[Fluent Inc. Logo] return to home search
next up previous contents index

18.3.9 Using ISAT Efficiently

Efficient use of ISAT requires thoughtful control. What follows are some detailed recommendations concerning the achievement of this goal.

figure   

The numerical error in the ISAT table is controlled by the ISAT Error Tolerance, which has a default value of 0.001. This value is relatively large, which allows faster convergence times. However, once the solution has converged, it is important to reduce this ISAT Error Tolerance and re-converge. This process should be repeated until the species that you are interested in modeling are unchanged. Note that as the error tolerance is decreased, the memory and time requirements to build the ISAT table will increase substantially. There is a large performance penalty in specifying an error tolerance smaller than is needed to achieve acceptable accuracy, and the error tolerance should be decreased gradually and judiciously.

figure   

Once the ISAT table is full, all queries that cannot be retrieved are directly integrated. Since retrieves are much quicker than direct integrations, larger ISAT tables are faster. Hence, you should set the ISAT Max. Storage to a large fraction of the available memory on your computer.

During the initial iterations, before a steady-state solution is attained, transient composition states occur that are not present in the steady-state solution. For example, you might patch a high temperature region in a cold fuel-air mixing zone to ignite the flame, whereas the converged solution never has hot reactants without products. Since all states that are realized in the simulation are tabulated in ISAT, these initial mappings are wasteful of memory, and can degrade ISAT performance. If the table fills the allocated memory and contains entries from an initial transient that are no longer accessed, it may be beneficial to purge the ISAT table. This is achieved by either clearing it in the Integration Parameters panel, or saving your case and data files, exiting FLUENT, then restarting FLUENT and reading in the case and data.

The optimum ISAT table is achieved when a new table is started from the converged FLUENT solution. If you are simulating a range of parametric cases where the flame changes gradually, it is likely beneficial to create such an optimum table for the first case, and then save it.

File $\rightarrow$ Write $\rightarrow$ ISAT Table...

Subsequent runs can start from this table by reading it into memory.

File $\rightarrow$ Read $\rightarrow$ ISAT Table...

See Section  18.3.10 for information about reading and writing ISAT tables in parallel.

ISAT efficiency may be increased by employing multiple tables (also called trees). Increasing the number of trees has the effect of decreasing the table size and hence the time needed to build the table, but increasing the retrieve time. Hence, for long simulations with simple chemistry, a small number of tables may be optimal. On the other hand, for short simulations with complex chemistry, computers with limited memory, or simulations with a small ISAT error tolerance, a large number of trees is likely optimal since most of the CPU time is spent building the table.

From experience, ISAT performs very well on premixed turbulent flames, where the range of composition states are smaller than in non-premixed flames. ISAT performance degrades in flames with large time-scales, where more work is required in the ODE integrator.


next up previous contents index Previous: 18.3.8 Monitoring ISAT
Up: 18.3 Steps for Using
Next: 18.3.10 Reading and Writing
© Fluent Inc. 2006-09-20