Fuel storage facilities concentrate large inventories of flammable substances, so that a single primary accident may escalate, through thermal radiation, into a cascading sequence involving several tanks (domino effect). This work presents the probabilistic core of a digital twin for such facilities, understood as a probabilistic replica of the facility with a fast query interface (live-data assimilation is not yet implemented): a permanently available, interactively queryable model that characterises, in probabilistic terms, the thermal radiation field generated by accident scenarios in a storage facility and derives from it an integrated assessment of its safety and economic risk. The framework combines (i) a Monte Carlo simulation of 108 facility-years, in which primary accidents and their fire-driven escalation are sampled using validated models of pool fires, thermal radiation and equipment failure; (ii) a point-wise condensation of the resulting radiation intensity into Gamma distributions fitted by maximum likelihood, whose parameter maps embody the frequency of every fire configuration; and (iii) a machine-learning surrogate, based on Generalized Regression Neural Networks, that renders the probabilistic field continuous over the terrain and evaluable in a fraction of a millisecond. On this backbone, the model delivers mutually consistent, receptor-level indicators: annual exceedance frequencies of harm thresholds (contour maps), the expected number of fatalities, distinguishing sheltered and outdoor population, and the Expected Annual Loss of every asset. In a demonstration on a five-tank facility storing light naphtha, jet fuel and diesel within a mixed industrial–residential complex, 16.5% of the accidents escalated to multiple tanks, the individual risk at the most exposed third-party location lay more than one order of magnitude below the 10−6 per year land-use criterion, and 97.5% of the expected losses were borne by the facility itself. The scope is limited to pool fires in fixed-roof atmospheric tanks, in wind-free conditions and without credited safety barriers.
No takes yet. Share an insight, caveat, or question.
Arana et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: