Multivariable modeling of reactor pressure vessel weld fracture toughness

Models for predicting J-R curves were developed, using advanced pattern recognition tools to help identify important variables and the optimal functional forms. Several models were developed for various databases; the ones reported here were calibrated to an extensive set of reactor pressure vessel (RPV) weld data. Three models are presented: the first predicts Jd as a function of crack extension, pre-irradiation Charpy energy, weld copper content, fluence, temperature, and specimen thickness; the second and third models predict Jd and Jm as a function of the same variables except for thickness. The resulting models predict J within 17-18% (±1Se) using 5 or 6 variables and 8-12 fitting constants. The radiation exposure of the material must be known, but the models require no measured properties of the irradiated material