The fatigue properties of un-notched polished specimens of an aluminium casting alloy have been measured for various heat-treatment conditions and at various mean stresses. The relation between fatigue life and alternating stress is insensitive to heat-treatment and, apparently, to mean stress. It was observed that failure initiated at interdendritic shrinkage defects: evidence of classical crack initiation from persistent slip bands was also seen but such cracks, being less severe than the casting defects, never caused failure. A fracture mechanics analysis for the growth of fatigue cracks from the pores is described. It shows that the fatigue life can be quantitatively predicted from a knowledge of the size of casting defects: in particular it explains the lack of effect of heat-treatment and the apparent absence of a mean stress effect is shown to be caused by the variation in size of maximum defect present among the specimens tested. It is shown that reducing the size of shrinkage defects will increase the life, but only up to the stage at which initiation from persistent slip bands becomes operative. NOMENCLATURE urn,, = maximum stress in the fatigue cycle urn," = minimum stress in the fatigue cycle C T~ = alternating stress amplitude [=0.5((rrnax -urn")] ACT = alternating stress range [=2u,] R = stress ratio [=urn,, /urnax] utS = tensile strength of the alloy n = log,, Nf K = stress intensity factor K,, = maximum value of K in the fatigue cycle Y ( a ) = compliance calibration factor AK = alternating stress intensity factor range AKeR = alternating effective stress intensity factor range U = ratio between AKer and AK [U = AK,,/AK] AKea ,,, = value of AK& at threshold U, = value of U at some value of R N, = number of cycles to initiate a crack Nf = total number of cycles to failure Np = number of cycles to grow a crack to failure [N, = Nf -N , ] a = crack length C, m = constants in the Pans fatigue crack growth equation a,, a, = initial and final crack lengths after Np cycles *Present address:
