Actions Imposed on Structures During Man-made Accidents: Prediction Via Simulation-based Uncertainty Propagation

Prediction of mechanical, thermal, and chemical actions induced during man‐made accidents (accidental actions) is of crucial importance to assessing potential damage to structures exposed to these actions. A logical result of such a prediction may be expressed in the form of probabilistic models describing likelihood of occurrence and characteristics of accidental actions. For many types of accidental actions the models are to be selected under the conditions of incomplete knowledge about and/or scarce statistical information on intensities and likelihood of imposition of the actions. This paper proposes a simulation‐based procedure intended for a selection of the probabilistic models under these conditions. The proposed procedure is formulated in the context of the classical Bayesian approach to risk assessment. The main idea of it is that statistical samples necessary for fitting the probabilistic action models can be acquired from a stochastic simulation of accident sequences leading to an imposition of accidental actions. Formally, the stochastic simulation of accidents serves the purpose of propagating uncertainties related to the physical phenomena capable of inducing accidental actions. These uncertainties are quantified in line with the classical Bayesian approach. The simulation‐based procedure can be used for damage assessment and risk studies within the methodological framework provided by the above‐mentioned approach.