The damage states in a storage tank subjected to seismic loading can induce loss of containment (LOC) with possible consequences (fire, explosion, etc..) both for the surrounding units and people. This aspect is particularly crucial for the Quantitative Risk Analysis (QRA) of industrial plants subjected to earthquakes. Classical QRA methodologies are based on standard LOC conditions whose frequency of occurrence is mainly related to technological accident rather than natural events and are thus useless. Therefore, it is evident the necessity of establishing new procedures for the evaluation of the frequencies of occurrence of LOC events in storage tanks when subjected to an earthquake.
Consequently, in this work a simple procedure founded on a probabilistic linear regression-based model is proposed, which uses simplified numerical models typically adopted for the seismic response of above ground storage tanks. Based on a set of predetermined LOC events (e.g. damage in the pipes, damage in the nozzles, etc..), whose probabilistic relationship with the local response (stress level, etc..) derives from experimental tests, the probabilistic relationship of selected response parameters with the seismic intensity measure (IM) is established. As result, for each LOC event, the cloud analysis method is used to derive the related fragility curve.