Track 1 Session 2

10:30 to 11:30 a.m. Tuesday, June 15, 2010

High Reliability and Systemic Risk: Predicting Rare Events and Failures in Modern Technological Systems

As an investor, owner, operator, manager or designer, the reduction of system failures and the prediction of reliability are key goals and attributes, thus reducing risk to the minimum achievable. But occasional events or outcomes still may occur and the probability of major loss is not negligible even when we have little or no prior data or experience. Such rare events are widely misunderstood. We already know from the world’s event data that the standard statistical distributions and methods used for probability and frequency of occurrence do not work for rare events, simply because the impact of learning, forgetting, randomness and experience are not properly accounted for. The events we experience include spectacular plane, train, space shuttle or stock market crashes; or may be more mundane auto collisions and medical errors. The rate of such events covers the whole spectrum from the frequent to the rare, so that they are unexpected. Rare events and/or limited data sets pose a special problem with highly uncertain or unknown outcome rates. In this presentation, we provide a new method to predict the future probability of such rare events based on the extreme case of insufficient learning. We compare the predictions against data for rare events and establish the uncertainties as an explicit function of future risk exposure.

Key Words: Reliability and Risk Exposure, Rare Events, Prediction and Influence of Human Risk Taking, Limitations on Standard Methods, Risk Under Prediction, Probability Estimation Case Studies

Romney B. Duffey

AECL Chalk River Laboratories

Chalk River, Ontario, Canada