No description available.

The thesis introduces a new quantitative safety modelling approach for large scale socio-technical systems that exhibit Low Probability High Consequence (LPHC) accidents. This approach uses Bayesian Belief Networks (BBN) and overcomes inherent difficulties of current methods, difficulties that make these methods incapable of coping with complexities of socio-technical systems.

The dissertation deals with the analysis of roundabout design elements and their impact on accidents. The analysis objective was to identify the important elements of roundabouts that have impact on accidents. In order to achieve this goal, the multifactorial statistical safety assessment method was used on the basis of a representative sample of data, by developing several safety performance functions, verifying them and interpreting the result. Several design elements, which from the point of view of the traffic accident mechanism belong to the infrastructure factor, have been identified: AADT, average diameter, entrance width, entry angle, direct passage angle, location and many others. The original sample contained about 1200 roundabouts, which were reduced to 200 based on data availability. Accident frequencies were monitored between 2009 and 2016, i.e. for eight years, resulting in total 2674 roundabouts accidents. The result is an accident prediction model, developer based on roundabout approach design elements, and map of critical roundabouts, identified based on empirical Bayes estimate of accident frequency. Following approach parameters were identified: AADT, entry angle, distance between collision points, deviation of angles between approaches, presence of apron, presence of bypass, entry type, presence of pedestrian crossing and surrounding area type.