On determination of ignition delay times for the modeling of DDT

PhD thesis.

A quantitative theory of three-dimensional detonations has not been yet developed, thereby researchers and engineers use experimental parameters to describe and categorize the dynamics of detonation. The experimental parameters usually taken into consideration are detonation cell size, ignition delay time, critical initiation energy, critical tube diameter, etc. The need of detonability estimation led to the formulation of correlations between these parameters and the ones which can be computed from ZND theory: ignition delay time behind the leading shock wave, induction zone length and reaction zone width. Nevertheless, these calculated parameters strongly differ depending on the detailed chemical reaction mechanism used. For example, in the case of induction zone length calculations in a stoichiometric ethane-air mixture with the use of GRI-mech 3.0, Konnov 0.5 and Aramco 2.0 noticeably different values of 1.9 mm, 1.0 mm and 0.85 mm, respectively, are produced. Ignition delay time itself is a key physical property controlling the transition to detonation and the propagation of detonation. Moreover, it is used for indirect mechanisms testing.

In order to enable more reliable numerical investigations of the transition to detonation and detonation itself, the analysis of detailed reaction mechanisms’ performance in the reproduction of ignition delay times was carried out. The analysis was devoted to C1–C4 hydrocarbons, as they form a group of globally used fuels and, moreover, their chemistry is important for non-aromatic species at high temperatures. An extensive literature review was done in search of ignition delay times from shock tube experiments as they most closely represent the conditions which appear in detonation. The accumulated ignition delay times were simulated with fifteen well established kinetic mechanisms and then the performance of the mechanisms was compared. The Author presented the comparison at different levels of generality, hence providing guidance as to which mechanism performs best across all investigated conditions, which differed strongly between mixtures. Newly developed techniques were used to assess the ignition delay time prediction capabilities of the detailed reaction mechanisms. Box-whisker plots summarized the statistical dispersion of the applied error function across pressure, temperature and equivalence ratio. Additional heat maps of the error function on pressure-temperature plots provide more insight into the performance of mechanisms and highlight areas of uncertainty and possible improvement.

Additionally, ignition delay time models based on a Deep Neural Network (DNN) were proposed. Two models were developed: one for argon diluted mixtures and a second for nitrogen. The DNN models were highly accurate, and comparable with the best mechanisms. And most importantly, Ignition Delay Time (IDT) computations using the DNN models were incomparably much shorter then the best mechanisms. The developed models might be a reasonable choice for CFD simulations of deflagration to detonation transition. The models can be easily extended to new experimental data as published.