On-line hazard aversion: a Bayes approach

Abstract

A Bayes theorem and fault tree (BFT) method is developed for online hazard aversion in process systems. BFT requires as input a prediction of the measurement patterns resulting from system operating nodes, equipment malfunctions and disturbances, and the prior rates of equipment malfunctions and disturbances. The Bayes theorem is used to estimate the posterior probability of each fault candidate, given the online measurements, producing a list of fault candidates ranked by their posterior probabilities. A fault tree is then used to estimate the current top event probability, given the online measurements in which basic event probabilities are the posterior probabilities of the faults. The future safety of the process is measured by safety meters which compare the estimated top event probability at given times over a time horizon. The estimated current top event probability and the safety meters indicate process safety, providing the basis for an unplanned shutdown, maintenance, or corrective action decision. A negative feedback process control loop illustrates the accuracy of BFT and lays the groundwork for applying BFT to more complicated process systems. >