Real time safety verification in the process industry using fault semantic networks (FSN)

Abstract

In the process industry, there are uncertainties associated with each variable, which might lead to process deviations and hazards. In order to accurately quantify the risks associated with these hazard scenarios, quantitative probability should be calculated. The process dynamically changes during plant operation, which requires continuous monitoring of process risks and real time safety verification. It is challenging to both dynamically and instantaneously estimate the risks for all faults and deviations. An FSN is introduced in this thesis to systematically and continuously estimate risks for all possible fault propagation scenarios. Intelligent reasoning algorithms are proposed using a BBN to accurately estimate risks. An FSN is used to analyze causes and consequences of different faults using automated forward and backward propagation learning techniques. Real time safety verification is applied to each fault propagation scenario. The TE process is used to illustrate the proposed real time safety verification. An FDS experimental setup is used to study real life data.