(772c) Risk Analysis In the Process Industries: A Perspective

Fortunately, accidents in the chemical process industries (CPIs) have very low frequencies of occurrence.  Typically, the probability of an event is estimated using its observed frequency, but for rare events like accidents, data are too sparse.  In such a case, risks of occurrence of these events are computed based on: (a) hazards identification, (b) their frequency estimation and (c) consequence modeling.  Since the late 1970s and early 1980s, these concepts (leading to techniques like HAZOP, PHA, LOPA) have received much attention in the process and nuclear industries.  A system (or plant) is decomposed into individual components – for example, power generators, pumps, and heat exchangers.  Possible hazards are identified, their probabilities of failure under demand and likely consequences are estimated using reliability (which could be generic or prior) data and expert knowledge. 

More recently, techniques have been developed to access the risk levels dynamically using alarm data recorded in plant databases.  Because the accident data is limited, the near-misses that occur in day-to-day operations are tracked, and the failure probabilities of regulating and protection systems, and the probabilities of incidents are computed. 

This presentation is intended to describe the advantages and disadvantages of many methods for analyzing alarm databases, including the estimation and online tracking of key indicators, and the use of leading indicators to project the occurrence of incidents sufficiently far in advance to permit operators to circumvent them.  Also, methods are being explored to identify the special causes that lead to the abnormal events that cause alarm flooding – which often significantly distracts operators as risk levels build.  In this presentation, a review of state-of-the-art techniques and practices for risk analysis is presented and promising new developments in future research are projected.