Dual fatigue failure modes in Ti–6Al–2Sn–4Zr–6Mo and consequences on probabilistic life prediction

Reducing uncertainty in fatigue life limits of turbine engine alloys

“…2. This separation in fatigue lifetimes, known as bimodal or competing-modes fatigue, has been observed in a wide range of alloys, including the superalloys: Rene'95 [21,22], Rene'88DT [23], IN100 [8,[24][25][26][27][28][29], Waspaloy [30], the single crystal alloy PWA 1484 [31], the titanium alloys Ti-10-2-3 [23,[32][33][34], Ti-6Al-2Sn-4Zr-6Mo [35][36][37][38][39][40][41][42], Ti-6Al-4V [43][44][45][46][47], gamma titanium aluminides [48,49], the aluminum alloy 7075-T651 [50], and others. Although such a separation of fatigue response has been known for some time [51], the significance of this behavior has not yet been generally captured in the strategies for fatigue design of turbine engine materials.…”

“…In exploring this behavior in turbine engine alloys, we have developed a physically-based approach for describing fatigue variability, and this approach has been integrated into a Monte Carlo probabilistic life prediction model for materials [35,40,52]. The primary tenets of our fatigue variability description are summarized by [40]: (i) under nominal microstructural and loading conditions, a hierarchy of local deformation heterogeneities develop in a fatigue sample, corresponding to grain-scale cyclic micro-plasticity in certain microstructural arrangements or features; (ii) as a result of this hierarchy, a probability exists of an extreme microstructural arrangement that may initiate a crack-growth dominated failure, producing a life-limiting failure distribution for the material; and (iii) there are often different degrees of influence of a given variable on minimum vs. the mean lifetimes.…”

“…Salient features of the model are (i) that the effect of any variable on lifetime distribution is separated into its influence on small-plus-large crack growth vs. the mean lifetime behaviors, and (ii) that the lifetime distribution is modeled as a superposition of these two responses. From a design life perspective (for example, a 1/1000, B0.1, lifetime), it can be shown that the prediction of life limit by the crack growth lifetime density, f l (N) (i.e., in the limit of p l = 1) forms a lower bound of predictions by the bimodal model [35], as discussed below. It is well known that the total fatigue life of a specimen is composed of the sum of the cycles from the successive stages of damage initiation and crack growth to failure, such that the total fatigue life (N T ) is given by:…”

Microstructural Influences on Very-High-Cycle Fatigue-Crack Initiation in Ti-6246

“…The 20 Hz tests were conducted at stress levels (r max ) of 820 MPa and higher at AFRL. [23] Ultrasonic-frequency fatigue testing was completed at 20 kHz at stresses (r max ) of 700 MPa and below. [24] Although the stress levels do not overlap, the data follow the same trend as expected for a typical SN curve.…”

“…[10,23,[32][33][34][35] The competition between surface and subsurface crack-initiation sites has alternately been attributed to the specimen surface to volume ratio, [27] the presence of compressive residual stresses on the surface, [34] environmental effects, [26] and the relative ease with which grains can deform at a free surface as compared to the specimen interior.…”