How Simulation of Novel Compressor Casings Got 28X Faster Through Meshing

With Pointwise meshes and ADSCFD GPU accelerated CFD, simulation times are reduced by nearly 30X; reducing the turnaround time for a simulation that would take a month with traditional CPU technology down to a single day. This reduction in turnaround time enables such an analysis to be moved from something exotic and reserved for research to something useable for a designer within a typical commercial design cycle. The Pointwise capability to create hybrid unstructured grids, where surface boundaries can be meshed with a fully structured surface mesh but then combined with a fully unstructured volume mesh, gives the best of both worlds for these simulations: the speed and accuracy available to structured grid sliding mesh methods and the ease of mesh generation typically only available with fully unstructured grids.

For turbomachinery compressor designers, a delicate balance exists between system performance and system stability. It can be the deciding factor on whether an engine program is successful or not, which affects billions of dollars in revenue for large OEMs. As a designer moves towards higher compression ratio systems, managing stall margin becomes a critical factor. Optimizing airfoil design to maximize both performance and stall margin is a mature area, so designers need to look outside of traditional airfoil design to make further gains. One very promising area that is a current area of interest is non-axisymmetric endwall contouring. By contouring the endwalls, with a specific focus on the outer casing, the designer has additional tools at their disposal to manipulate the unsteady loading seen by the airfoils which allow for improvement to performance and stall margin at various points within the compressor map.

Novel casing treatment geometries have been shown experimentally to improve stall margin. However, they have not been widely adopted by industry due to the analysis complexity and the requirement of large scale unsteady simulations to attain meaningful results. Using current technologies, this analysis often requires hundreds of CPUs and weeks to complete just one analysis, precluding its use during the design cycle. The problem is also complicated by requiring extremely high-quality transfer of data along the sliding mesh interface. Traditional multi-block structured grids facilitate accurate and computationally inexpensive interpolation methods but are extremely difficult to generate for complicated endwall treatment designs. Unstructured grids are much easier to generate but require more complex and computationally expensive interpolation approaches to achieve similar levels of accuracy.

The details of this simulation are included in the case study 28X Increase in Simulation Speed by Using Pointwise Meshes on our website. Read it today and then contact us to see how this type of meshing flexibility can work for you.

This entry was posted in Applications, Software and tagged , , , , , . Bookmark the permalink.

Leave a Reply