A Safety Model for Industrial Environment Based on the Bayesian Network Paradigm

A safety model for identifying hazards and predicting their possible consequences in an industrial facility is proposed. Complex cause-effect relations between safety-related events, starting from the initial ones, through the intermediate, to the final, are represented in the form of a Bayesian network. The input data originate from sensors and meters installed in safety-critical locations. Using the Bayesian network methodology, the impending hazards, accidents, or machinery breakdowns can be predicted from symptoms indicated by the monitoring devices. Also, the "reverse analysis" of the network can establish the root causes of these undesired events so that a preemptive maintenance can be carried out in order to avoid them. For illustration, a simplified safety model of a biogas plant is presented along with its basic analysis. Although there is abundant literature on Bayesian networks in the safety and reliability context, much of it is limited to theoretical considerations or provides only general guidelines for the construction of such networks. Thus, publications reporting specific applications of this methodology are rather rare. The current paper aims to contribute to filling this gap.

Keywords: Workplace safety, Bayesian network, Biogas plant, Hazards identification and prediction, Root cause analysis.