PRA update discipline before peer review
The peer-review findings that surprise utilities most often are the ones that come from the data update, not the model update. Our checklist for what to lock down first.
The model update gets the bulk of the attention in PRA refresh programs. The fault-tree changes, the event-tree changes, the new initiating events, the integration of operating-experience data: all visible, all reviewed by the engineering team multiple times, all defensible at peer review. The data update happens in parallel and is, in most utilities we have worked with, less rigorously reviewed.
The data update covers the component failure rates, the human-error probabilities, the common-cause failure parameters, and the success criteria for plant systems against the current operating configuration. Each of these is a place where a quiet update has cascading effects on the model results. A failure-rate change for a frequently-used component can swing the cutset frequencies enough to change the dominant accident sequences. A human-error-probability update against new operator training data can move the importance ranking of operator actions. A common-cause-failure parameter update against industry data can change the conclusion on a system's reliability.
Our practice is to require, before the PRA goes to peer review, that every data input that has changed since the last peer review has documented justification, a sensitivity-study output showing how the change affects the headline results, and an explicit acceptance or rejection of the change by the PRA technical lead. The documentation is not just for the peer review; it is for the technical lead's own ability to defend the data choices when they are questioned.
The peer review process under the NRC-endorsed PRA standards explicitly looks for this discipline. A PRA that has good model documentation and weak data documentation typically gets findings on the data side that are uncomfortable to close in the time available. Doing the data discipline before the peer review is much cheaper than doing it during the peer review under deadline pressure.