Development of a dynamic risk assessment and control methodology for industrial accidents

Laburpena

Abstract The most affected types of industries affected by accidents in the European Union in the period of years 1979-2019 are chemical-pharmaceutical and petrochemical-refining with causes covering operations, failure in equipment, explosion and fire and human errors. This situation is followed by an occupational accidentability rate of 2 fatal accidents by 100,000 persons employed in 2015. To manage situations that affect risks in industrial processes and occupational risk at work, a review of existing tools is first carried out taking into account three prevention, simultaneity and immediacy characteristics. As a result, a new dynamic methodology called Statistical Risk Control (SRC) based on Bayesian inference, control charts and analysis applying hidden Markov chains for the initiating causes and safety barriers is presented. The objective is to detect the situation outside the limits early enough to allow corrective actions to reduce the risk before an accident occurs under the concept of immediacy. Several cases are developed testing different Bayesian inference models. The methodology covers industrial risks, occupational accidents, loss of containment with possible domino effect and deviations in cost-time A total of 12 inference models have been tested performing the analysis of collected observations of initiating causes of risk and safety barriers failure using a Metropolis-Hastings sampling when it is needed. Collected values are presented in tables or control charts visualizing when the situation is out of limits. The results show that the methodology offers a formal procedure for to have a determination of failure probabilities that can prevent or mitigate accidents and occupational hazards in manufacturing scenarios and industrial processes warning of the existence of a risk to act in advance correcting causes and offering a complete vision in the simplest and most practical way possible and responding to the three characteristics of prevention (P), simultaneity (S) and immediacy (I). Keywords— Risk, Bayesian, Assessment, Control, Occupational accident, Cost, Domino effect