Subcritical Crack Growth Models For Static Fatigue Of Hi-Nicalon(Tm)-S Sic Fiber In Air And Steam

Three subcritical crack growth (SCG) laws were used to model strain-rates and failure times for static fatigue of Hi-Nicalon(TM)-S SiC fiber tows in air and Si(OH)(4)(g)-saturated steam. Models were fit to tow failure times (t(f)) and steady-state strain rates (e) for brittle creep measured at 700 to 1100 degrees C under initial applied stresses (sigma(A)) of 260 to 1260 MPa. A power law, a reaction-rate law, and a bond-energy law were used to describe SCG that caused sequential filament failure, and ultimately tow failure. Two versions of each model were developed. One allowed access of chemisorbed species to flaws throughout the fiber (mode 1) and another only allowed access to flaws at the SiC-SiO2 interface (mode 2). The stress increase on intact filaments as others fractured and as filaments oxidized, and the increase in stress intensity geometric factors (Y) as crack size increased were incorporated in the models. The fits to data were compared for the different models by using both simple regression analysis and orthogonal distance regression (ODR) analysis. Faster convergence and more consistent results were achieved using ODR analysis. Regression analyses found parameters for all models with similar error in data fits, so validity of a model could not be distinguished by regression analysis alone. For all models, the stress dependence of SCG rates was much stronger in steam than in air, and for most models activation energies were between 300 and 420 kJ/mol, regardless of environment. For the steam environment, the bond-length parameter (delta) for the bond-energy model was very close to the lattice parameter of beta-SiC (.436 nm), but in air it was significantly lower at 0.25-0.26 nm, but still larger than the Si-C bond length of 0.189 nm. Other factors suggest that either a bond-energy based law or a modified version of a reaction-rate law are the best choices for a SCG law. Filament strength distributions initially described by Weibull distributions could not be described by such distributions after application of the models. SCG mechanisms are discussed.