Analysis of Warm Prestress Data

Abstract

Loading a cracked structure at elevated temperature, or warm prestressing (WPS), enhances its fracture resistance at a lower temperature. Five data sets, comprising 119 unclad pressure vessel steel specimens, were combined to derive correlations for WPS-enhanced fracture toughness (KIfrac) in the absence of ductile tearing. New WPS test results for 27 surface-flawed specimens, eight subclad-flawed specimens, and five strain-aged specimens are discussed. KIfrac exceeded non-WPS fracture toughness, KIc, for all experiments. The WPS data showed that no specimens failed while K was decreasing, and that at least an additional 7% additional reloading from the minimum value of applied KI took place prior to final fracture. The data included complete and partial unloading after WPS prior to final fracture. Crack tip three-dimensional elastic-plastic finite element (3DEPFE) analysis was performed to support statistical analysis of the data. Regression models were compared with the Chell WPS model. Crack tip 3DEPFE analysis indicated that partially unloaded and completely unloaded data should be treated separately, and that the amount of unloading is unimportant for partially unloaded data. The regression models, which use KI at WPS (KIwps) and KIc as independent variables, better represented the WPS benefit than did the more complicated Chell model. An adequate accounting was made for constraint in the WPS experiments. The subclad flaw data support the use of the partial unload regression model provided that some care is taken to represent the effect of intact cladding if present. The effect of strain aging at or below 260°C (500°F) on the WPS benefit was of no consequence for the pressure vessel steels and WPS temperatures used to derive the regression models. The presence of ductile tearing precludes the use of the regression models. The regression model for partial unloading accurately predicted the behavior of full-scale pressure vessel WPS experiments. All but one of the 174 experiments considered lie above the lower 2σ estimate of the regressions. The experiments all supported Type I WPS, i.e., there was no fracture during cooling until reloading occurred. However, the regression equations apply to the reload and are inapplicable to Type I WPS.