One of the requirements for validating nuclear reactor plant models is to benchmark the predicted results of selected transients against measured plant data or another qualified code. A major challenge with benchmarking is the criteria for validating a model are not always well defined and rely heavily on human judgment, thus introducing the possibility of human bias or inconsistent conclusions. The validation process can also be time consuming.
A new method is presented to aid in the validation of nuclear reactor plant models, using the Automated Code Assessment Program (ACAP), which is a tool developed at Pennsylvania State University under contract by the U. S. Nuclear Regulatory Commission (NRC). The proposed method was developed specifically for real-time best-estimate nuclear operator training simulator transients. However, the tool can be easily customized for most applications (e.g., design models, steady state data). Four distinct statistical metrics and weightings were chosen, as deemed appropriate for transient nuclear operator training simulator applications. The metrics account for errors in magnitude and trend, and incorporate an experimental uncertainty. The four metrics are then combined into a single Figure of Merit (i.e., a statistical level of agreement between the predicted and benchmarking data sets).
The use of ACAP for benchmarking is demonstrated by comparing experimental data from the Loss-of-Fluid-Test (LOFT) facility Large Break Loss-of-Coolant Experiment L2-5 with code predictions from a RELAP5-3D (Version 2.9.3+) model previously developed and published by Idaho National Laboratories. The proposed method is shown to have several advantages over conventional validation methods, in that it greatly reduces the possibility of human bias, generates reproducible results, can be easily automated to improve efficiency, and can be easily documented. After the initial validation, the tool can also be used to re-validate models after computer hardware changes, model changes, tool upgrades, and operating system upgrades.
Digitizing Native American Collections