Computational AeroAcoustics: how they got it wrong (repeatedly)

Add to your list(s) Download to your calendar using vCal

• Dr Ed Brambley, University of Warwick, UK
• Tuesday 23 October 2018, 10:30-11:30
• CMS, MR15.

If you have a question about this talk, please contact Matthew Priddin.

Despite the best efforts of mathematicians, aircraft still do not use infinite straight engine-ducts with thin zero-angle-of-attack blades, and so unfortunately good Computational AeroAcoustics (CAA) simulations remain necessary in order to calculate the sound produced in the actual geometry and CFD -computed flows within aircraft engines. CAA simulations push the limits of current computing power, and so try to be as efficient as possible, while also being precariously balanced on the knife edge between being overly dissipative and being unstable.

Most CAA simulations today use optimized Dispersion Relation Preserving (DRP) spatial derivatives, which claim to be accurate down to 5 points per wavelength. They aren’t.

Most CAA researchers are aware there are problems with numerical instability near boundaries due to the use of asymmetric stencils. There are a number of papers proposing optimized spatial derivatives near boundaries that are supposedly stable. They aren’t.

Most CAA simulations use optimized Runge-Kutta time stepping to evolve the solution forwards in time, (most commonly LDDRK56 , Low Dispersion and Dissipation Runge-Kutta 5stage/6stage), which are claimed to be optimized for wave propagation. They aren’t.

There is some theory of finite difference schemes that are provably stable, and which do work near boundaries. These are called Summation By Parts (SBP) schemes, with the important property that they have a discretized version of Integration By Parts (IBP), which could be used for CAA simulations. They aren’t.

This talk will explain some of the maths behind the claims above, and some recent work to bring adapt recent best practice (SBP) finite differences to CAA simulations.

This talk is part of the Waves Group (DAMTP) series.

This talk is included in these lists:

• CMS, MR15
• Waves Group (DAMTP)

Note that ex-directory lists are not shown.