NR) method is used for solution of Eqs. (13) and (14). The initial guess for the NR method is obtained as follows: (a) calculate Nf with the rst term on the right hand side discarded; (b) calculate Nf with the second term on the right hand side discarded; (c) take the larger value from (a) or (b) as the initial guess. In cases where the NR method does not converge, a trial-and-error approach is used in the computer code. 12. Determine fatigue damage associated with the current cycle of strain on the kth candidate plane as per tensile

744

J. Das, S.M. Sivakumar / International Journal of Pressure Vessels and Piping 76 (1999) 741746

Fig. 2. Geometry, portion used for FE modelling, and loading condition of the SAE notched shaft specimen. Dimensions in millimetres.

and shear damage models: dtcycle 1=Nft dscycle 1=Nfs 16

13. Determine fatigue damage associated with a loading block by summing damage associated with each cycle in the block (for i 1 to n) using Miners rule as follows: dtblock
n i1

15. Identify the critical plane as the plane associated with the maximum damage Dtmax as per tensile model and Dsmax as per shear model. 16. Determine fatigue life of the component in blocks to failure based on the damage on the critical plane as follows: Nbt 1=Dtmax Nbs 1=Dsmax 18

1=Nfti

dsblock

n i1

1=Nfsi

17

17. Take the smaller of the two values, Nbt and Nbs, as the nal fatigue life estimate of the component. The procedure incorporating the rst variant of critical plane models is similar to the one outlined above except for differences in the denitions of damage parameters and critical planes as mentioned earlier. 3. Evaluation of results The foregoing procedures were applied to correlate results from tests conducted on normalized SAE 1045 steel notched shaft specimens (taken from Ref. [7]), subjected to bending and in-phase bending and torsion loads. Geometry and loading condition of the notched shaft are shown in Fig. 2. The notch radius is 5 mm and

where n is the number of cycles in the loading block and Nfti and Nfsi refer to the number of cycles to failure for each cycle i in the block for the tensile and shear models, respectively. 14. Vary u R from 0 to 180 in user-specied increments (say, 2) and for every u R vary u from 0 to 180 in user-specied increments (say, 2) in order to locate possible orientations of the normal vector n to the candidate plane and for each orientation of n, repeat Steps 513 to calculate dtblock and dsblock for various orientations of the candidate plane. Let Dtmax and Dsmax be the maximum values of dtblock and dsblock, respectively.

J. Das, S.M. Sivakumar / International Journal of Pressure Vessels and Piping 76 (1999) 741746 Table 1 Notched shaft fatigue test results under constant amplitude loading [7] Load case 1 2 3 4 Loading Pure Bending M, Nm 1475 1730 2600 990 T, Nm 0 0 0 1390 Number of cycles to crack initiation 354 500 49 200 5676 716 382 201 800 84 467 3000 350 000

745

60 000 3918

Combined bending and torsion, T =M 1:4

5 6

1220 1850

1710 2550

72 000 2045

71 142

60 800

Table 2 Predicted versus experimental fatigue lives (number of cycles to crack initiation) Load case Nexperimental, cycles Npredicted, cycles Critical-plane approach variant-1 1 2 3 4 5 6 354 500 49 200 5676 716 382 72 000 2045 201 800 84 467 3000 350 000 71 142 1 074 000 249 151 11785 272 225 15 120 920 Critical-plane approach variant2 661 827 161 837 8118 279 388 16 560 975 Local strain approach

60 000 3918 60 800

1 500 000 250 000 21 000 1 800 000 370 000 5500

Table 3 Ratio of predicted and experimental fatigue lives Load case Nexperimental, cycles Npredicted/Nexperimental Critical-plane approach variant1 1 2 3 4 5 6 354 500 49 200 5676 716 382 72 000 2045 201 800 84 467 3000 350 000 71 142 3.03, 5.32 5.06, 2.95, 4.15 2.08, 3.93, 3.01 0.38, 0.78 0.21, 0.212, 0.25 0.45 Critical-plane approach variant2 1.86, 3.27 3.28, 1.91, 2.69 1.43, 2.7, 2.07 0.39, 0.80 0.23, 0.233, 0.27 0.47 Local strain approach

60 000 3918 60 800

4.23, 7.43 5.08, 2.96, 4.16 3.70, 7.0, 5.36 2.51, 5.14 5.13, 5.2, 6.08 2.68

the net section diameter is 40 mm. Bending loading is achieved by a vertical force at a distance of 150 mm from the notch root. This vertical load causes a shear force V and bending moment Mo as shown in Fig. 2. The applied loads and resulting number of cycles to crack initiation (taken from Ref. [7]) are given in Table 1. Multiple results indicate test replications. Finite element analyses were carried out for the notched shaft under constant amplitude cyclic bending and in-phase bending and torsion loadings using ansys nite element code [8]. Kinematic hardening with von Mises yield criterion and associative ow rule was used for elasticplastic FEA. The load data may be found in Table 1. Material properties are (taken from Ref. [9]): s Hf 948:2 MPa; 1 Hf

0:239; b 0:092; c 0:435, E 202375 MPa: The fatigue constants and the E value from Deere and Company [9] were used. The cyclic stressstrain curve with K H 1283 MPa and n H 0:211 (taken from Ref. [10]) was used. The FE mesh for combined bending and torsion loading is shown in Fig. 3 and it contains 6289 nodes and 1515 elements. Isoparametric solid elements with 20 nodes were used. For bending loading, only half model was used with symmetric boundary conditions. Fatigue life estimates were obtained from the two versions of computer code incorporating the two variants of critical plane approach. In addition, life estimate was obtained from the conventional local strain based life prediction method, usually followed in the industry, using

746

J. Das, S.M. Sivakumar / International Journal of Pressure Vessels and Piping 76 (1999) 741746

Fig. 3. Finite element mesh for the SAE notched shaft specimen.

notch root effective strain amplitude, given by seq =E 1eqpl ; and low-cycle fatigue curve (Fig. 4) obtained from uniaxial fatigue tests [9]. The low-cycle fatigue curve from Deere and Company [9] was used. Tables 2 and 3 present a summary of the results of the comparisons between the experimental and predicted fatigue lives. As can be seen from Table 3, the second variant of critical plane approach is clearly the best out of the three approaches considered in correlating experimental data. Correlations for pure bending loadings are within a factor of 2 for the best correlations in life estimates. For combined bending and torsion loadings, with more complex stress strain states, the results fall within a factor of 5 for the best correlations. 4. Conclusions 1. From experimental data correlations, it is found that the second variant of the critical plane models (in which critical planes are dened as the planes experiencing the maximum damage) provides better life estimates than the rst variant (in which critical planes are dened as the planes experiencing the maximum range of strain) or the conventional local strain based life prediction method, even for simple and short loading histories made up of repeating cycles. 2. The results of this work emphasize the need to account for the local multiaxial stress state in the fatigue life predictions using appropriate multiaxial damage models. 3. FE analysis procedure is used in this work for the entire loading history which may prove costly and time consuming. Less complex analysis procedures which give reasonable good analysis results with a lot of cost and time saving is the need of the day. But because of the robustness of the FE analysis procedures, it is used in this work for general setting. However, for some specic

Fig. 4. Low-cycle fatigue curves for SAE 1045 steel ([9]).

problems and situations, it may be possible to use simpler analysis techniques, if available.

References
[1] Bannantine JA, Socie DF. A multiaxial fatigue life estimation technique. In: Mitchell MR, Landgraf RW, editors. Advances in fatigue lifetime predictive techniques, ASTM STP 1122, Philadelphia: American Society for Testing and Materials, 1992. pp. 249275. [2] Socie D. A summary and interpretation of the Society of Automotive Engineers biaxial testing program, multiaxial fatigue: analysis and experiments, SAE AE-14, 1989, Society of Automotive Engineers, Warrendale, PA, pp. 111. [3] You BR, Lee SB. A critical review on multiaxial fatigue assessments of metals. International Journal of Fatigue 1996;18(4):235244. [4] Smith RN, Watson P, Topper TH. A stressstrain function for the fatigue of metals. Journal of Materials 1970;5(4):767778. [5] Fatemi A, Socie DF. A critical plane approach to multiaxial fatigue damage including out-of-phase loading. Fatigue and Fracture of Engineering Materials and Structures 1988;11(3):149165. [6] Socie D. Critical plane approaches for multiaxial fatigue damage assessment. In: McDowell DL, Ellis R, editors. Advances in multiaxial fatigue, ASTM STP 1191, Philadelphia: American Society for Testing and Materials, 1993. pp. 736. [7] Fash JW. An evaluation of damage development during multiaxial fatigue of smooth and notched specimens, Materials engineering report no. 123, University of Illinois at Urbana-Champaign, 1985. [8] ansys theory reference, release 5.3, ANSYS, Inc., Houston, PA, 1996. [9] Tipton SM, Fash JW. Multiaxial fatigue life predictions of the SAE specimen using strain based approaches, multiaxial fatigue: analysis and experiments, SAE AE-14, 1989, Society of automotive engineers, Warrendale, PA, pp. 6780. [10] Hoffman M, Seeger T. Stressstrain analysis and life predictions of a notched shaft under multiaxial loading, multiaxial fatigue: analysis and Experiments, SAE AE-14, 1989, Society of automotive engineers, Warrendale, PA, pp. 8199.

You might also like