Fire risk assessment with quantification of uncertainty for offshore installations

Abstract

This dissertation investigated the significance of categorizing and selecting representative leak sizes in the Quantitative Risk Assessment (QRA) for estimating fire risks. In most fire QRA procedures, the full-bore sizes of leaks have been overlooked even though they causes the catastrophic consequences with relatively high frequencies. So, in the present study, the effect of particular leak sizes on the individual risks (IRs) were investigated through the fire risk assessment. As case studies, LNG (Liquefied Natural Gas) fuel gas supply system for a large ship were considered with three different approaches. Approach 1 considers only three categories of leak sizes

small, medium, and large. Approach 2 includes the full-bore leak in addition to categories of Approach 1, and Approach 3 reflects more categories of leak sizes with every 1 mm increment in diameter. Compared with the total IR of Approach 3, the total IR of Approach 1 was two times lower, whereas Approach 2 resulted in the comparable total IR with an approximately 4 % deviation. Admitting that Approach 3 should be close to the exact solution, Approach 2 gave acceptable results, whereas Approach 1 unsatisfactorily underestimated the total IR. The results indicated that full-bore leak accidents are critically important scenarios and omission of them can lead to misjudgment in the risk-informed decision-making process. Taking the full-bore accidents into additional account to the conventional simplified QRA can reduce the time and effort for estimating the leak-related risks without a significant loss of accuracy in the estimated risks. This dissertation also investigated the critical issues for determining the DAL (Design Accidental Load) fire procedure based on the QRA for offshore installations. Considerable attention was paid to parametric uncertainty in choosing the numerical values used for the frequency and consequence analysis. In particular, selecting the initial leak size was one of the most critical aspects, and inconsistent approaches for selecting this value resulted in different risks for identical systems. Frequency analysis of past investigations also overlooked the inaccuracy and unsuitability of statistical data. Accordingly, the estimated risks were significantly uncertain, and the lack of information about the results increased the risk of making the wrong decision. In this study, the Latin hypercube sampling (LHS) technique was used to treat parametric uncertainty in the QRA. Different fire exceedance curves and DAL fires were demonstrated by selecting different sets of representative values. The distribution and confidence interval of the DAL fires showed a wide distribution with varying uncertain and critical parameters. Therefore, this procedure provided quantitative information on inherent uncertainty, and such additional information regarding DAL fires can lead to better decision making.