(393f) A New Mixing Model for Turbulent Reacting Flows Using Hierarchical Parcel Swapping (HiPS)
AIChE Annual Meeting
2017
2017 Annual Meeting
North American Mixing Forum
The Use of CFD in Simulation of Mixing Processes
Tuesday, October 31, 2017 - 2:15pm to 2:36pm
Turbulent reacting flows are ubiquitous in Chemical Engineering and a great
deal of work has been performed to understand and quantify them. This is an
extremely challenging problem, however, due to the complexity of turbulent
flows. Such flows exhibit a wide range of length and time scales. Resolving all
continuum scales is not possible for engineering applications due to large
Reynolds numbers. The computational cost of direct simulations that do resolve
all scales increases with the cube of the Reynolds number. Large Eddy
Simulations (LES) overcome this restriction by only resolving the large scale
eddies, while modeling the subgrid scale mixing and reaction processes. The key
challenge for LES is developing accurate subgrid mixing models, and there are a
number of models of varying success that have been published. These include the
Interaction by Exchange with the Mean (EIM) model, the Euclidian Minimum
Spanning Trees (EMST) model, Curl's model, and the Shadow Position Mixing
Model. The central challenge of these models is representing the mixing
processes in a manner that is computationally efficient while retaining as much
physics as possible. We present a new mixing model termed Hierarchical Parcel
Swapping (HiPS). HiPS was recently introduced and applied to channel flow in
two publications. HiPS consists of a collection of notional fluid parcels with
adjacency defined using a binary tree. The levels of the tree correspond to
geometrically-spaced length scales, the number of which are user-defined. The
mixing and pairing of parcels is done at rates consistent with inertial range
turbulent scaling, and this is the key physical processes enabling the
potential improvement of HiPS over other models. The model has aspects in
common with the Linear Eddy and One Dimensional Turbulence models, but is more
efficient and flexible, and does not directly imply diffusive mixing on a
physical domain. We present an overview of the HiPS model, and then extend the
model to reacting flows. Reaction adds complexity to the mixing process through
variable properties (including dilatation associated with heat release), the
need to carry multiple scalars, which may have varying Schmidt numbers, and the
development of complex reaction-transport structures, such as flames. At this
early stage of HiPS development, we demonstrate the model for simple parallel
competitive reaction mechanisms. We consider variable Schmidt number and
dilatational effects. Results are presented at varying Damkohler and Reynolds
numbers. Product selectively is discussed.
deal of work has been performed to understand and quantify them. This is an
extremely challenging problem, however, due to the complexity of turbulent
flows. Such flows exhibit a wide range of length and time scales. Resolving all
continuum scales is not possible for engineering applications due to large
Reynolds numbers. The computational cost of direct simulations that do resolve
all scales increases with the cube of the Reynolds number. Large Eddy
Simulations (LES) overcome this restriction by only resolving the large scale
eddies, while modeling the subgrid scale mixing and reaction processes. The key
challenge for LES is developing accurate subgrid mixing models, and there are a
number of models of varying success that have been published. These include the
Interaction by Exchange with the Mean (EIM) model, the Euclidian Minimum
Spanning Trees (EMST) model, Curl's model, and the Shadow Position Mixing
Model. The central challenge of these models is representing the mixing
processes in a manner that is computationally efficient while retaining as much
physics as possible. We present a new mixing model termed Hierarchical Parcel
Swapping (HiPS). HiPS was recently introduced and applied to channel flow in
two publications. HiPS consists of a collection of notional fluid parcels with
adjacency defined using a binary tree. The levels of the tree correspond to
geometrically-spaced length scales, the number of which are user-defined. The
mixing and pairing of parcels is done at rates consistent with inertial range
turbulent scaling, and this is the key physical processes enabling the
potential improvement of HiPS over other models. The model has aspects in
common with the Linear Eddy and One Dimensional Turbulence models, but is more
efficient and flexible, and does not directly imply diffusive mixing on a
physical domain. We present an overview of the HiPS model, and then extend the
model to reacting flows. Reaction adds complexity to the mixing process through
variable properties (including dilatation associated with heat release), the
need to carry multiple scalars, which may have varying Schmidt numbers, and the
development of complex reaction-transport structures, such as flames. At this
early stage of HiPS development, we demonstrate the model for simple parallel
competitive reaction mechanisms. We consider variable Schmidt number and
dilatational effects. Results are presented at varying Damkohler and Reynolds
numbers. Product selectively is discussed.