Вы находитесь на странице: 1из 9

International Journal of Fatigue 23 (2001) S375S383 www.elsevier.

com/locate/ijfatigue

Residual stress distributions and their inuence on fatigue lifetimes


G.A. Webster
b

a,*

, A.N. Ezeilo

a Department of Mechanical Engineering, Imperial College, London, SW7 2BX, UK MSX International Ltd., Endeavour Drive, Festival Business Park, Basildon, SS14 3WF, UK

Abstract Residual stresses can have a signicant inuence on the fatigue lives of structural engineering components. For the accurate assessment of fatigue lifetimes a detailed knowledge of the residual stress prole is required. The neutron diffraction non-destructive method of measuring these stresses is described. Examples of the detrimental residual stress distributions introduced by welding and of the benecial stresses produced by such processes as autofrettage, cold hole expansion and shot peening are presented. It is demonstrated that, provided an accurate knowledge of the residual stress proles generated is available and allowance is made for stress redistribution and the multiaxial nature of residual stresses, reliable predictions of fatigue performance can be made. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Residual stress; Neutron diffraction; Welding; Manufacturing processes; Fatigue performance

1. Introduction Residual stresses can have a signicant inuence on the fatigue lives of engineering components [112]. In particular, near surface tensile residual stresses tend to accelerate the initiation and growth phases of the fatigue process while compressive residual stresses close to a surface may prolong fatigue life. Residual stresses can be introduced unintentionally into components by, for example, forging, bending and welding operations which may be employed during manufacture or by forces and thermal gradients imposed during use [13]. They can also be generated deliberately by such processes as autofrettage, shot-peening and a variety of surface treatments. For the accurate assessment of fatigue lifetimes a detailed knowledge of the residual stress proles generated is required together with methods for incorporating these into fatigue lifetime calculation procedures. Signicant advances have been made in recent years for obtaining accurate and reliable determinations of residual stress distributions. These include both experimental [13,14] and numerical methods [15,16]. In this paper a brief outline of these is provided. Some typical

residual stress distributions are presented and it will be shown how these inuence resistance to fatigue failure.

2. Determination of residual stress Several experimental methods are available for determining residual stresses. They include X-ray diffraction [17,18], neutron diffraction [1924], surface and deep hole drilling [2527], boring [2830], slicing [31 33], and magnetic methods [34]. Only the neutron diffraction technique is capable of establishing these stresses in the interior of components non-destructively. An attraction of a non-destructive procedure is that a fatigue test can subsequently be carried out on a sample in which the residual stresses have already been measured. Also any redistribution of residual stress throughout life can be estimated by making further repeat measurements [35,36]. With the neutron diffraction method, components of strain are obtained from measurements of changes d in the lattice spacings d of crystals as shown in Fig. 1. When a beam of neutrons of xed wavelength l is incident upon a crystalline sample, a diffraction pattern with sharp maxima is produced. The positions of the peaks are given by the Bragg equation 2d sin q nl (1)

* Corresponding author. Fax: +44-020-7595-7017. E-mail address: g.webster@ic.ac.uk (G.A. Webster).

0142-1123/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 1 4 2 - 1 1 2 3 ( 0 1 ) 0 0 1 3 3 - 5

S376

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

Fig. 1. Principles of the neutron diffraction technique showing Bragg reection from crystal planes d apart.

where n is an integer and 2q the diffraction angle. A small change in the lattice parameter d will result in a change q in the Bragg angle so that the lattice elastic strain in the direction of the scattering vector Q is given by e d d q.cot q (2)
Fig. 2. Peak proles for the (211) reection, measured for the hoop direction in a 5 mm autofrettaged steel sample, showing angular shifts as a function of radius.

An illustration of the peak proles and shifts observed as a function of distance r from the bore of a piece of autofrettaged tubing with wall thickness W is presented in Fig. 2 [20]. Typically in order to obtain a resolution in strain of 10 4 (corresponding to a stress of about 20 MPa in most engineering materials) it is necessary to measure peak shifts to an accuracy of 0.005. In general, to completely dene the stress tensor at a point, measurements of strain in six orientations are required. However, when the principal directions are known three orientations will sufce. When these coincide with the coordinate directions x, y and z, the principal stresses are given by sx E [(1 n)ex n(ey ez)] (1+n)(12n) (3)

cated in tests on tensile bars in Fig. 3 [37]. After yielding, some planes gradually shed stress and others gain stress which results in a non-linear stress strain response and apparent residual stresses (shown by the offsets in strain in the gure) on returning to zero load. It is

with corresponding expressions for sy and sz. In this equation E is the elastic modulus and n Poissons ratio, respectively. The values appropriate to the specic (hkl) crystallographic plane on which the strain is being measured are used [37,38]. Several factors need to be taken into account when making residual stress measurements for engineering purposes by neutron diffraction. An accurate estimate of the unstressed lattice spacing do must be known. This may change with position in inhomogeneous material and particularly adjacent to welds. Not all crystallographic planes respond similarly to applied stress as indi-

Fig. 3. Effect of yielding on neutron diffraction stress measurements on particular crystallographic planes.

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

S377

important for continuum mechanics analyses that measurements are only made on those planes which represent bulk behaviour. To obtain the residual stress pattern across a region, it is necessary to identify a suitably small sampling volume (by neutron beam masks as shown in Fig. 1) and translate this volume across the region in sufciently small steps to achieve adequate resolution. The neutron diffraction technique can also be employed to measure live loads in specimens. This is achieved by loading a specimen in-situ in a diffractometer and making measurements as the load is changed. Stroboscopic methods can be employed to determine stresses at a particular point in a fatigue cycle if desired. In this way measurements can be made in specic (hkl) crystallographic directions [38], to correspond with slip directions, to assist in interpreting the mechanisms responsible for fatigue crack initiation and growth. It should be borne in mind, however, that with neutrons only normal strains are measured. This means that measurements must be made in sufcient orientations for shear stresses to be calculated if needed. Analytical and numerical methods can also be employed to establish residual stress distributions [39 41]. These require accurate models of the processes responsible for introducing the residual stresses and of the material behaviour. For describing material behaviour a yield criterion and work hardening characteristics need to be specied. Typically kinematic or isotropic or a mixed hardening law is assumed [12]. Often, however, as illustrated in Fig. 4 for a high strength steel, the unloading behaviour is quite different to the initial loading behaviour and a Bauschinger effect is apparent where the yield stress in compression is only a fraction a of that in tension. Consequently, in this case, when unloading is involved accurate predictions of residual stress distributions cannot be expected from materials models which are based simply on the initial loading stressstrain curve. This is demonstrated in Fig. 5 for an autofrettaged ring with an inner to outer radius ratio of 2.07 [40]. This gure compares predicted residual hoop stress distributions with those measured experimentally. The predictions were obtained from the initial stress strain loading curve shown in Fig. 4 when assuming nonwork hardening behaviour and the yield criteria indicated. It is clear that quite different predictions are obtained depending on the extent of the Bauschinger effect (value of a) assumed. The greatest limitation in modelling the stressstrain properties of the steel with this approach is that it makes no allowance for the materials work-hardening behaviour during the unloading phase. The shape of the unloading curve shown in Fig. 4 will be expected to be dependent on the stress (or plastic strain) at which unloading takes place [42]. In general, this will need to be taken into account

Fig. 4. Stressstrain properties of tube steel with inset showing unloading behaviour from a plastic strain of 0.45%.

for more accurate predictions of residual stress distributions to be obtained [4345]. During use, materials may cyclically harden or soften with repeated loading which may cause some redistribution of any residual stresses present. This is illustrated in Fig. 6. This gure shows the experimentally measured residual hoop stress distributions produced adjacent to an expanded hole before and after fatigue loading for 105,000 cycles. Also included in the gure is a nite element (FE) prediction of the redistributed residual stress distribution based on the cyclic stressstrain properties of the material. The close correspondence achieved with the experimentally measured residual stresses after fatigue cycling demonstrates that accurate predictions can be achieved when proper allowance is made for material behaviour.

3. Characterisation of residual stress distributions It is evident from the previous section that substantially different residual stress distributions can be calculated to those measured experimentally. Accurate estimates of these residual stress distributions are required

S378

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

Fig. 5. Inuence of the Bauschinger effect on predicted residual hoop stress distributions for autofrettaged tube.

if reliable predictions of fatigue lifetimes and fracture are to be obtained. For conservative assessments, it is necessary that an overestimate of detrimental residual tension and an underestimate of benecial residual compression is made at sites of crack initiation and growth. These stresses must be balanced by stresses of opposite sign elsewhere to satisfy equilibrium conditions. Welds are a major source of detrimental tensile residual stresses [21,27,4651]. They often exist at sites of stress concentration. Fusion welds are also prone to containing regions of poor microstructure. They may also contain defects. They are therefore a frequent source of fatigue failures. An example of a typical residual stress distribution that has been measured in a multi-pass butt weld of an aluminium alloy is presented in Fig. 7 [21]. Typically maximum tensile stresses in the region of the yield stress of a material are obtained. Guidance is provided in some codes (e.g. BS 7910 [52]) on the residual stress distributions to be used in calculations when other estimates are not available. Often, however, these stress distributions do not satisfy equilibrium conditions because of the conservative assumptions made in producing them. This means that the inuence of residual tension may be overestimated and that of residual compression underestimated. In addition these stresses may shake down with cyclic loading.

Fig. 6. Comparison between residual hoop stress distributions around an expanded hole before and after fatigue cycling.

Fig. 7. Comparison of transverse residual stress distributions sy across a multi-pass butt weld in an aluminium alloy.

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

S379

Fig. 8. Residual stress distributions resulting from a circumferential inertia weld in a nickel alloy cylinder as a function of distance from the weld.

For high performance applications other types of welds [3,24,53] are being employed in an attempt to counteract some of the limitations of fusion welds. These include electron beam, linear friction, inertia and friction stir welds. They all exhibit narrow weld metal and heat affected zones and in some instances avoid melting. An example of a residual stress distribution that has been measured in an inertia weld in a nickel alloy cylinder is shown in Fig. 8. This indicates the state of high triaxial residual tension that is developed local to the weld. This state of residual stress is representative of that generated by the other welding processes. It is evident that residual stresses well above yield can be produced. Fig. 9 indicates a similar residual stress pattern that is generated by electron beam welding of a nickel alloy ring. It also shows the substantial reduction in stress that can be achieved by post weld heat-treatment (PWHT), although

about a third of the original peak tensile stress is still retained. Several processes are available for introducing benecial compressive residual stresses adjacent to surfaces to enhance fatigue performance [36,45,5458]. These include autofrettage, hole expansion, shot-peening, laser surface treatment, induction-hardening and carburization. The residual stress distributions produced by autofrettage and hole expansion have already been presented in Figs. 5 and 6. These processes can be optimized to give most resistance to fatigue failure. It has already been established that accurate representation of material behaviour is required to obtain reliable predictions of residual stress distributions. Fig. 6 shows the residual stresses produced at mid plate thickness caused by hole expansion. More detailed [29,41,59] investigations of the hole expansion process have shown that lower compressive residual stress is produced at the mandrel entry than that at the exit and that this is the reason why fatigue cracks usually initiate at the entry side rst. An illustration of the residual stress pattern caused by shot-peening and the improvement in fatigue life that can be gained by increasing the intensity of the peening operation is shown in Figs. 10 and 11, respectively [13,57,60]. A feature of the peening process is that it produces a predominantly biaxial residual stress eld with a near surface stress gradient of up to 2000 MPa/mm. Typically surface compression approaching yield is generated with a subsurface balancing tensile peak of about a third of this value. Again the process can be optimized to give the maximum improvement in fatigue performance although the increase in fatigue performance becomes more pronounced with increasing number of cycles to failure due to less redistribution (relaxation) of the residual stresses with decrease in cycle amplitude. For large amplitudes, the cyclic plastic

Fig. 9. Residual stress distributions produced by electron beam welding of a nickel alloy ring.

Fig. 10. Residual stresses generated by shot peening in a Udimet 720 nickel alloy plate.

S380

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

Their inuence on fatigue crack initiation and growth will now be examined [2,5,10,55,6365]. In order to obtain accurate assessments of fatigue life when residual stresses are present it is necessary to account for residual stress redistribution. For shallow depths of residual stresses such as in the case of shot peening, crack initiation or total life methods are usually used to assess inuence on fatigue life. When residual stress depths are large it is usually more appropriate to use crack growth models to assess the fatigue performance. Most fatigue crack growth data are plotted in terms of crack growth/cycle, da/dN, against stress intensity factor range K which is the difference between the maximum Kmax and minimum Kmin stress intensity factors applied during the load cycle. A representative fatigue crack propagation curve is shown in Fig. 13. Generally it exhibits three regionsA, B and C. Usually a threshold stress intensity factor range KTh is observed below which measurable crack growth is not detected and failure is predicted when Kmax approaches the fracture toughness of the material. Typically region B is relatively insensitive to minimum to maximum load ratio R (=Kmin/Kmax) and can be described by
Fig. 11. Improvement in fatigue strength due to near surface compression induced by shot peening.

da C Km dN

(4)

strains introduced may be sufcient to cause so much stress redistribution as to almost wipe out any effects from the residual stress eld. The effect of residual stress on fatigue performance at elevated temperatures is expected to be less pronounced than at ambient temperature [61,62]. This is due to relaxation of this stress by creep, as indicated in Fig. 12, as well as by fatigue.

where C and m are material parameters. An increase in R has the effect of moving regions A and C to lower K values. The presence of residual stress can alter R and also the stress intensity factor range over which the crack

4. Inuence on fatigue Some examples of the residual stress distributions that can be produced in components have been considered.

Fig. 12. Relaxation of residual stress in a C-shaped steel specimen caused by soaking for up to 1000 hours at 550C.

Fig. 13.

Representative fatigue crack growth curve.

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

S381

remains open. This can be explained by applying the principle of superposition such that R will change to Reff Kmin+Kres Kmax+Kres (5)

where Kres is the stress intensity factor due to the residual stress distribution alone. The value of Kres will be positive for a crack in a tensile residual stress eld leading to Reff R. An effective stress intensity factor range Keff can be dened which is regarded as being responsible for the cracking process. It is supposed that no fatigue damage is incurred when compression is transmitted across the crack faces. In the present approach this is assumed to occur at K=0. Consequently for Kmin+Kres 0 only R will change and Keff= K, since the crack faces will always remain open throughout a cycle. In this instance the stress intensity factor range will remain unaltered by the residual stress. When Kmin+Kres 0, R will change according to Eq. (5) as before but Keff Kmax Kres 0 (6)

to allow for the fact that the crack faces will be closed for part of the cycle. With this approach Eq. (4) becomes modied to da C Km eff dN (7)
Fig. 15. Comparison of fatigue crack growth rate characteristics for as-received and autofrettage ring specimens tested at R=0.01.

for a material with fatigue crack growth properties that are insensitive to R in region B. In this circumstance, compressive residual stress should have most inuence on fatigue crack growth when Kmin |Kres|. This is most likely to occur at low R and low K. An illustration of the use of this procedure in dealing with the effects of compressive residual stress on fatigue crack growth is shown in Figs. 1416. Fig. 14 shows

Fig. 14. Residual stress intensity factor solutions for as-received and autofrettaged ring and C-shaped specimens.

Kres values determined from measurements of residual stress made on ring and C-shaped fracture mechanics specimens taken from as-received and autofrettaged thick walled tubing [39]. Fig. 15 demonstrates the reduction in fatigue crack growth rate produced at a given value of K by the autofrettage process for the propagation of a crack from the bore of a ring specimen containing the residual stress distribution shown in Fig. 5 [55]. The maximum benet gained in reduced cracking rate is obtained close to the bore where the greatest residual compression is measured. When the results of this test and other data [55] are plotted in terms of Keff, as shown in Fig. 16, close correspondence is achieved with as-received data demonstrating the relevance of Eq. (7). For the examples considered in Figs. 1416, signicant residual stress redistribution did not take place with cycling so that Kres could be calculated from the initial residual stress distribution. Allowance for residual stress redistribution must be taken into account in calculating Kres when it occurs. Also for other materials, sensitivity to R ratio may need to be considered and an allowance made for crack closure not taking place at Keff=0 due to crack surface roughness, oxide bridging and plasticity effects [5,66]. Most total life methods rely on establishing modied fatigue limits to account for the presence of residual stresses. Residual stresses are dealt with in the same way

S382

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

nique has been described. It has been shown how it can be applied to measure residual stresses introduced by a selection of manufacturing processes and monitor residual stress re-distribution caused by fatigue loading. It has been applied to determine the undesirable stresses introduced by welding and the benecial residual stresses generated by autofrettage, hole expansion and shot peening to improve fatigue performance. It has been demonstrated that accurate models of fabrication processes and of a materials monotonic and cyclic stressstrain behaviour are required for reliable estimates of residual stress distributions to be made. When these are available, it has been shown that satisfactory predictions of fatigue failure can be obtained provided proper allowance is made for the multiaxial stress state of most residual stress proles and the effect these may have on the R ratio and the proportion of the fatigue cycle over which a crack remains open.

References
[1] Champoux RL, Underwood JH, Kapp JA, editors. Analytical and experimental methods for residual stress effects in fatigue. ASTM STP 1004, 1988. [2] Glinka G. Advances in surface treatments. In: Kiku-Lari A, editor. Pergamon Press, 1986:41354. [3] Webster GA. In: Lieurade HP, editor. Fatigue and stress. Gournay-sur-Marne (France): IITT International, 1989:920. [4] Suresh S. Fatigue of materials. Cambridge University Press, 1991. [5] Newman JC. Eng Mater Tech 1995;117:4339. [6] Almer JD, Cohen JB, Winholtz RA. Metall Mater Trans 1998;29A:212736. [7] Donald JK. Naval Res Rev 1998;50(4):4654. [8] Sadananda K, Vasudevan AK. Fracture mechanics. In: Erdogen F, editor. ASTM STP 1220, vol. 25. American Society for Testing and Materials, Philadelphia, 1995, pp. 484500. [9] Sadananda K, Vasudevan AK. Naval Res Rev 1998;50(4). [10] Moshier MA, Hillberry BM. Fatigue Fract Eng Mater Struct 1999;22(2):51926. [11] Wang H, Buchholz F-G, Richard HA, Jagg S, Scholtes B. Comput Mater Sci 1999;16:10412. [12] Almer JD, Cohen JB, Moran B. Mater Sci Eng 2000;A284:26879. [13] Webster GA. ECRS-5, Holland, Sept 1999. Materials Science Forum, 347349, 19, Trans Tech Publ, Switzerland, 2000. [14] Webster GA, Ezeilo AN. Phys B 1997;234-236:94955. [15] Tauchert TR, Webster GA, Reed RC. In: Crespo da Silva MRM, Mazzilli CEN, editors. Mechanics Pan-America 1993. Appl Mech Rev 1993;46(Part 2):S12S20. [16] Tauchert TR, Webster GA, Ezeilo A. In: Godoy LA, Idelsohn SR, Laura PAA, Mook DT, editors. Applied mechanics in the Americas, vol. III. Santa Fe (Argentina): AAM and AMCA, 1995:1859. [17] Hughes H. Strain 1967;3:2631. [18] Webster PJ, Vaughan GBM, Mills G, Kang WP. In: Fifth Euro Powder Diffraction Conference, Parma (Italy), May 1997, 1997. [19] Allen A, Andreani C, Hutchings MT, Windsor CG. NDT Int 1981;14:24954. [20] Stacey A, MacGillivray HJ, Webster GA, Webster PJ, Ziebeck KRA. J Strain Anal 1985;20:93100.

Fig. 16. Correlation of fatigue crack propagation in as received and autogrettage specimens with effective stress intensity factor range.

as mean stresses [67]. The residual stress is considered as a mean uniaxial stress at the surface with no allowance being made for stress gradients. Extensions [68] to this approach have been proposed to incorporate the cyclic stressstrain behaviour of materials, microstructural changes at surfaces and the inuence of constraint, due to the multi-axial state of most residual stress elds [5,69,70]. The MansonCofn equation can also be modied to account for residual stresses by taking the residual stress sres as a mean stress correction as follows: sf sres e ef(2N)c (2N)b 2 E (8)

is the local total strain range at the fatigue where initiation point, N is the cycles to failure, f , sf , b and c are material constants to represent the fatigue properties of a material and E is the modulus of elasticity. Some success has been achieved with this approach [71,72] for several surface treatment processes.

5. Conclusions Methods of measuring residual stress have been reviewed. The neutron diffraction non-destructive tech-

G.A. Webster, A.N. Ezeilo / International Journal of Fatigue 23 (2001) S375S383

S383

[21] Smith DJ, Leggatt RH, Webster GA, MacGillivray HJ, Webster PJ, Mills G. J Strain Anal 1988;23(4):20111. [22] Bourke MAM, Smith DJ, Webster GA, Webster PJ. In: Ninth International Conference on Experimental Mechanics (Copenhagen) vol. 3. 1990:1198206. [23] Ezeilo AN, Webster GA. Textures and microstructures, vol. 33. Overseas Publishers Association, 1999:15171. [24] Bonner NW, Ezeilo AN. Inst Mater 2000;2:138592. [25] Beaney EM, Proctor E. Strain 1974;10:714. [26] Smith DJ, Bonner NW. PVP, ASME 1996;327:5363. [27] Smith DJ, Bouchard PJ, George D. J Strain Anal IMechE 2000;35(4):287306. [28] Sachs G. Z Metall 1927;19:3527. [29] Ozdemir AT, Edwards L. J Strain Anal 1996;31:7019. [30] Garcia-Granada AA, Smith DJ, Pavier MJ. Int J Mech Sci 2000;42:102747. [31] Cheng W, Finnie I, Gremaud M, Prime MB. ASME J Eng Mater Tech 1994;116(1):107. [32] Prime MB, Gonzales AR. In: Proceedings of the ICRS-6, 1012 July, 2000, Oxford, vol. 1. Institute of Materials, 2000:61724. [33] Nowell D, Tochilin S, Hills DA. J Strain Anal, IMechE 2000;35:27786. [34] Buttle DJ. In: Proceedings of ICRS-6, 1012 July, 2000, Oxford, vol. 1. Institute of Materials, 2000:56788. [35] Bourke MAM, Webster GA, Webster PJ, Wang X. In: Fujiwara H, Abe T, Tanaka K, editors. Residual stresses III Science and Technology ICRS3, vol. 2. 1991:10308. [36] Ezeilo AN, Webster GA, Webster PS, Webster PJ. In: Proceedings of the 4th International Conference on Residual Stress, Baltimore, USA, June 1994. Society of Experimental Mechanics, 1994:127584. [37] Ezeilo AN, Webster GA, Webster PJ, Wang X. Physica B 1992;180-181:10446. [38] Daymond MR, Bourke MAM, Von Dreele RB. J Appl Phys 1997;82(4):155462. [39] Stacey A, Webster GA. In: Champoux RL, Underwood JH, Kapp JA, editors. ASTM Symposium, ASTM STP 1004. American Society for Testing and Materials, 1988:3753. [40] Stacey A, Webster GA. Int J Pres Ves Piping 1988;31:20520. [41] Pavier MJ, Poussard CGC, Smith DJ. J Strain Anal Eng Design 1998;33(4):27589. [42] Smith DJ, Bourke MAM, Hodgson AP, Webster GA, Webster PJ. J Strain Anal 1992;27(2):7783. [43] Parker AP, Endersby SN, Bond TJ, Underwood JH, Lee SL, Higgins J. J Pres Ves Tech, Trans ASME 1997;119(4):48893. [44] Feng H, Mughrabi H, Donth B. Modell Simul Mater Sci Eng 1998;6:5169. [45] Koh S-K, Na E-G. Int J Fatigue 1999;21:13546. [46] Greasley A, Naylor S. Eng Fract Mech 1986;24(5):71726. [47] Glinka G, Dharmavasan S, Dover WD. In: Noordhoek C, de Back J, editors. Steel in marine structures. Amsterdam: Elsevier Science Publishers, 1987:50112.

[48] Nguyen TN, Wahab MA. J Materials Processing Tech 1998;77:2018. [49] Berkovits A, Kelly DW, Di S. Fatigue Fract Eng Mater Struct 1998;21(2):15970. [50] Dye D, Roberts SM, Withers PJ, Reed RC. J Strain Anal, IMechE 2000;35:24760. [51] May PS, ODowd NP, Webster GA. In: Proceedings ICRS-6, 10 12 July, 2000, Oxford, vol. 2. Institute of Materials, 2000:136572. [52] BS 7910. Guide on methods for assessing the acceptability of aws in fusion welded structures. BSI, 1999. [53] Ezeilo AN, Webster GA. In: Proceedings ICRS-6, 1012 July 2000, Oxford, vol. 2. Institute of Materials, 2000:150310. [54] Schlosberg WH, Cohen JB. Metall Trans 1982, 19881995;13A. [55] Stacey A, Webster GA. In: Champoux RL, Underwood JH, Kapp JA, editors. ASTM Symposium, ASTM STP 1004. American Society for Testing and Materials, 1988:10721. [56] Ezeilo AN, Webster GA, Webster PJ, Roth M, Muster WJ. In: Clyne TW, Withers PJ, editors. Proceedings of the 2nd European Conference on Advanced Materials and Processes. Cambridge: Institute of Materials, 1991:38994. [57] Ezeilo AN, Webster GA, Webster PJ, Webster PS. In: Proceedings of the 5th International Conference on Shot Peening, Oxford University, September 1993, 1993:27481. [58] Toparli M, Ozel A, Aksoy T. Mater Sci Eng 1997;A225:196203. [59] Ozdemir AT, Edwards L. Fatigue Fract Eng Mater Struct 1997;20:144351. [60] Shaw GA, Korsunsky AM, Evans JT. ECF 12, Shefeld 1998;1:157. [61] Chorlton M, Nikbin K, Webster GA. In: Edwards JH et al, editors. Lifetime management and evaluation of plant, structures and components. UK: EMAS Publishing, 1998:4757. [62] Sadananda K, Vasudevan AK. Int J Fatigue 1998;19(Suppl. 1):S1839. [63] Louat N, Sadananda M, Duesbery M, Vasudevan AK. Metall Trans A 1993;24A:222532. [64] Vasudevan AK, Sadananda K. Metall Mater Trans A 1995;26A:122134. [65] Sadananda K, Vasudevan AK. Fatigue and fracture mechanics. In: Piascik RS, editor. American Society for Testing and Materials, 1997:30116. [66] Galatola R, Lanciaotti A. Int J Fatigue 1997;1:439. [67] SAE fatigue design handbook, 3rd ed., 1997. [68] Kliman V, Bily M, Prohacka J. Int J Fatigue 1993;15(2):93107. [69] Flavenot J, Skalli N. In: Second International Conference on Biaxial/multiaxial Fatigue, Shefeld (UK), 1989. [70] Ezeilo AN, Heyes P. Conference on Comp. and Experimental Reciprocating Engines, November 2000. I. Mech. E. Conf. transaction 20008, pp. 203210, ISBN 1 80058 275 3. [71] Farrahi G, Lebrun J, Couratin D. Fatigue Fract Eng Mater Struct 1995;18(2):21120. [72] Farreira J, Borrego L, Costa J. Fatigue Fract Eng Mater Struct 1996;19(1):1117.

Вам также может понравиться