ABSTRACT
Considerable attention has been focused on the assessment of human error in risk analysis. In such analyses there is a requirement to comprehensively model the structure of the task to reflect the manner in which it is actually performed in practice. This model is an essential starting point for both qualitative and quantitative analyses of human reliability. A systematic process for performing such analyses called SHERPA (Systematic Human Error Reduction and Prediction Approach) was developed in the 1980s and has subsequently been refined and applied in a wide variety of industries. In SHERPA, a task or process is broken down in a hierarchical manner into the subtasks and finally task elements required to achieve its overall goal. The decision to break down a task in further detail is based on criteria such as the likelihood of error, severity of consequences and likelihood of recovery before the consequences ensue. This process allows the analysis to drill down into the areas of greatest risk. Once a task has been decomposed to a suitable level of detail, a structured risk identification process is applied in which a series of guide words, e.g. action omitted, right action on wrong object, is applied to identify specific errors, their consequences and possible control measures. The methodology is therefore similar to both FMEA and HAZOP techniques used in engineering risk analyses. The results of a SHERPA analysis can be used as the basis for generating improved operating procedures and competence management processes. The paper provides an example of the methodology applied to a safety critical unloading task in the oil and gas sector.
