Incineration of organophosphorus simulants at high-temperature conditions is an effective method for the destruction of chemical warfare agents. However, uncertainties associated with the behavior of simulant combustion pose a significant challenge to develop reliable destruction strategies. Inverse uncertainty quantification (UQ) is an effective computational approach to quantify the input parameters uncertainty for available measured data from experimental studies of simulant combustion. A non-intrusive Bayesian framework for inverse UQ of chemically reacting flows is desirable for its simplicity, where inverse UQ can be performed either using direct or surrogate modeling techniques. While the direct modeling approach is popular, it tends to be computationally prohibitive when it comes to the investigation of reacting flows in practical systems. To this end, surrogate modeling techniques offer a computationally tractable approach, but they need to be established for such studies. In this study, three surrogate approaches, namely, polynomial chaos expansion (PCE), stochastic collocation (SC), and Gaussian process (GP) are first assessed for their predictive capabilities for inverse UQ of a freely propagating laminar premixed flame by comparing with the results obtained from using the direct Markovian Chain Monte Carlo sampling (MCMC) technique. Based on the results, two surrogate approaches, namely, PCE and GP are then used to conduct inverse UQ of oxidation of diisopropyl methyl phosphonate (DIMP), a representative sarin simulant, in a shock tube setup. The results demonstrated the accuracy and efficiency aspects of surrogate modeling techniques for inverse UQ of simulant combustion.
展开▼
