Analysis of constraining a chemical kinetic mechanism using hybrid response surface networks

This paper presents an artificial intelligence based analysis for fast development of jet fuel chemical kinetic mechanisms through optimal constraining of parameters. The need to generate chemical kinetic mechanisms rapidly with less uncertainty is a critical requirement to efficiently assess newly introduced sustainable aviation fuels. To overcome the under-constrained nature of the optimization process with readily available but limited data, a hybrid response surface technique was developed to rapidly repeat this fitting process and provide a distribution of solutions. Through this approach, not only can the uncertainties be quantified, but the distribution of solutions can also be used to identify additional data that can reduce those uncertainties. Since extensive experimental measurements can be costly, the ability to identify a limited set of additional data can be of great importance. In this study, we demonstrate this approach using a Jet-A chemical kinetic mechanism that was optimized towards new experimental ignition delay measurements using the hybrid response surface network approach. This mechanism is shown to produce well-constrained ignition delay predictions at conditions of other ignition delay data in the literature, but significant uncertainty in chemical species were observed when compared to shock tube pyrolysis species measurements. By constraining one key chemical species from the response surface analysis, it is shown that most of the uncertainty in the remaining species was also reduced and utilizing two species provided extremely strong constraints. The results suggest that adding even a single species measurement to the mechanism development would significantly reduce the uncertainties in the optimization process.