Abstract
In the petrochemical industry, storage tank failures may lead to unfavorable consequences or even catastrophic events. The risk management of chemical tank farms must identify all possible failure modes and reduce the risk for critical failure modes. However, identifying and assessing failure modes often depends on experts' subjective judgment, and epistemic uncertainty will affect the validity of assessment results. This paper proposes a novel hybrid methodology for identifying and assessing failure modes of loss of containment (LOC) of storage tanks considering expert epistemic uncertainty. The improved FMECA method considering the expert epistemic uncertainty is used to compare and analyze the possible risk priority numbers (RPN) and RPN critical thresholds of the failure modes. In this process, the Dempster-Shafer theory is used to obtain all possible RPN intervals, and Data Mining is used to provide a reference for the occurrence of failure modes in the FMECA process, and the possible failure modes are expanded. This methodology is applied to a storage tank system, and the ranking of critical failure modes causing LOC of the storage tank is obtained, among which the most critical failure mode is Rupture disc failure. The assessment results will be more valid after considering the expert epistemic uncertainty. Risk management based on the current risk assessment results can effectively reduce the risk of LOC accidents, and then reduce the occurrence and severity of more serious accidents. The relevant results can provide a reference for subsequent related research and risk management in chemical industrial parks.
