ON THE QUANTIFICATION OF HUMAN RELIABILITY MODELS BASED ON COGNITIVE MECHANISMS OF HUMAN ERRORS

摘要

The second generation human reliability models have as a main feature the incorporation of cognitive models and a detailed taxonomy of human errors, in such a way that they present coherence in terms of error classification with the adopted cognitive model. This work has as a main goal a preliminary revision of the main second-generation human reliability models, trying to show their advantages and shortcomings, particularly in what concerns the quantification of the human error in complex tasks of industrial plants. Three second-generation models will be examined: ATHEANA, CREAM and DYLAM. These three models are sufficiently complex to be named cognitive reliability models. Two main aspects will be explored in these models: the error mechanisms and the adopted or proposed quantification methods. Concerning the error mechanisms, the chosen models will be compared with the Reason's model of human errors in three performance levels: skills, rules and knowledge, considering the concept of unsafe acts and the basic error types related to such acts, in other words, slips, lapses and mistakes. Besides the basic error types, the error forms and the activation mechanisms that occur in the retrieval of incomplete semantic knowledge related to such error forms will also be investigated.