Analysis of the behavior of a gas-generating system under runaway conditions in a closed vessel

In the event of a runaway reaction, emergency relief systems (ERS) act as the last line of defense against the vessel explosion. To date, ERS design for scenarios involving the runaway of chemical mixtures that generate gaseous reaction products remains very challenging as it requires the prediction of the gas generation rate under runaway conditions at large scale. Current methodologies for the assessment of the gas generation rate are based on pseudo-adiabatic calorimetric experiments at laboratory scale in which the temperature and pressure profiles resulting from the runaway are used to assess the gas generation rate using the ideal gas law. The gas generation rate needs to be further corrected for the thermal inertia ( φ -factor), of the calorimetric cell used ( φ >1) to predict large scale behavior ( φ ≈1). The work described in this paper focuses on evaluating the capability of this approach to accurately assess the gas generation rate. It uses a dynamic simulator tool based on rigorous thermodynamics to calculate the temperature, pressure, and the composition of each component in a closed vessel containing a gas generating reactive mixture (decomposition of 20 % w/w di-tert butyl peroxide in toluene) under runaway conditions over a range of initial vessel fill levels. The results were analyzed to develop a better understanding of the phenomena that govern the thermal behavior of the mixture, the overall pressurization of the vessel, rate of generation of the gaseous products and its distribution in the liquid and vapor phase of the vessel as the runaway reaction occurs. The simulated pressure, temperature and phase composition data rigorously calculated by the simulator were used to assess and highlight the limitations of using the temperature and pressure in a closed vessel and the ideal gas law to evaluate the gas generation rate. The simulations were also used to evaluate the validity of φ -factor correction methods currently available in the literature.