Application of a Quantitative Risk Assessment (QRA) in a refinery unit

Abstract

In this study a common basic distillation unit has been chosen to perform a Quantitative Risk Analysis following the Purple Book and the OGP database. There has been an explanation of the decisions taken in order to perform this analysis. The main objective of this study is to see the differences between using those two databases particularly in terms of frequency and consequence calculation. For this to be done, there has been a selection of Initial Events situated in representative places of the process that could lead to significant accidents. The Initial Events have been the same for both databases, the only variables that changed between these two ways on analysis has been the types of leaks for each Initial Event and the frequencies used to calculate frequencies for each type of leak and final Initial Event frequency (summation of every leak present in an Initial Event). Consequences obtained with these two databases have been similar, the only significant difference between these two database in terms of consequences assessment, was the number of types of leaks in every Initial Event. When talking of frequencies assessment, these two databases had different values for final Initial Event frequency. In the majority of Initial Events, the OGP database was more conservative. When the risk values were obtained, the results following the Purple Book database were more conservative contradicting the initial hypothesis. After an analysis of the results there was seen that the Initial Event with worse consequences was also one of the few who had greater values of frequency for the Purple Book database. This higher values of frequency for this Initial Event have an explanation related to frequency values with large diameter pipelines. In the OGP databases, in case of full rupture of a large nominal diameter pipeline, the frequency of this scenario tends to zero, not being that in the case of Purple Book. Therefore, it is correct to conclude that the OGP database is more realistic in terms of frequency values but in this case, this values have given less conservative results