Use of Statistical Tools for Comparison between Different Analytical and Semi-Empirical Models of the Bleve Fireball

Abstract

The Bleve is an explosion involving both the rapid vaporization of liquid and the rapid expansion of vapor in a vessel. The loss of containment results in a large fireball if the stored chemical is flammable. In order to predict the damage generated by a Bleve, several authors propose analytical or semi-empirical correlations, which consist in predicting the diameter and the lifetime of the fireballs according to the quantity of fuel. These models are based on previous experience, which makes their validity arbitrary in relation to the initial conditions and the nature of the product concerned. The article delves into uncertainty analysis associated with analytical and semi-empirical models of the BLEVE fireball. It could explore how uncertainties in input data, and the choice of a more or less inappropriate model, propagate into the model results. Statistical techniques such as global sensitivity analysis or uncertainty analysis are employed to quantify these uncertainties. In this paper, an attempt is made to evaluate and select reasonable models available in the literature for characterizing fireballs and their consequences. Correlations were analyzed using statistical methods and BLEVE data (experimental and estimated data by correlation) to determine the residual sum of squares (RSS) and average absolute deviation (AAD). Analysis revealed that the Center for Chemical Process Safety (CCPS), the TNO (Netherlands Organization for Applied Scientific Research), and the Gayle model revealed a high degree of satisfaction between the experimental and estimated data through correlation.