Time-dependent creep induced failure is a major concern for structural components (i.e. IGT components, Gen IV nuclear reactor components) operating at elevated temperature. The likelihood of a failure is aggravated by randomness in several sources of uncertainty. Creep rupture data shows expanding scatter bands for long-duration creep tests where uncertainty can span multiple logarithmic decades of life. This experimental uncertainty is exacerbated by the uncertainties that exist during service. The continuum damage mechanics (CDM) based creep-damage model readily available in literature does not consider the uncertainty effect while predicting the long-term reliability of the components; rather the problem is tackled deterministically. Introduction of probabilistic phenomena into the existing model to predict the minimum-creep-strain-rate (MCSR) and stress-rupture (SR) would present a pathway for estimation of effect of uncertainty ensuing high reliability in the assessment.
The objective of this paper is to develop a probabilistic model for MCSR and SR that is capable of predicting experimental uncertainty and extrapolating the expanded scatter bands observed in long-duration creep data. The Sine-hyperbolic (Sinh) CDM model is selected. Multi-isotherm MCSR and SR data for 304 (18Cr-8Ni) and 316 (18Cr-12Ni-Mo) stainless steel are gathered from the NIMS material database. A deterministic calibration is performed where the optimal material constants are obtained with no initial damage and perfect loading conditions. Probabilistic calibration begins with adding ASTM-specified temperature and stress tolerances (± X°C, ±Y% MPa) to capture a portion of the experimental uncertainty. The initial damage tolerances is then calibrated to capture the remaining uncertainty in the data. Probability distribution functions (pdfs) are assigned to each uncertainty parameter. Monte Carlo simulations are performed over a range of stress and temperature. The probabilistic Sinh model is shown to predict the expanding scatter band observed in long-term MCSR and SR data. Parametric simulations are performed where service condition uncertainty is added to the probabilistic model. It is determined that service condition uncertainties further degrade the creep resistance of a material.