As a nuclear engineer, I have experience working at nuclear plants in probabilistic risk assessment. At one particular nuclear plant, they use bayesian updating to calculate the risk of a component failing in the nuclear plant. They basically collect data by starting the component up and every time the component starts up initially, that’s a pass and every time it stalls, that’s a fail. They use this data along with bayesian updating to calculate the risk of failure of the component and assess when the component needs to be replaced. The pros of this method is that is uses actual data from the component, and when properly done, it is at least somewhat accurate. The problem I have with this particular nuclear plant using this method is that it didn’t appear to be standarized. They used an excel spreadsheet to do the calculations and I would constantly find errors in the spreadsheet. They had nothing in place to find these errors. They could have resolved it by having multiple people reviewing the calculations and redoing the calculations to ensure that they all came to the same answer. Instead, they had a manager who knew little to nothing about the actual calculations sign off on the calcs. As an actual nuclear engineer, I understand that a nuclear plant is designed to have failures of components and still function properly. There is no risk to the public or the actual plant due to this but it is just one step closer to being a problem to the public or actual plant due to them not reviewing their work.