The current best-estimate model describing the fracture toughness of ferritic steels is the Master Curve methodology standardized in ASTM E1921. Shortly following standardization by ASTM, efforts were undertaken to incorporate this best-estimate model into the framework of the ASME Code to reduce the conservatisms resulting from use of a reference temperature based on the nil-ductility temperature (RTNDT) to index the plane strain fracture initiation toughness (KIc). The reference temperature RTT0, which is based on the ASTM E1921-defined T0 value, was introduced in ASME Code Cases N-629 (replaced by Code Case N-851) and N-631 to replace RTNDT for indexing the ASME KIc curve. Efforts are continuing within the ASME Code to implement direct use of the Master Curve model; using the T0 reference temperature to index an elastic-plastic, KJc fracture toughness curve.
Transitioning to a direct T0-based fracture toughness assessment methodology requires the availability of T0 estimates for all materials to be assessed. The historical Charpy and NDT-based regulatory approach to characterizing toughness for reactor pressure vessel (RPV) steels results in a lack of T0 values for a large population of the US nuclear fleet. The expense of the fracture toughness testing required to estimate a valid T0 value makes it unlikely that T0 will ever be widely available. Since direct implementation of best-estimate, fracture toughness models in codes and regulatory actions requires an estimate of T0 for all materials of interest it is necessary to develop an alternative means of estimating T0.
A project has been undertaken to develop a combined model approach to estimating T0 from data that may include limited elastic-plastic fracture toughness KJc, Charpy, tensile, ductile initiation toughness, arrest toughness, and/or nil-ductility temperature data. Using correlations between these properties and T0 a methodology for combining estimates of T0 from several sources of data was developed. T0 estimates obtained independently from the Master Curve model, the Simple T28J correlation model, and a more complex Charpy correlation model were combined using the Mixture Probability Density Function (PDF) method to provide a single estimate for T0. Using this method, the individual T0 estimates were combined using weighting factors that accounted for sample size and individual model accuracy to optimize the accuracy and precision of the combined T0 estimate. Combining weighted estimates of T0 from several sources of data was found to provide a more refined estimate of T0 than could be obtained from any of the models alone.