19.2.3. Using Dynamic Mechanism Reduction Effectively

As described in Dynamic Mechanism Reduction in the Fluent Theory Guide, mechanism reduction is performed at each cell or particle, once per time step (for transient simulations) or per flow iteration (for steady-state simulations). It is assumed that the reduced mechanism created for the starting conditions will remain valid for the entire transport time step over which the chemistry ODE is integrated. If the chemistry integration time interval is very large, the chemical state can change significantly, degrading the accuracy of the reduced mechanism. In some cases (particularly transient simulations such as ignition) it may be necessary to enforce smaller flow time steps in order to effectively use Dynamic Mechanism Reduction. In this way the reduced mechanisms are updated more frequently to match the changing chemical states.

Note that Dynamic Mechanism Reduction works best when there are significant regions in a computational domain and/or times during a simulation with relatively low chemical activity (for example, low temperature, low mixing of fuel/oxidizer, low concentrations of reactive species, and so on). Consider a simple opposed flow diffusion flame problem with pure fuel as one stream and pure oxygen as the other stream. Large mechanisms (which, depending on the error tolerance, may be close to the full mechanism size) will likely be used in the mixing region where the flame is located, while smaller mechanisms would be used elsewhere in the domain. You should not expect much speed-up if a very large fraction of your grid cells (for example, 90%) are in the flame region.

In steady-state simulations, you can improve the solution time by first running to convergence with a larger error tolerance , and then restarting the simulation repeatedly, gradually decreasing the error tolerance until you reach the desired accuracy level. For example, if you first converge the solution with the error tolerance , then iterate the solution further to convergence with the lower error tolerance, , and then with the desired error tolerance , you will generally be able to achieve faster convergence than iterating from initial conditions with the error tolerance .


Important:  As described in Mesh Partitioning and Load Balancing, you can select portions of the simulation to consider when performing dynamic load balancing in multiprocessor simulations. When using the Dynamic Mechanism Reduction along with ISAT, it is recommended that you select the ISAT option in the Weighting tab of the Partitioning and Load Balancing Dialog Box when solving stiff chemistry. If ISAT is selected, the time required for solving chemistry will be factored in when assigning the computational cells to each available processor; thus, more resources can be allocated for the cells with the larger mechanisms.



Release 18.2 - © ANSYS, Inc. All rights reserved.