|
Computational fluid dynamics has long held the promise of producing improved designs for chemical reactors for fast chemical reactions. It is widely recognized that this promise has not been fulfilled. Reaction engineers point out that commercial CFD codes do not adequately deal with turbulence, multiphase flow, and resolution and speed issues. The Council for Chemical Research convened an international workshop in 1996 on CFD for the chemical industries to plan how to improve commercial CFD packages. The software developers themselves were credited with improving the interface for users, especially for the difficult aspects of complex geometric internal flows, and postprocessing. Surprisingly, it was recognized that commercial CFD companies lacked the capital resources to invest in fundamental fluid dynamics and reaction engineering to substantially improve the physical models in their codes. Academic and industrial partnerships would be necessary.Professor Broadkey of Ohio State University, the keynote speaker at the recent IChemE Mixing 6 conference outlined the academic paradigm for improving commercial CFD-based reactor design. Direct numerical simulations (DNS) of the molecular scale phenomena must be carried out to large enough scales that universal statistical relations parameterizing low resolution statistically averaged models; either time-dependent Reynolds averaged equations or large eddy simulations. This orthodoxy has been accepted for many years, yet the proper implementation still eludes the direct numerical simulation community. The received wisdom is that the large scales of turbulent flow may be anisotropic, statistically non-stationary and inhomogeneous, yet somewhere below grid resolution, the motion can be modelled as isotropic, homogeneous and stationary. It is this paradigm I want to shift during this advanced fellowship study.In the early 1980's, several siren articles in the fluid physics literature raised the issue that helicity, a 3-D swirling property, must be a fundamental feature of all scales of turbulent motion. The theory became well developed for incompressible turbulence, but required,, DNS validation. Rogers and Moin (1987) conducted fully resolved DNS of isotropic, homogeneous turbulence apparently contradicting the unresolved DNS findings of Pelz et al. (1986) that helicity played an important role in isotropic turbulence. Further investigation into the importance of helicity at small scales was dropped until recent DNS studies by Oriandi (1997) have shown that when the mean flow is helical, helicity fluctuations play an important role at all scales of the flow. I propose to capitalise on this pivotal result as follows. Conventional turbulence models are derived from shear turbulence experiments and simulations. Since conventional turbulence models are inadequate for predictive purposes for typical reactors used for fast reactions, this proposal posits that treatment of helical dynamics (swirl flow superposed over a mean flow) is the major component that is untreated in these models. It is proposed that direct numerical simulations of helical turbulence be constructed; the results will be used to parameterise moment closure models of helical turbulence, mixing, and reaction for use with conventional CFD solvers. The helical dynamics of coherent structures and the inverse cascade of these structures to large scale, posited theoretically in the fluid physics literature, will be investigated experimentally. Finally, the tools developed will be explored in a number of chemical engineering applications for improved design of combustors, incinerators, mixers, pipelines, and novel reactors.
|