This paper presents a numerical study of the effects of the methods used for calibrating the Chemical-Diffusive Model (CDM) with regard to propagation, diffraction, and reinitiation of detonations in stoichiometric C2H4- 2 H 4- air mixtures at near atmospheric pressures. The CDM is a model for the chemical reaction and diffusive transport processes designed to be coupled with the compressible, reactive Navier-Stokes equations. The CDM parameters are derived using machine learning methods on large datasets of detailed chemical kinetic mechanisms, theoretical results, and experiments. When optimized, using the CDM reproduces correct flame and detonation properties, including the detonation cell size. The inputs to the optimization procedure can come from detailed chemical kinetics or experiments. When the CDM is constrained to reproduce the correct detonation cell size (A-optimized), the induction zone of the ZND profile is larger compared to when the CDM is optimized using inputs solely from detailed kinetics. The detonation cell sizes are in better agreement with experimental data when the A-optimized CDM is used. The diffraction and reinitiation are studied by numerical simulations of detonation diffraction over an obstacle of large blockage (70-98 percent) in a half- channel. The shock and reaction front are decoupled on diffraction, and reinitiation occurs following formation of hotspots in the decoupled region. The results identify two regimes depending on the extent of blockage. For lower blockage (70-80 percent), the A-optimized CDM has a greater propensity for reinitiation by the direct ignition mechanism. Hotspots are formed in the reactants during diffraction or immediately on shock reflection. The CDM optimized using inputs from detailed kinetics requires multiple shock collisions for hotspot formation. For larger blockage (80-90 percent), the hotspot formation mechanisms are independent of the CDM calibration strategy used. The results agree with prior literature on the reinitiation mechanisms of detonations after diffraction. The results indicate that it may be important to account for the variability in cell sizes when developing simplified chemical reaction models for hydrocarbon mixtures.
