A neural network approach to predict the time-to-failure of atmospheric tanks exposed to external fire
Nicola Tamascelli,Giordano Emrys Scarponi,Md Tanjin Amin,Zaman Sajid,Nicola Paltrinieri,Faisal Khan,Valerio Cozzani
DOI: https://doi.org/10.1016/j.ress.2024.109974
IF: 7.247
2024-02-03
Reliability Engineering & System Safety
Abstract:Domino scenarios triggered by fire pose severe risks to workers, assets, and the environment. Accurate quantitative models are needed to support mitigation actions addressing the prevention of fire escalation, especially considering sensitive targets such as atmospheric tanks containing large quantities of dangerous substances. A novel approach based on neural networks was developed, allowing the accurate quantification of the time-to-failure (TTF) of atmospheric tanks exposed to external fires accounting for mitigation actions. Data from a lumped parameter model were used to train and assess neural networks' performance. The toolbox of models obtained provides the TTF of atmospheric tanks both in the case of unmitigated fire scenarios and considering safety barriers and protection measures, such as water deluges and fire monitors. Model predictions are fast, accurate, and supplemented with confidence intervals. The comparative analysis demonstrated the better performance of the model developed compared to simplified correlations widely used in the literature to predict TTF. The approach developed, based on the integration of neural networks in consequence analysis tools, shows significant potential for the advancement of a quantitative assessment of domino scenarios, providing accurate and user-friendly tools for a quick evaluation of domino fire scenarios under both mitigated and unmitigated conditions.
engineering, industrial,operations research & management science