Current Human Reliability Analysis models express error probabilities as a function of task types and operational context, without explicitly modelling the influence of different crew behavioral characteristics on the error probability. The influence of such variability is treated only implicitly, by variability and uncertainty distributions with bounds primarily obtained by expert judgment. This paper presents a methodology to empirically incorporate crew performance variability in error probability quantification, from simulator data. Crew behaviors are represented by a set of “behavioral patterns” that emerge in the observation of operating crews (e.g. in information sharing or in adhering to procedural guidance). The paper demonstrates the use of a Bayesian hierarchical model to explicitly capture the performance variability emerging from data. The methodology is applied to a case study from literature. Numerical demonstrations are performed in order to compare the proposed approach to the existing quantification models used in HRA for treating simulator data.

当前的人类可靠性分析模型将错误概率表示为任务类型和操作环境的函数，而没有明确地建模不同机组行为特征对错误概率的影响。这种可变性的影响只能通过可变性和不确定性分布来隐含地处理，其界限主要是由专家判断得出的。本文提出了一种方法，可根据模拟器数据将机组人员的性能变化经验性地纳入误差概率量化中。机组人员的行为由观察操作人员时出现的一组“行为模式”来表示（例如，在信息共享或遵守程序指导方面）。本文演示了如何使用贝叶斯层次模型来明确捕获数据中出现的性能差异。该方法适用于文献中的案例研究。进行了数值演示，以便将建议的方法与HRA中用于处理模拟器数据的现有量化模型进行比较。