Interdisciplinary safety analysis of complex socio-technological systems based on the functional resonance accident model: An application to railway trafficsupervision

This paper presents an application of Functional Resonance Accident Models (FRAM) for the safety analysis of complex socio-technological systems, i.e. systems which include not only technological, but also human and organizational components. The supervision of certain industrial domains provides a good example of such systems, because although more and more actions for piloting installations are now automatized, there always remains a decision level (at least in the management of degraded modes) involving human behavior and organizations. The field of application of the study presented here is railway traffic supervision, using modern Automatic Train Supervision (ATS) systems. Examples taken from railway traffic supervision Preprint submitted to Reliability Engineering and System Safety August 27, 2010 illustrate the principal advantage of FRAM in comparison to classical safety analysis models, i.e. their ability to take into account technical as well as human and organizational aspects within a single model, thus allowing a true multidisciplinary cooperation between specialists from the different domains involved. A FRAM analysis is used to interpret experimental results obtained from a real ATS system linked to a railway simulator that places operators (experimental subjects) in simulated situations involving incidents. The first results show a significant dispersion in performances among different operators when detecting incidents. Some subsequent work in progress aims to make these “performance conditions” more homogeneous, mainly by ergonomic modifications. It is clear that the current Human-Machine Interface (HMI) in ATS systems (a legacy of past technologies that used LED displays) has reached its limits and needs to be improved, for example, by highlighting the most pertinent information for a given situation (and, conversely, by removing irrelevant information likely to distract operators).