Abstract: The parameter calibration for a strain-energy-density-based fracture failure (SDFF) model was improved by a maximum likelihood estimation method, and the multi-temperature SDFF model whose parameters could be calibrated by the fracture toughness test data at different temperatures was established. The model parameters were calculated with low-temperature fracture toughness test results of a one-inch thick compact tension (1T CT) specimen and a 0.5-inch thick compact tension (0.5T CT) specimen sampled from reactor pressure vessel (RPV) steel. The ductile-brittle transition temperatures of the two specimens were predicted, and the lower boundary lines of fracture toughness were analyzed. The results show that the ductile-to-brittle transition temperatures of the 0.5T CT specimen and 1T CT specimen calculated by the multi-temperature SDFF model were −81 ℃ and −64 ℃, respectively, and the calculated results were in good agreement with the test results with a relative error about 6%. The fracture toughness data measured at −90 ℃ to 20 ℃ were all within the 5% to 95% failure probability envelope, and the model had good adaptability to specimens of different size from −90 ℃ to normal temperature. When the 0.5T CT specimen was used to analyze the fracture toughness of the material in the ductile-brittle transition zone, with whose 5% fracture failure probability curve as the lower boundary line of fracture toughness could cover all the test data of 0.5T CT specimen and 1T CT specimen, and the margin was small; the margin with the 1% fracture failure probability curve as the lower boundary line of fracture toughness was larger.