A Human Operator Function Reliability Statistical Modelling For Preliminary System Safety And Risk Assessment

This paper describes key concepts, approach and modelling details for preliminary safety (risk) assessment of human functions in complex systems controlled by a human operator or a group of operators. Human function consists of sequence of tasks and each task can be “abnormally” executed by human potentially leading to undesired events compromising system safety. Usually function reliability is measured in terms of function failure probability (likelihood) and the actual figure comes from a subject expert. A new approach has been developed to assist human factor and safety experts in human function reliability estimations and assessment. Monte-Carlo statistical modelling is used for human function assessment. In the model each function execution (as a set of tasks) generates an output stochastic sequence of normal or failed tasks. This output is subject for a safety analysis, conducted by a system specific algorithm which generates output safety index value (representing levels of safety), for example, an accident, major or minor system mishap. Number of tasks in a function and individual task failure probabilities will determine number of iterations to produce credible safety index distribution. The prime focus is “spontaneous” human errors not complicated by unhelpful system events, like equipment failure, or other human performance shaping factors, like, fatigue. The approach gives an estimation of a “pure” human impact on function or operational procedure reliability. However, this model and the associated software tool can be used in the different contexts, for example, in estimating function reliability under fatigue by applying “fatigue” multiplier to each task failure probability. The heart of the model is human agent which takes input information: task, failure task nominal probability, nominal task safety level associated with the particular task (provided by safety expert) and produces stochastic output result: task failure or normal task execution. Human agent model behaviour has interesting features based on psychological observations, for example, subjective task failure probability correction, ability to notice and correct (mitigate) failed tasks. A new concept of “human error entropy” as a human performance factor influencing failure rate probability introduced in the human agent model will be discussed.