ABSTRACT
Safety performance indicators (like FAR, TRIR, LTIF and severity) express the distilled information from accident reporting. The present work extends available Bayesian tools, and properly adapts them to receive evidence in the form of safety performance indicators over a given period of observation. The models evaluate accident frequencies, duration of recovery from an accident, and worker unavailability. The extended methodology enables the exploitation of readily available data sources, like open databases of accidents in the Petrochemical industry worldwide. The results provide predictions (and prediction uncertainties) on accident statistics and work time losses, depending on the site location.
