ABSTRACT
There are many different methods of human reliability assessment, each differing in its performance according to a number of criteria such as accuracy, resources usage, usefulness and auditability. For a large Probabilistic Risk Assessment (PRA), it is necessary to be resources-efficient when applying techniaues of human reliability assessment (HRA), so that the real human error problems get the attention in terms of depth of analysis which they need, and the more trivial problems (i.e. with less risk impact) are dealt with in a more expedient fashion.
In practice this is not so simple as often the perception of the severity of the problem alters as a function of the depth of analysis, and thus in real PRA’s the decision as to whether to evaluate a scenario in depth or not is based on fairly broad risk sensitivity criteria and analysts judgement.
BNFL are utilising two human reliability assessment approaches. The first is a ‘screening’ type of approach for putting broadly accurate estimates into risk assessments, but which also concentrates on making such assessments auditable and involves limited task and human error analysis. The second approach is a far more powerful approach to in-depth human reliability assessment (which is also far more resources - intensive), which involves full task and human error analysis, Performance-Shaping-Factor-based quantification, and error reduction analysis. The way in which these two approaches are utilised in order to make the PRA/HRA process efficient, is discussed in the paper with examples of the decision criteria for deciding when to implement the more detailed approach.
