O-180 Probabilistic Aspects of the Initiation of Explosives and Ammunition

The probability of an accidental initiation of explosives and ammunition is an important parameter in Quantitative Risk Analysis (QRA). Existing models for the frequency of event are primarily based on historical data, and do not address details such as the involved threat stimuli and type of munitions. Also the response level of the ammunition is not specifically addressed. Current models may therefore overestimate the risk posed by Insensitive Munitions (IM) and conventional munitions in situations where mitigation measures are deployed.

This paper proposes a methodology to make more realistic predictions for the event frequency based on fault tree analysis, and by modelling uncertainty and variability. The Bayesian Belief Network (BBN) technique was selected to conduct two case studies.

Shock induced detonation due to fragment impact was analysed with the Jacobs-Roslund (JR) model. By taking into account variation in all model parameters, a realistic prediction of the probability of exceeding the critical impact velocity could be obtained. A second case study deals with the prediction of the Inhabited Building Distance for ammunition storage, taking into account variability in parameters like magazine filling level over time and TNT equivalence.

The presented case studies clearly show the benefits of probabilistic analysis in the field of explosives and ammunition safety. Probabilistic analysis gives a more realistic result than a deterministic and typically conservative analysis. Also it enables determination of more accurate event frequencies to be used in QRA when the input data is representative. Challenges are to define the required probability distributions as input to BBNs, and caution should be taken not to miss black swan events. Future case studies are planned on among others dropping of explosives, range safety and IM.