  The problematic associated to the performance of unsteady simulations is well known, and the main consequence is large computation times to get an accurate solution. To avoid the time step limit from the smaller cells, allowing acceleration techniques such as multigrid and giving the possibility to choose the convergence accurateness depending on the error required by the project, the dual time step technique is an industry standard in unsteady simulations. Despite of that, the convergence time is still a big issue.

  The dual time step technique itself has its own critical points, as can be the election of the dual time step iterations (or its convergence level) as it impacts on both, the accuracy (where different requirements are possible depending on the zones/features of interest inside the same simulation) and the convergence time. The best solution means a compromise that is advisable to be conservative to avoid having a worthless computation.

  A special case of unsteady simulations is the time periodic simulation, of paramount importance in main industrial applications as can be the rotating turbomachinery or the aero-engines. Some examples could be the vibration problem, the forced response simulation or the inlet distortion problem just to mention a few.

  In unsteady periodic cases some techniques have arisen to improve further this behavior, from which, the harmonic balance method can be highlighted. If we start from a typical U-RANS code with dual time step method, the harmonic balance technique requires a considerable effort in implementation, and a vast memory requirement as all the time instants are computed together.

  A more straightforward approach is to re-use the solution computed in the previous cycle as initial solution for the current cycle.

  This approximation has some clear advantages over the plain dual time step method, and even over the harmonic balance method, but it has also its drawbacks. A brief review of all of them will be presented.

  To correctly assess the performance of this approach is mandatory to do an optimization of its usage. The main points that will be addressed in the paper are:

-          - How to introduce the technique in the simulation to avoid the penalty in the first cycles when the convergence is still poor.

-          - Optimum number of dual time steps for each physical time step motivated in terms of the different boundary conditions types used and the calculation times of your scheme.

-          - Possible benefits of a specific implementation depending on the methodologies used. In our case a specific implementation for the double-passage methodology will be reviewed.

-          - A cheap re-starting procedure, as the straightforward re-starting means to deal with very big files for standard turbomachinery cases.

  The optimized version will be evaluated and compared qualitatively and quantitatively with the plain dual time step technique and with the harmonic balance methodology for different simulation conditions in typical turbomachinery cases.

  The evaluation and performance of the technique depends on the specific implementation used. Therefore, for this evaluation to be representative of the turbomachinery industry standards, the optimization and comparison will be done using a state-of-the-art in-house code with full parallel GPUs and CPUs computation featuring the main acceleration techniques such as overlapping communications, multigrid, implicit scheme, or Block-Jacobi among others.


Quantitative Evaluation and Optimization of Feeding-Back Previous Cycles Information for CFD Periodic Convergence

Division / Sponsor:
International Gas Turbine Institute (IGTI)

Paper Proceedings: https://asmedigitalcollection.asme.org/GT/GT2020/volume/84058

Turbo Expo - Virtual Conference

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.