The increasing technological complexity and scale of process plants demands proportionate attention to the contribution of human performance if public safety and capital investment are not to be jeopardised. In some industries there exists a strong long-term commitment to human factors engineering to maximise human reliability and to ensure that the generally good engineering reliability inherent in most plants is not to be frustrated by human error which now provides a major contribution to unreliability. This unreliability results from activities in design, construction, operation, but predominantly in maintenance and testing.

Human Performance: What Improvement from Human Reliability Assessment

Risk assessment: Controlling hazardous materials

The prior is chosen somewhat conservatively, but not arbitrarily, in Bayes analysis for the event whose characteristics are not well understood. It is desirable to get the prior through an objective procedure for the given information. The author's approach of assigning the pre-prior is by coupling the principle of maximum entropy with the moment-matching method; i.e., to find upper and lower bounds of the population parameter sets-based on upper and lower bounds of the entropy for the given information. The methodology is demonstrated by applying it to the data sets of initiating events taken from the performance of nuclear power plants. >

Systematic Bayes prior-assignment by coupling the mini-max entropy and moment-matching methods

An application of the two-stage bayesian method to estimate the outrage rate of oil-fired boilers

Expert judgement concerning human performance in nuclear power plants is quantitatively coupled with actuarial data on such performance in order to derive plant-specific human-error rate probability distributions. The coupling procedure consists of a two-stage application of Bayes' theorem to information which is grouped by type. The first information type contains expert judgement concerning human performance at nuclear power plants in general. Data collected on human performance at a group of similar plants forms the second information type. The third information type consists of data on human performance in a specific plant which has the same characteristics as the group members. The first and second information types are coupled in the first application of Bayes' theorem to derive a probability distribution for population performance. This distribution is then combined with the third information type in a second application of Bayes' theorem to determine a plant-specific human-error rate probability distribution. The two-stage Bayesian procedure thus provides a means to quantitatively couple sparse data with expert judgement in order to obtain a human performance probability distribution based upon available information. Example calculations for a group of like reactors are also given.

Earl I. Patterson

Papers

Plant specification of generic human error data through a two-stage bayesian approach

Plant specification of generic human-error data through a two-stage Bayesian approach