A shock tube experiment and an improved high-temperature diisopropyl ketone model by Bayesian optimization

Abstract

Ignition delay times (IDTs) for di-isopropyl ketone (DIPK) were measured at pressures of 6 and 10 atm, equivalence ratios (ϕ) of 0.5, 1.0, and 2, and over a temperature range of 1100 - 1500 K. In the DIPK sub-model proposed by Barari et al., DIPK⇔IC3H7+IC3H7CO (R2) was erroneously written as DIPK⇔IC3H7+ IC4H7O and the consumption paths of dimethyl ketene (DMK) are missing. A DIPK sub-model was proposed on the basis of Barari model by correcting these errors, reassigning the rate coefficients of the DIPK unimolecular decomposition reactions (R1, R2), H-abstraction reactions by O˙H (R3, R4), fuel radical isomerization reaction (R5), and β-scission decomposition reactions (R5-R10) using the analogy method, and adopting the theoretical calculations of the H-abstraction reactions by H˙, C˙H3, and HO˙2 (R11-R16) provided by Allen et al.. The rate coefficients of R1-R10 were globally optimized by the Bayesian optimization algorithm using the experimental IDTs and profiles of DIPK and DMK. The optimized model well predicts not only the IDTs and profiles of DIPK and DMK, but also the laminar flame speeds and the profiles of many intermediates (propyne, propylene, and methyl ketene) and small species (H2, CO, CH4) during the pyrolysis of DIPK. The calculation using the method of Chen et al. shows that no optimized reaction violates the collision limit. The sensitivity and rate of production analysis were used to discuss the DIPK unimolecular decompositions, DIPK H-abstraction reactions, DIPK radicals isomerization and β-scission reactions, and the optimization of the DIPK sub-model.